HIPERFIT

Talk on Extracting Functional Programs from Coq, in Coq by Danil Annenkov

2021-09-15T00:00:00+00:00

Talk: Extracting functional programs from Coq, in Coq

Presenter: Danil Annenkov, Aarhus University

Time: Wednesday, September 15, 2021, 11:15-12:00.

Place: PLTC meeting room (HCØ-01-0-029), Universitetsparken 5.

Abstract

Many proof assistants offer a possibility for obtaining code in a conventional functional programming language from formalised developments. This functionality is called code extraction. The Coq proof assistant supports extraction to OCaml, Haskell, and Scheme out of the box. However, the extraction functionality itself is not verified. Moreover, there are many interesting target languages not covered by the standard extraction. We address these challenges by developing an extraction pipeline entirely in Coq by extending the MetaCoq verified erasure procedure. We also develop pretty-printing functionality to new target languages: Elm, Rust, Liquidity, and CameLIGO. In total, this gives us a way to write dependently typed programs in Coq, verify, and extract them to several target languages while retaining a small trusted computing base.

In this talk, we present the current status of our development, motivate the transformations we currently have in the pipeline, and discuss further opportunities for extending the pipeline with new target languages.

Joint work with Mikkel Milo, Jakob Botsch Nielsen, and Bas Spitters. arXiv preprint: https://arxiv.org/abs/2108.02995

Bio

Danil Annenkov is a postdoc researcher at the Concordium Blockchain Research Center, Aarhus University working on formal verification of smart contracts. His research areas include programming language semantics, formal verification, proof assistants, and type theory. Danil Annenkov received his PhD degree in Computer Science from the University of Copenhagen, DIKU in 2018 under the supervision of Martin Elsman. After receiving his PhD degree, Annenkov was a postdoc researcher at INRIA Nantes, France, where he worked on extending the Coq proof assistant with new reasoning principles.

Host: Martin Elsman

MSc Thesis defence on Sum types in Futhark

2020-01-06T00:00:00+00:00

MSc thesis defense by Robert Schenck.

Time: Monday, January 6, 2020, 13:00-14:00.

Place: PLTC meeting room (HCØ-01-0-029), Universitetsparken 5.

Abstract

While a common staple of many modern functional programming languages, sum types are an unusual feature in specialized, computeoriented languages. We advocate that sum types have merit even in this restricted setting: providing increased safety and better abstraction at manageable cost.

This thesis explores adding sum types to Futhark—a compute-oriented, purely functional, data-parallel array programming language. The work includes a theoretical development of sum types in model type system closely related to Futhark’s type system, implementation details for adding sum types to the Futhark compiler, and an analysis of the implementation—including suggestions to improve performance.

Supervisors: Martin Elsman and Troels Henriksen (co-supervisor)

External examiner: Mads Rosendahl, RUC

ConCert: A Smart Contract Certification Framework in Coq

2019-12-12T00:00:00+00:00

HIPERFIT Talk: “ConCert: A Smart Contract Certification Framework in Coq”

Presenter: Danil Annenkov, Post-doctorial Researcher, the Concordium Blockchain Research Center, Aarhus University

Time: Thursday, December 12, 2019, 09:30-10:15

Place: PLTC Meeting Room (HCØ-01-0-029), HCØ, Universitetsparken 5, University of Copenhagen.

Abstract:

Modern smart contract languages for blockchains tend to employ the functional programming paradigm. This fact makes them perfect targets for embedding into proof assistants allowing for convenient reasoning about their properties. But we are also interested in the meta-theory of smart contract languages and would like to have strong correctness guarantees for the embedding. We introduce ConCert: a framework allowing for both deep (AST) and shallow (Coq functions) embeddings of a smart contract language along with the soundness theorem connecting the two embeddings. The framework also features an execution model that enables reasoning about safety and temporal properties of interacting smart contracts. As an application of the developed framework, we show how to verify programs in Acorn - functional smart contract for the Concordium blockchain.

The developed techniques could be applied to the verification of programs in various (not necessarily smart contract) functional languages.

Joint work with Bas Spitters and Jakob Botsch Nielsen.

Biography

Danil Annenkov is a postdoc researcher at the Concordium Blockchain Research Center, Aarhus University working on formal verification of smart contracts. His research areas include programming language semantics, formal verification, proof assistants and type theory. Danil Annenkov received his PhD degree in Computer Science from the University of Copenhagen in 2018. After receiving his PhD degree, Annenkov was a postdoc researcher at INRIA Nantes, France, where he worked on extending the Coq proof assistant with new reasoning principles.

Host: Martin Elsman

PhD Defence: Danil Annenkov

2018-01-25T00:00:00+00:00

Title: Adventures in Formalisation: Financial Contracts, Modules, and Two-Level Type Theory

Thesis: Revised April 16, 2018. pdf.

Presenter: Danil Annenkov, PhD. Student, DIKU/University of Copenhagen

Time: Thursday, January 25, 2018, 13:15

Place: HCØ, Auditorium 5, Universitetsparken 5, 2100 Copenhagen Ø

Committee:

Associate Professor Andrzej Filinski (Chairman), Department of Computer Science, University of Copenhagen, Denmark
Professor Lars Birkedal, Department of Computer Science, Aarhus University, Denmark
Professor Andrew Pitts, University of Cambridge, United Kingdom

Abstract

We present three projects concerned with applications of certified programming techniques and proof assistants in the area of programming language theory and mathematics.

The first project is about a certified compilation technique for a domain-specific programming language for financial contracts (the CL language). The code in CL is translated into a simple expression language well-suited for integration with software components implementing Monte Carlo simulation techniques (pricing engines). The compilation procedure is accompanied with formal proofs of correctness carried out in the Coq proof assistant.

The second project presents a number of techniques that allow for formal reasoning with nested and mutually inductive structures built up from finite maps and sets (also called semantic objects). The techniques, which build on the theory of nominal sets combined with the ability to work with multiple isomorphic representations of finite maps, make it possible to give a formal treatment, in Coq, of a higher-order module system, including the ability to eliminate entirely, at compile time, abstraction barriers introduced by the module system. The development is based on earlier work on static interpretation of modules and provides the foundation for a higher-order module language for Futhark, an optimising compiler targeting data-parallel architectures, such as GPGPUs.

The third project presents an implementation of two-level type theory, a version of Martin-Lof type theory with two equality types: the first acts as the usual equality of homotopy type theory, while the second allows us to reason about strict equality. In this system, we can formalise results of partially meta-theoretic nature. We develop and explore in details how two-level type theory can be implemented in a proof assistant, providing a prototype implementation in the proof assistant Lean. We demonstrate an application of two-level type theory by developing some results on the theory of inverse diagrams using our Lean implementation.

Biography

Danil Annenkov is a PhD student at DIKU/University of Copenhagen under supervision of Associate Professor Martin Elsman, DIKU.

Host: DIKU and HIPERFIT

HIPERFIT Summit

2017-11-17T00:00:00+00:00

Computation and Finance for the 21st Century

On November 17, 2017, in the afternoon, HIPERFIT is hosting a 3 hour summit on Computation and Finance for the 21st Century. This summit will give highlights, in the form of a potpourri, of the technical outcomes of HIPERFIT (looking backwards) followed by presentations of insights into the needs of the future, given by a number of prominent guests who, in one way or the other, play a role in forming the future of the it-finance sector.

This event, together with the technical workshop on November 16, 2017, marks the end of the HIPERFIT research center, as we know it. We will, continue to operate under the name of HIPERFIT, but funding for a number of the research projects are ending, which, of course, opens up avenues to new opportunities.

Registration

The Venue

DIKU, Aud 3, HCØ, Universitetsparken 5, DK-2100 Copenhagen.

Schedule November 17, 2017 – Summit

13:00

Welcome (Martin Elsman, DIKU)

13:05

HIPERFIT highlights---a summary of project results (HIPERFIT Team)

Predicting Brexit Effects
Functional Programming on Speed
100x Speedup in One Line
Financial Contracts in the Wild--no Brokers Added
...

14:00

Coffee Break

14:20

Putting Computation and Finance to Work

14:20

Enterprise fintech opportunities (Anders Kirkeby, VP Enterprise Architecture, SimCorp)

14:45

Putting parallel processing at the fingertips of the domain expert (Gitte Christensen, CEO, Dyalog)

15:10

I krydsfeltet af FinTech og AI. Hvordan holder jeg mig, og min organisation, på forkanten af nye teknologier indenfor FinTech og AI.(Mikael Munck, CEO 2021.AI, Chairman, Copenhagen Fintech)

15:35

Strategic Research and Innovation -- how and who pays?

15:35

Introduction (Fritz Henglein, DIKU)

15:45

Panel Discussion

Gitte Christensen, CEO, Dyalog
Anders Kirkeby, VP Enterprise Architecture, SimCorp
Anette Broløs, Independent Fintech Analyst, Broløs Consult
Kourosh Marjani Rasmussen, Associate Professor, DTU
Mikael Munck, CEO 2021.AI, Chairman, Copenhagen Fintech

16:10

Closing Remarks (Fritz Henglein, DIKU)

16:15

Networking, Snacks, and Wine

(Updates and Changes to the program may occur.)

Registration

HIPERFIT Workshop

2017-11-16T00:00:00+00:00

HIPERFIT Technical Workshop

On November 16, 2017, HIPERFIT is hosting a workshop for HIPERFIT partners, faculty, and other people interested in technical topics related to HIPERFIT activities. The presentations will include presentations both by researchers external to HIPERFIT and by HIPERFIT researchers themselves.

This event, together with the HIPERFIT Summit held on November 17, 2017, marks the end of the HIPERFIT research center, as we know it. We will, continue to operate under the name of HIPERFIT, but funding for a number of the research projects are ending, which, of course, opens up avenues to new opportunities.

Registration

The Venue

DIKU, Aud 3, HCØ, Universitetsparken 5, DK-2100 Copenhagen.

Schedule November 16, 2017

08:55: Welcome (Fritz Henglein, DIKU)
09:00: Partial Evaluation and Formalisation (Chair: Cosmin Oancea)
09:00: Object-oriented partial evaluation and the expression problem (Alan Mycroft, U. of Cambridge)
09:30: Nominal techniques in Coq (Danil Annenkov, DIKU)
10:00: Coffee Break
10:30: Finance (Chair: Mogens Steffensen)
10:30: Special FX (Rolf Poulsen, MATH)
11:00: Household finance (Maj-Britt Nordfang, MATH)
11:30: Modelling stochastic volatility in forward markets (Fred Esben Benth, U. of Oslo)
12:00: Lunch
13:00: Blockchain Technology (Chair: Martin Elsman)
13:00: Automated execution of financial contracts on blockchain (Omri Ross, DIKU)
13:30: Mini Coffee Break
13:45: Streaming and Life-Insurance on GPUs (Chair: Ken Friis Larsen)
13:45: Streaming data-parallelism (Andrzej Filinski, DIKU)
14:15: Stream fusion: from staging to ahead-of-time compilation through Scala Native (Aggelos Biboudis, École Polytechnique Fédérale de Lausanne)
14:45: Domain-specific languages and GPGPUs in life insurance and pensions (Peter Sestoft, ITU)
15:15: Coffee Break
15:45: Data-parallel Programming (Chair: Fritz Henglein)
15:45: Design and implementation of the Futhark programming language (Troels Henriksen, DIKU)
16:15: Futhark: Challenges and Future Research Directions (Cosmin Oancea, DIKU)
16:45: APL on GPUs--a progress report with a touch of machine learning (Martin Elsman, DIKU)
17:15: Closing Remarks (Fritz Henglein, DIKU)
17:20: Chat and Snacks

(Updates and Changes to the program may occur.)

Registration

Abstracts for Talks on November 16, 2017

09:00-10:00 Partial Evaluation and Formalisation

Object-oriented partial evaluation and the expression problem (Alan Mycroft, U. of Cambridge)

Object-oriented and functional languages rest on differing foundations and the ‘expression problem’ captures the idea that their inter-translation (‘transpilation’) is non-modular; this suggests that automated transpilation is likely to be ad-hoc or clumsy.

We show this to be false. By creating a partial evaluator for a pure Java-like language with a focus on class specialisation and applying it to an interpreter for an ML-like language we show that the seemingly idiomatic transformation from sum types to a class hierarchy arises naturally.

In common with many partial-evaluation tools, our initial translation is untyped, but we also show how slightly richer analysis can produce a typed translation which is close to what a human might produce.

Nominal techniques in Coq (Danil Annenkov, DIKU)

Variables and binding are ubiquitous concepts in programming language research. In pen-and-paper proofs about programming language semantics one usually applies Barendregt’s informal variable convention: all bound variables can be chosen to be different from the free variables. This convention allows for reasoning in the presence of restrictions on freshness of bound variables involved in a proof. It is essential to give a formal justification for the variable convention at least for two reasons. First, applying the convention may lead to unsound reasoning. Second, it is impossible to formalise proofs, which use the informal variable convention. In this talk we present a theory of nominal sets (Gabbay, Pitts 2002) and techniques for dealing with bound variables based on this theory. We discuss applications of these techniques to development of proofs in the Coq proof assistant, show examples in simply-typed lambda calculus, and outline some applications to ongoing work on module system formalisation.

10:30-12:00 Finance

Special FX (Rolf Poulsen, MATH)

Special events call for special techniques – also in quantitative finance. In this talk I look at two such fairly recent events from the world of exchange rates (FX): The Swiss Floor and Brexit/Trump.

Household finance (Maj-Britt Nordfang, MATH)

What is the optimal mortgage and how are pension savings optimally invested? In this presentation I will discuss the design of financial contracts in relation to the preferences of individual household consumers.

Modelling stochastic volatility in forward markets (Fred Esben Benth, U. of Oslo)

TBA.

13:00-13:30 Blockchain Technology

Automated execution of financial contracts on blockchain (Omri Ross, DIKU)

In this presentation we outline how certain financial contracts can be managed and executed automatically on the Ethereum block-chain system. The system is based on a domain-specific language for financial contracts that is capable of expressing complex multi-party derivatives and is conducive to automated execution. We propose an architecture for separating contractual terms from contract execution. A contract engine encapsulates the syntax and semantics of financial contracts without actively performing contractual actions; such actions are handled by user-definable contract managers that administer strategies for the execution of contracts. Hosting contracts and contract managers on a distributed ledger, side-by-side with digital assets, facilitates automated settlement of commitments without the need for an intermediary. We discuss how the proposed technology may change the way financial institutions, regulators, and individuals interact in a financial system based on distributed ledgers.

13:45-15:15 Streaming and Life-Insurance on GPUs

Streaming data-parallelism (Andrzej Filinski, DIKU)

TBA.

Stream fusion: from staging to ahead-of-time compilation through Scala Native (Aggelos Biboudis, École Polytechnique Fédérale de Lausanne)

Stream processing is mainstream (again): Widely-used stream libraries are now available for virtually all modern OO and functional languages, from Java to C# to Scala to OCaml to Haskell. Yet expressivity and performance are still lacking. In this talk we present two topics. Firstly, we present the strymonas library, the first approach that represents the full generality of stream processing and eliminates overheads, via the use of staging. It is based on an unusually rich semantic model of stream interaction. We support any combination of zipping, nesting (or flat-mapping), sub-ranging, filtering, mapping-of finite or infinite streams. Secondly, gaining inspiration from Futhark we present our vision for Scala Native enhanced with capabilities to offload computations to GPGPU supporting fusion and fission. Scala Native is the ahead-of-time compiler for Scala developed at EPFL and in this talk we will introduce a possible common ground for ideas to flow between the two worlds.

Domain-specific languages and GPGPUs in life insurance and pensions (Peter Sestoft, ITU)

The pension and life insurance industry works with a very long-term financial contracts where payments depend on the life state of the insured individual(s). We present a domain-specific language called the Actulus Modelling Language for Products (AML-P) for describing the payment streams of life insurance and pension contracts, and the envisioned uses of this language, including the computation of net present value, future cashflows and so on. The latter quantities can be estimated by numerical solution of (Thiele) differential equations, and we highlight the computational flexibility obtained by using a domain-specific language to separate specification and implementations. This reflects work done in the Actulus project 2011-2016, a collaboration between Edlund A/S, Copenhagen University and the IT University of Copenhagen. We also sketch some ideas for further development.

15:45-17:15 Data-parallel Programming

Design and implementation of the Futhark programming language (Troels Henriksen, DIKU)

Futhark is a data-parallel programming language under development at HIPERFIT. The language supports nested data-parallelism and features an optimising compiler under rabit development. This presentation gives a status update on recent Futhark development, with a particular focus on generation of efficient OpenCL GPU kernels and the steps automatically taken by the Futhark compiler to ensure fusion and efficient memory accesses. In the presentation, we will discuss various tradeoffs of the code generation, and how a new versioning technique can defer to runtime the decision between different parallelisation strategies.

Futhark: Challenges and Future Research Directions (Cosmin Oancea, DIKU)

The study of programming-model technology that can effectively utilize modern hardware is an open, hard, and fundamental topic of research. The previous talk has introduced such a data-parallel language – in summary, Futhark has already brought a number of research contributions and has already demonstrated its potential for efficiently utilizing heterogeneous GPGPU hardware on a number of applications, but one could remark that so far, Futhark’s design has followed a rather conventional path. We believe that currently, Futhark merely provides a solid foundation for studying and experimenting with novel and aggressive techniques in the language and compiler-construction fields.

This talk outlines several such research opportunities: First, we use the motivating example of trinomial pricing to show how locality of reference can be optimised by using inspector-executor techniques coupled with a more aggressive flattening of nested parallelism. Second, we demonstrate how the large-memory footprint and the significant copying overhead inherent to purely-functional approaches can be optimised by combining ideas from region-based memory management and register allocation (the later is lifted to operate on arrays rather than scalars). Finally, we discuss language extensions for (automatically) supporting module systems, automatic differentiation, and relational algebra.

APL on GPUs–a progress report with a touch of machine learning (Martin Elsman, DIKU)

We give a status report of the APL->TAIL->Futhark compiler, which compiles a subset of APL into code executable on GPUs. The compiler handles quite a number of APL functions and operators and is, for the subset of APL it supports, highly compatible with code written for Dyalog APL. Besides reporting on the performance of a number of APL benchmarks, we demonstrate, by example, how the APL compiler tool chain can be used to teach efficiently a neural network to recognise handwritten digits.

PhD Defence: Troels Henriksen

2017-11-15T00:00:00+00:00

Title: Design and Implementation of the Futhark Programming Language

Thesis: To be published later (For an electronic copy of the thesis, please contact ‘PhDadmin@di.ku.dk’)

Presenter: Troels Henriksen, PhD. Student, DIKU/University of Copenhagen

Time: Wednesday, November 15, 2017, 13:15

Place: HCØ, Auditorium 6, Universitetsparken 5, 2100 Copenhagen Ø

Committee:

Associate Professor Torben Mogensen (Chairman), Department of Computer Science, University of Copenhagen, Denmark
Professor Mary Sheeran, Chalmers University of Technology, Sweden
Professor Alan Mycroft, University of Cambridge, United Kingdom

Abstract

In this thesis we describe the design and implementation of Futhark, a small data-parallel purely functional array language that offers a machine-neutral programming model, and an optimising compiler that generates efficient OpenCL code for GPUs. The overall philosophy is based on seeking a middle ground between functional and imperative approaches. The specific contributions are as follows:

First, we present a moderate flattening transformation aimed at enhancing the degree of parallelism, which is capable of exploiting easily accessible parallelism. Excess parallelism is efficiently sequentialised, while keeping access patterns intact, which then permits further locality-of-reference optimisations. We demonstrate this capability by showing instances of automatic loop tiling, as well as optimising memory access patterns.

Second, to support the flattening transformation, we present a lightweight system of size-dependent types that enables the compiler to reason symbolically about the size of arrays in the program, and that reuses general-purpose compiler optimisations to infer relationships between sizes.

Third, we furnish Futhark with novel parallel combinators capable of expressing efficient sequentialisation of excess parallelism, as well as their fusion rules.

Fourth, in order to express efficient programmer-written sequential code inside parallel constructs, we introduce support for safe in-place updates, with type system support to ensure referential transparency and equational reasoning.

Fifth, we perform an evaluation on 21 benchmarks that demonstrates the impact of the language and compiler features, and shows application-level performance that is in many cases competitive with hand-written GPU code.

Sixth, we make the Futhark compiler freely available with full source code and documentation, to serve both as a basis for further research into functional array programming, and as a useful tool for parallel programming in practise.

Biography

Troels Henriksen is a PhD student at DIKU/University of Copenhagen under supervision of Professor Fritz Henglein, DIKU and (co-supervisor) Assistant Professor Cosmin Oancea, DIKU.

Host: DIKU and HIPERFIT

PhD Defence: Martin Dybdal

2017-08-09T00:00:00+00:00

Title: Array abstractions for GPU Programming

Thesis: To be published later (For an electronic copy of the thesis, please contact ‘PhDadmin@di.ku.dk’)

Presenter: Martin Dybdal, PhD. Student, DIKU/University of Copenhagen

Time: Wednesday, August 9, 2017, 13:15

Place: HCØ, Auditorium 10, Universitetsparken 5, 2100 Copenhagen Ø

Committee:

Associate Professor Andrzej Filinski (chairman), Department of Computer Science, University of Copenhagen, Denmark
Associate Professor Gabriele Keller, University of New South Wales, Australia
Professor Peter Sestoft, IT University of Copenhagen, Denmark

Abstract

The shift towards massively parallel hardware platforms for high-performance computing tasks has introduced a need for improved programming models that facilitate ease of reasoning for both users and compiler optimization.

A promising direction is the field of functional data-parallel programming, for which functional invariants can be utilized by optimizing compilers to perform large program transformations automatically. However, the previous work in this area allow users only limited ability to reason about the performance of algorithms. For this reason, such languages have yet to see wide industrial adoption.

We present two programming languages that attempt at both supporting industrial applications and providing reasoning tools for hierarchical data-parallel architectures, such as GPUs.

First, we present TAIL, an array based intermediate language and compiler framework for compiling a large subset of APL, a language which have been used in the financial industry for decades. The TAIL language is a typed functional intermediate language that allows compilation to data-parallel platforms, thereby providing high-performance at the fingertips of APL programmers.

Second, we present FCL, a purely functional data-parallel language, that allows for expressing data-parallel algorithms in a fashion where users at a low-level can reason about data-movement through the memory hierarchy and control fusion will and will not happen. We demonstrate through a number of micro benchmarks that FCL compiles to efficient GPU code.

Biography

Martin Dybdal is a PhD student at DIKU/University of Copenhagen under supervision of Associate Professor Martin Elsman, DIKU.

Host: DIKU and HIPERFIT (Assoc. Prof. Martin Elsman)

PhD Defence: Maj-Britt Nordfang

2017-06-16T00:00:00+00:00

Title: Preferences, behaviour and the design of financial contracts

Thesis: To be published later

Presenter: Maj-Britt Nordfang, PhD. Student, MATH/University of Copenhagen

Time: June 16, 2017, 14:00

Place: HCØ, Auditorium 4, Universitetsparken 5, 2100 Copenhagen Ø

Committee:

Prof. Rolf Poulsen (chairman), University of Copenhagen
Ass. Prof Matthias Albrecht Fahrenwaldt, Heriot Watt University
Prof. An Chen, Ulm University

Abstract

Today’s insurance and financial markets are characterized by an ever-increasing complexity and product choice. How do we assist individual household consumers in making informed choices about their financial positions in these markets? This thesis consists of five separate research papers that are all in some way concerned with the design and regulation of financial contracts in relation to consumer preferences under uncertainty. First, the design of mortgage products in relation to consumer preferences in a market with a stochastic interest rate is examined. We derive the (extreme) characteristics of an investor who optimally chooses either a fixed rate mortgage or an adjustable rate mortgage. We also show how adjustable rate mortgages by construction have a build-in cap on the adjustable rate in contrast to what could be perceived from standard financial advice. Next, optimal investments of an investor with time-inconsistent preferences in a Black-Scholes market is explored. We derive the optimal investment strategy of an investor with a scaled mean-variance objective and derive an approximation to the optimal investment strategy of an investor who continuously measures utility with respects to an updated wealth-dependent reference point. Finally, the implications of a ban on risk-classification in an insurance market with a low-risk and high-risk customer groups with homogeneous utility preferences is examined and discussed. We illustrate how solidarity in premiums between the low- and a high-risk group of insurance customers may lower the premiums of the high-risk customers without affecting the premium level of the low-risk customers when a solvency capital requirement is imposed. The research papers highlight how the risk preferences of individual consumers relate to the design of financial contracts and the regulation hereof. Through the insights gained in the thesis, we intend to contribute to a better understanding of the complex financial decisions faced by individual consumers today.

Biography

Maj-Britt Nordfang is a PhD student at MATH/University of Copenhagen under supervision of Professor Mogens Steffensen, MATH. The PhD thesis work has been co-supervised by Trine Krogh Boomsma, MATH, and Ken Friis Larsen, DIKU.

Host: DIKU and HIPERFIT (Prof. Mogens Steffensen)

Blockchain Meeting

2017-01-11T00:00:00+00:00

Meeting in DIKU Business Club on Blockchain Technology

In relation to the DIKU Business Club, DIKU is hosting an afternoon workshop on Blockchain technology on January 11, 2017. Attendance is free, but you need to signup at the official announcement page, which also contains abstracts for the talks below.

Program

16.00-16.10: Welcome by Head of Dept. Mads Nielsen, DIKU
16.10-16.20: Blockchain - brief intro by professor Fritz Henglein, DIKU Business Club
16.20-18.15: Talks by confirmed speakers
Blockchain–-hope or hype? by Lars Stage Thomsen, Chief Consultant, Concept Development and Digital Hub at Danske Bank
Blockchain and the Emergence of the Trust-free Economy by Prof. Dr. Roman Beck, IT University of Copenhagen
Blockchain at SKAT-–Challenges and Solutions by Mikkel Christiansen, Special Consultant, SKAT
Break
In Search for the Sweet Spots of Blockchain by Michel Avital, Copenhagen Business School
Intro to legal challenges of Blockchain by Martin Haller von Grønbæk, Partner at Bird & Bird Copenhagen Office
Blockchain technology deconstructed (distributed systems aspects) by Fritz Henglein, DIKU
18.15-18.30: Reserved for speed case session by members of DIKU Business Club
18.30ff: Food, drinks and mingling in the Southern end of HC Ørsteds Instituttet

Location

Auditorium 1, HCØ, Universitetsparken 5, 2100 Copenhagen.

PhD Defence: Frederik Meisner Madsen

2016-12-15T00:00:00+00:00

Title: Streaming for Functional Data-Parallel Languages

Presenter: Frederik Meisner Madsen, PhD. Student, DIKU/University of Copenhagen

Time: Thursday, December 15, 2016, 13:15

Place: Room Aud 05, HCØ, Universitetsparken 5, 2100 Copenhagen Ø

Committee:

Assocociate Professor Martin Elsman (chairman), DIKU, University of Copenhagen
Professor John Reppy, Department of Computer Science, University of Chicago, USA
Professor Mary Sheeran, Department of Computer Science and Engineering, Chalmers University of Technology, Sweden

Thesis (revised): pdf

Abstract

In this thesis, we investigate streaming as a general solution to the space inefficiency commonly found in functional data-parallel programming languages. The data-parallel paradigm maps well to parallel SIMD-style hardware. However, the traditional fully materializing execution strategy, and the limited memory in these architectures, severely constrains the data sets that can be processed. Moreover, the language-integrated cost semantics for nested data parallelism pioneered by NESL depends on a parallelism-flattening execution strategy that only exacerbates the problem. This is because flattening necessitates all sub-computations to materialize at the same time. For example, naive n by n matrix multiplication requires n^3 space in NESL because the algorithm contains n^3 independent scalar multiplications. For large values of n, this is completely unacceptable.

We address the problem by extending two existing data-parallel languages: NESL and Accelerate. In the extensions we map bulk operations to data-parallel streams that can evaluate fully sequential, fully parallel or anything in between. By a dataflow, piecewise parallel execution strategy, the runtime system can adjust to any target machine without any changes in the specification. We expose streams as sequences in the frontend languages to provide the programmer with high-level information and control over streamable and non-streamable computations. In particular, we can extend NESL’s intuitive and high-level work–depth model for time complexity with similarly intuitive and high-level model for space complexity that guarantees streamability.

Our implementations are backed by empirical evidence. For Streaming Accelerate we demonstrate performance on par with Accelerate without streams for a series of benchmark including the PageRank algorithm and a MD5 dictionary attack algorithm. For Streaming NESL we show that for several examples of simple, but not trivially parallelizable, text-processing tasks, we obtain single-core performance on par with off-the-shelf GNU Coreutils code, and near-linear speedups for multiple cores.

Notice: For an electronic copy of the thesis, please contact phdadmin@di.ku.dk.

Biography

Frederik Meisner Madsen is a PhD student at DIKU/University of Copenhagen under supervision of Associate Professor Andrzej Filinski, DIKU.

Host: DIKU and HIPERFIT (Associate Professor Andrzej Filinski)

Talk on Nessie by John Reppy

2016-12-14T00:00:00+00:00

HIPERFIT Seminar Talk

Title: Nessie: A NESL to CUDA Compiler

Presenter: John Reppy, Computation Institute, University of Chicago

Time: Wednesday, December 14th, 2016, 13:15-15:00

Place: Universitetsparken 1, Small Auditorium, University of Copenhagen.

Abstract:

Modern GPUs provide supercomputer-level performance at commodity prices, but they are notoriously hard to program. To address this problem, we have been exploring the use of Nested Data Parallelism (NDP), and specifically Guy Blelloch’s first-order functional language NESL, as a way to raise the level of abstraction for programming GPUs. Preliminary results suggest that NDP can be effectively mapped onto GPUs, but there is significant work required to make this mapping competitive with hand-written GPU code. This talk describes ongoing work on a new compiler for the NESL language that generates CUDA code. Specifically, we will describe several aspects of the compiler that address some of the challenges of generating efficient NDP code for GPUs.

This work is joint with Nora Sandler and Joeseph Wingerter.

Notice: The presentation is part of the course on Parallel Functional Programming (PFP) at DIKU.

Biography

John Reppy is a Professor of Computer Science and a Senior Fellow of the Computation Institute at the University of Chicago. He received his Ph.D. from Cornell University in 1992 and spent the first eleven years of his career at Bell Labs in Murray Hill NJ. He has been exploring issues in language design and implementation since the late 1980’s, with a focus on higher-order, typed, functional languages. His work includes the invention of Concurrent ML and work on combining object-oriented and functional language features. His current research is on high-level languages for parallel programming, including the Manticore and Diderot projects.

Host: Martin Elsman, DIKU, HIPERFIT.

PhD Defence: Martin A. Jönsson

2016-08-26T00:00:00+00:00

Title: Essays in Quantitative Finance

Presenter: Martin A. Jönsson, PhD. Student, Math/University of Copenhagen

Time: Friday, August 26, 2016, 13:15

Place: HCØ, Auditorium 3, Universitetsparken 5, 2100 Copenhagen Ø

Committee:

Prof. Mogens Steffensen (chairman), MATH, University of Copenhagen
Prof. Antje Mahayni, Universität Duisburg-Essen
Prof. Claus Munk, Copenhagen Business School

Abstract

The first two papers concern option pricing from two different perspectives: First, we study stochastic volatility models subject to parameter uncertainty in order to approach worst-case option prices based on a control theoretic approach. We explore a formulation with backward stochastic differential equations and detail several numerical methods for their solution. In an empirical study, we then compare the conservative model-derived prices with real market data of European call options. In the second paper, we study the delta-hedge for options in a setting where the hedger employs an erroneously specified volatility. We derive results for the incurred profit-and-loss from the hedge portfolio and explore two special cases: hedging with option-market implied volatility and hedging in accordance with the “true” dynamics. The theoretical implications invite the hedger to arbitrage opportunities; we scrutinize on their applicability in an empirical study.

The third paper considers a practical hedging situation commonly faced by retailers in electricity markets. We look at the problem of price-quantity risk and study static hedging strategies for fixed-price-agreement contracts. We propose a bivariate model for this purpose and employ risk-minimization techniques. The approach is then empirically tested on market data from the Nordic power market and compared to industry standards. In the forth paper we look at the problem of optimal investment in a bond-stock-option economy driven by a stochastic volatility model. Based on martingale methods, we derive explicit formulas for the optimal portfolio strategy when the market option is plain vanilla and we set the suggested plan to work in an empirical study. In the final paper of the thesis we ask the simple question of what is a good model of volatility. In contrast to the usual measure of model-to-market fit for option prices, we focus on the volatility process itself and how well stochastic volatility models match its distributional properties. We suggest a goodness-of-fit analysis for this purpose and perform an extensive empirical study based on a market index and several individual stocks.

Biography

Martin A. Jönsson is a PhD student at Math/University of Copenhagen under supervision of Prof. Rolf Poulsen, Math.

Host: Math and HIPERFIT (Prof. Rolf Poulsen)

Talk on File System Scalability by André Brinkmann

2016-05-17T00:00:00+00:00

HIPERFIT Seminar Talk

Title: File System Scalability with Highly Decentralized Metadata on Independent Storage Devices

Presenter: André Brinkmann, Johannes Gutenberg Universität, Mainz

Time: Tuesday, May 17, 2016, 15:01-16:00

Place: Universitetsparken 5, Auditorium 10 (mezzanine, south end of building), University of Copenhagen.

Abstract:

Data handling concepts for Cloud Computing and High-Performance Computing have been developed nearly independently from each other and are still mostly restricted to their initial use cases. Cloud Computing approaches try to scale out by removing the file system abstraction and applying a flat key-value namespace, while HPC environments rely on file system semantics and the corresponding management capabilities. This talk discusses an approach to overcoming the resulting gap by building a file system with a (very slightly) relaxed Posix-semantics on top of the key-value interface of the Seagate Kinetic storage platform. Taking advantage of higher-level functionality to handle metadata on the drives themselves a server-less system architecture is proposed. Skipping path component traversal during the lookup operation is the key technique discussed in this paper to avoid performance degradation with highly decentralized metadata. Scalability implications are reviewed based on a fuse file system implementation.

Biography

André Brinkmann is full professor at the computer science department of Johannes Gutenberg University in Mainz, where he heads the university’s data processing center. He received his Ph.D. in electrical engineering in 2004 from the University of Paderborn, where he subsequently was assistant professor and managing director of the Paderborn Centre for Parallel Computing (PC2) from 2008 to 2011. His research interests focus on algorithm engineering techniques in data centre management, cloud computing, and storage systems. He has published more than 100 papers in renowned conferences and journals. He is associate editor of ACM Transactions on Storage and steering committee member of French Grid’5000 and the IEEE Symposium on Massive Storage Systems and Technologies (MSST) and IEEE International Conference on Networking, Architecture, and Storage (NAS).

All are welcome. No registration required.

Hosts: Fritz Henglein, DIKU, HIPERFIT.

PhD Status Seminar for Troels Henriksen

2016-05-03T00:00:00+00:00

HIPERFIT Seminar Talk: PhD Status Seminar for Troels Henriksen

Presenter: Troels Henriksen, PhD. Student, DIKU/University of Copenhagen

Time: Tuesday, May 3, 2016, 15:15-16:00

Place: HCØ, AUD 10, Universitetsparken 5, University of Copenhagen.

Abstract:

I will be presenting the work I have been doing during my two years of PhD studies at DIKU. I will talk about:

…how I started my PhD work a year before I started my PhD.
…what happens when you apply Haskell-inspired DSL techniques to C++. (The results are now hidden from the world in a DOE supercomputer.)
…how the Futhark compiler grew from a leaky toy to something that can actually produce decent-ish GPU code.

There will be pretty pictures.

Biography

Troels Henriksen is a PhD student at DIKU and works on the implementation and design of the HIPERFIT parallel-functional programming language Futhark.

Host: Martin Elsman and Cosmin Oancea

Accelerate

2016-04-05T00:00:00+00:00

HIPERFIT Seminar Talk

Title: Accelerate

Presenter: Gabriele Keller, University of New South Wales

Time: Tuesday, April 5th, 2016, 15:00-16:00

Place: Universitetsparken 5, Auditorium 7 (mezzanine, south end of building), University of Copenhagen.

Abstract:

In this talk I will present an overview of the Accelerate project - a domain specific language for GPU programming embedded in Haskell - and report on ongoing work. I will also talk about the objectives of a new project supported by the Australian Research Council, to extend Accelerate with support for streams, both for single as well as multicore GPU architectures.

Biography

Gabriele Keller is an Associate Professor in the Computer Science and Engineering Department at UNSW Australia and a Principal Researcher at Data61 (formerly NICTA) in the Trustworthy Systems project. Her research interest are type systems, functional languages, and how these languages can be used to reduce the costs of software development, in particular in the context of high-performance computing and safety critical systems.also a Principal Researcher at Data61 (formerly NICTA) in the Trustworthy Systems project. Her research interest are type systems, functional languages, and how these languages can be used to reduce the costs of software development, in particular in the context of high-performance computing (Accelerate and Repa Haskell packages) and safety critical systems (Trustworthy Systems Project, Data61).

Hosts: Fritz Henglein, DIKU, HIPERFIT.

PhD Defence: Lykke Rasmussen

2016-03-31T00:00:00+00:00

Title: Computational Finance - on the search for performance

Thesis: PDF

Presenter: Lykke Rasmussen, PhD. Student, Math/University of Copenhagen

Time: Thursday, March 31, 2016, 13:15

Place: HCØ, Auditorium 10, Universitetsparken 5, 2100 Copenhagen Ø

Committee:

Ass. Prof. David Skovmand (chairman), MATH, University of Copenhagen
Prof. Natalie Packham, Frankfurt School of Finance & Management
Antoine Savine, Chief Quantitative analyst, Danske Bank, Copenhagen

Abstract

The PhD thesis ‘Computational Finance - on the search for performance’ consists of two papers. The first examines the performance of five different methods for calibrating Dupire’s deterministic local volatility function, these are assessed in terms of accuracy, smoothness, speed and robustness. The second presents four different methods for evaluating the Greeks of an interest rate swap, these are assessed in terms of accuracy, stability and run time.

Biography

Lykke Rasmussen is a PhD student at Math/University of Copenhagen under supervision of Prof. Rolf Poulsen, Math.

Host: Math and HIPERFIT (Prof. Rolf Poulsen)

HIPERFIT Workshop

2016-03-03T00:00:00+00:00

2016 HIPERFIT Workshop for Partners and Faculty

On March 3, 2016, HIPERFIT is hosting a workshop for HIPERFIT HIPERFIT partners and faculty. At the workshop, HIPERFIT faculty, including PhD students, will present the latest developments in the various HIPERFIT project. A separate session is devoted to partners for presenting opportunities for involving HIPERFIT researchers in obstacles encountered by the partners with respect to high-performance computational problems in the finance domain.

The year 2016 is formally the last HIPERFIT year and much has happened since the HIPERFIT launch in 2011. We are therefore also happy to demonstrate our more long term findings and how we envision that the HIPERFIT work can continue to make a difference in the years to come.

The Venue

Kurhotel Skodsborg, Skodsborg Strandvej 139, 2942 Skodsborg (45585800)

Schedule March 3, 2016

08:30

Breakfast served

08:55

Welcome (Fritz Henglein)

09:00

Compiling for Parallel Computations (Chair: Ken Friis Larsen)

Streaming nested data parallelism (Frederik Meisner Madsen, DIKU)
A low-level functional GPU language (Martin Dybdal, DIKU)
Verifying the generation of payoff-language expressions (Danil Annenkov, DIKU)

10:15

Break

10:30

The Futhark Optimizing Parallel Compiler (Chair: Fritz Henglein)

Introduction to the Futhark language and compiler (Troels Henriksen, DIKU)
Experiences porting Accelerate code to Futhark (Niels Gustav Westphal Serup)

11:30

Break

11:45

Bohrium (Chair: Andrzej Filinski)

SME and BPU: High-level development for low-level hardware (Kenneth Skovhede, NBI)
Fusion of Parallel Array Operations (Mads Kristensen, NBI)
BDAE – Big Data Analysis Engine (Brian Vinter, NBI)

13:00

Lunch

14:00

Risk Management (Chair: Rolf Poulsen)

Automated Loan Credit Rating (Omri Ross, IMF)
The MC-squared problem (Oleksandr Shturmov, DIKU)

15:00

Break

15:15

Miscellaneous Topics (Chair: Cosmin Oancea)

The HIPERFIT portfolio management prototype (Martin Elsman, DIKU)
Hedge Funds Don't Hedge - And 50-Odd Other Odd Things in Finance (Rolf Poulsen, IMF)

16:00

Industry feedback (Chair: Fritz Henglein)

17:00

Chat and a drink - free time until dinner

18:30

Dinner

A table of participants for the workshop on March 3, and for the discussion and planning sessions (HIPERFIT staff only) on March 4, appears below. The schedule for March 4 appears at the very end of the page.

Participation in the workshop is by invitation only, but please contact the HIPERFIT management if you think you should have received an invitation.

As always, please visit http://www.hiperfit.dk for news and information about published papers, our Seminars, and news in general from the HIPERFIT research center.

Participants

Name	March 3	Dinner March 3	March 4	Room
Anders Pall Skött (CFIR)	1	1	0	0
Andrzej Filinski (DIKU)	1	1	1	1
Anette Broløs (CFIR)	1	0	0	0
Brian Vinter (NBI)	1	1	1	1
Carl Balslev Clausen (SimCorp)	0	0	0	0
Christian Andreetta (Nordea)	1	1	0	0
Cosmin Oancea (DIKU)	1	1	1	1
Danil Annenkov (DIKU)	1	1	1	1
Frederik Meisner Madsen (DIKU)	1	1	1	1
Fritz Henglein (DIKU)	1	1	1	1
James Avery (NBI)	0	0	0	0
Jonas Bardino (NBI)	1	1	1	1
Ken Friis Larsen (DIKU)	1	1	1	1
Kenneth Skovhede (NBI)	1	1	1	1
Klaus Birkelund Jensen (NBI)	1	1	1	1
Lykke Rasmussen (IMF)	0	0	0	0
Mads R. B. Kristensen (NBI)	1	1	1	1
Maj-Britt Nordfang (IMF)	1	0	1	0
Marcos Vaz Salles (DIKU)	1	1	0	0
Martin Dybdal (DIKU)	1	1	1	1
Martin Elsman (DIKU)	1	1	1	1
Mads Ohm Larsen (NBI)	1	1	1	1
Martin Rehr (NBI)	1	1	1	1
Mogens Steffensen (IMF)	0	0	0	0
Niels Hallenberg (SimCorp)	1	1	0	0
Oleksandr Shturmov (DIKU)	1	1	1	1
Omri Ross (IMF)	1	1	1	1
Patrick Bahr (ITU)	1	1	0	0
Rolf Poulsen (IMF)	1	1	1	1
Simon Ellersgaard (IMF)	0	0	0	0
Simon Lund (NBI)	0	0	0	0
Troels Blum (NBI)	1	1	1	1
Troels Henriksen (DIKU)	1	1	1	1
Uwe Heissner (Nordea)	1	0	0	0
Niels Gustav Westphal Serup (DIKU)	1	1	0	0
Vivek Shah (DIKU)	0	0	0	0
Total	29	26	21	20

Schedule March 4, 2016 (HIPERFIT staff only)

On March 4, 2016, HIPERFIT staff will meet and discuss feedback and plan future activities and projects within HIPERFIT. Here is a draft schedule.

08:30: Breakfast served
08:55: Outline for the day (Plenum)
09:00: Session 1 (Groups)
12:00: Lunch
13:00: Session 2 (Groups)
15:00: Break
15:15: Group Presentations (Plenum)
16:30: End

Abstracts for Talks on March 3, 2016

09:00-10:15 Compiling for Parallel Computations

Streaming Nested Data Parallelism (Frederik Meisner Madsen)

This talk presents ongoing work on a streaming dataflow model for nested data parallelism. By evaluating bulk operations in chunks in a network of stream transformers, the streaming model saves significantly more space than traditional data-parallel languages in many common cases. On GPUs, the device memory is very limited relative to the available parallelism, making streaming essential in large computations. On CPUs, streaming can help improve cache performance, and since fusion is not as essential on CPUs as on GPUs, the streaming model can leverage pre-compiled kernels that are hand-optimized with SIMD instructions and multi-threading.

A low-level functional GPU language (Martin Dybdal)

Obsidian is an embedded functional language that allows playfulness and experimentation when developing data-parallel algorithms for GPUs. Obsidian gives explicit control of shared-memory usage, fusion, loop-unrolling, warp-level computations etc., when developing new GPU algorithms. This feature makes it possible to generate high performance GPU-kernels, in a simpler framework than plain CUDA/OpenCL.

In this presentation, we show how to “unembed” Obsidian to create a standalone language FCL, a language at the same level as Obsidian, but designed as an intermediate language. Our plan is to use FCL as an intermediate language for our APL-compiler, and to allow users to inline FCL-terms, when performance tuning is necessary.

Verifying the generation of payoff-language expressions (Danil Annnenkov)

In this presentation, we present some results on pushing forward code certification in connection with the HIPERFIT Portfolio Management Prototype project. The contract DSL, which is used to express a large variety of financial contracts, is translated into an intermediate language (IL) inspired by traditional approaches to payoff languages. The IL is mapped relatively straightforwardly to a subset of language constructs in other languages, such as OpenCL, Haskell, and Futhark. The generated code is then “fused” into an appropriate contract valuation engine. Translation from the contract DSL to the IL is proven correct with respect to a language semantics specified in the Coq proof assistant. The extracted translation code works nicely with the certified code for contract analysis and transformation.

10:30-11:30 The Futhark Parallel Optimizing Compiler

Introduction to the Futhark language and compiler (Troels Henriksen, DIKU)

Compiling real-world applications to efficient parallel code, especially when targeting vector hardware such as GPUs, requires an optimiser and code generator that can exploit a combination of high-level invariants and low-level optimisations. We present a purely functional data-parallel programming language, named Futhark, that supports nested map-reduce parallelism on regular arrays, but also a set of “imperative” constructs, such as in-place updates and do-loops. We report in-progress work on our optimising Futhark compiler, and demonstrate significant speedups for a range of benchmark programs.

Experiences porting Accelerate code to Futhark (Niels Gustav Westphal Serup, DIKU)

The people working on the embedded array language Accelerate have a repository of computation-heavy benchmarks. In this talk I present how I have ported some of their benchmarks to Futhark. I then describe how I have creatively measured the correctness of the Futhark code, which can be challenging due to several technical factors. Finally I compare and discuss the execution time of hand-written Futhark code and hand-written Accelerate code.

11:45-13:00 Bohrium

SME and BPU: High-level development for low-level hardware (Kenneth Skovhede, NBI)

In this talk I present the motivation and ideas behind the FPGA backend for the Bohrium Runtime System. I start by covering the evolution and experiments we performed with Synchronous Message Exchange (SME) as a design pattern for developing hardware. I then describe how we use the SME model and our VHDL generation system to rapidly prototype hardware. Finally I show the current state of the BPU and show how the processor maps to FPGA hardware.

Fusion of Parallel Array Operations (Mads Kristensen, NBI)

Array operation fusion is a program transformation that combines, or fuses, multiple array operations into a kernel of operations. When it is applicable, the technique can drastically improve cache utilization through temporal data locality and enables other program transformations such as streaming and array contraction. Thus, the challenge is to find which array operations to fuse in order to maximize memory access reuse.

The traditional formulation of this problem is as a graph partitioning problem. Given a DAG that represents array operation dependencies and is extended with undirected weight edges and fuse-preventing edges, find a partition that minimizes the sum of the weight edges that cuts between partition blocks while nodes connected with fuse-preventing edges are not in the same partition block. This graph problem is known as that Weighted Loop Fusion Problem. In my talk, I will show that this formulation is insufficient when optimizing for memory access reuse and introduce a new graph formulation that is sufficient.

BDAE – Big Data Analysis Engine (Brian Vinter, NBI)

In Big Data Analytics the primary processing approach is MapReduce, which based on a physical distribution of the data so that processing is done in parallel as a direct consequence of the data locality. This approach is very scalable and to a large degree convenient for programmers since the programmer initially need only to write code to process the data that is on the individual node. Unfortunately, the physical distribution of the data is done with conventional file-system approaches, i.e. in fixed size blocks, in the most popular HDFS the standard is 64MB blocks. This fixed size approach means that data that is semantically correlated is split between two storage nodes, and to handle this scenario the programmer must then add code to manage communication between the two nodes so that the correct data-representation can be processed.

In BDAE we address this challenge, the talk will outline the BDAE architecture and seek to demonstrate how easy programming new analysis codes becomes with this setup.

14:00-15:00 Risk Management

Automated Loan Credit Rating (Omri Ross, IMF)

The current process of credit rating for a loan approval in Denmark is manual, cumbersome and non-optimal. We would like to propose a process allowing for a quick credit decision based on the available personal data. Our process has two steps: First using a non-linear dimensional reduction algorithm on the original data and then running a classification algorithm on the results of the first step. We improve existing results compared to the standard benchmarks.

The MC-squared problem (Oleksandr Shturmov, DIKU)

15:15-16:00 Miscellaneous Topics

The HIPERFIT portfolio management prototype (Martin Elsman, DIKU)

We present the HIPERFIT portfolio management prototype, which aims at integrating various HIPERFIT research projects, such as a financial contract modeling framework and generic parallel monte-carlo pricing.

Integration work and design has primarily been undertaken by Danil Annenkov. Features added by BSc student projects.

Hedge Funds Don’t Hedge - And 50-Odd Other Odd Things in Finance (Rolf Poulsen, IMF)

A tour de force of things in finance that are not exactly what they sound like. For instance: Risk-neutral pricing does not assume risk-neutrality. The FED isn’t federal. Hedge funds don’t hedge. What’s up with the Mertons? Default isn’t default. Change of numeraire changes the denominator, not the numerator.

PhD Defence: Simon Ellersgaard Nielsen

2016-02-29T00:00:00+00:00

Title: Essays on Rational Portfolio Theory

Presenter: Simon Ellersgaard Nielsen, PhD. Student, Math/University of Copenhagen

Time: February 29, 2016, 15:15

Place: HCØ, Auditorium 4, Universitetsparken 5, 2100 Copenhagen Ø

Committee:

Prof. Mogens Steffensen (chairman), MATH, University of Copenhagen
Prof. Nicole Branger, University of Muenster
Prof. Claus Munk, Copenhagen Business School

Abstract

This dissertation is a potpourri of articles on the topics of model risk and optimal portfolio selection in stochastic environments. The main concerns are:

Demonstrating the effect of erroneous delta hedging in a jump-diffusion economy. What is the nature of the profit-and-loss from a theoretical perspective? From an empirical perspective?
Analysing the method by which the non-linear Hamilton-Jacobi-Bellman equations should be solved in connection with Merton type optimisation problems. A thorough review of the explicit and implicit methods is provided in one and more spatial dimensions.
Exposing the optimal investment ratios for a utility maximising investor who trades in bonds and stocks in a stochastic volatility environment. Various models are considered, including their effect in empirical trading experiments.
Extending the above analysis to include the derivatives markets. How much do the portfolio weights change? What is the effect of hedging stochastic volatility per se versus merely including a second asset? Is the bond-stock-derivative strategy truly superior when applied to real market data?
Hedging derivatives in a limit order book with the option of placing both limit and market orders. Assuming a certain tolerance towards deviating from a targeted hedge strategy when should a rational investor place which type of order?

Overall, the project strives to strike a careful balance between theoretical advances and empirical testing of results. An extensive appendix introducing some of the underlying mathematics is provided.

Biography

Simon Ellersgaard Nielsen is a PhD student at Math/University of Copenhagen under supervision of Prof. Rolf Poulsen, Math and co-supervisor Assoc. Prof. Ken Friis Larsen, Computer Sciense, University of Copenhagen.

Host: Math and HIPERFIT (Prof. Rolf Poulsen)

Chapel: Parallel Programmability from Desktops to Supercomputers

2016-02-04T00:00:00+00:00

HIPERFIT Seminar Talk

Title: Chapel: Parallel Programmability from Desktops to Supercomputers.

Presenter: Brad Chamberlain, Principal Engineer, Cray Inc.

Time: Thursday, February 4, 2016, 15:15-16:15

Place: DIKU Building, Small Auditorium, UP1, Universitetsparken 1, University of Copenhagen.

Abstract:

Chapel (http://chapel.cray.com) is an emerging open-source language whose goal is to vastly improve the programmability of parallel systems while also enhancing generality and portability compared to conventional techniques. Chapel is seeing growing levels of interest not only among HPC users, but also in the data analytic, academic, and mainstream programming communities. Chapel’s design and implementation are portable and open-source, supporting a wide variety of compute platforms, from desktops (Mac, Linux, *nix) to commodity clusters, the cloud, and large-scale systems developed by Cray and other vendors.

In this talk, I’ll start by providing motivation and context for Chapel before giving an overview of its unique feature set. I’ll also describe the status and organization of the Chapel project itself, highlighting opportunities for collaboration.

Slides

Biography

Brad Chamberlain is a Principal Engineer at Cray Inc. where he works on parallel programming models, focusing primarily on the design and implementation of the Chapel language in his role as technical lead for that project. Brad received his Ph.D. in Computer Science & Engineering from the University of Washington in 2001 where his work focused on the design and implementation of the ZPL parallel array language. His thesis explored the concept of ‘regions’ — a first-class index set supporting global-view data parallelism and a syntactic performance model. Brad remains associated with the University of Washington as an affiliate professor. He received his Bachelor’s degree in Computer Science with honors from Stanford University in 1992.

Hosts: Martin Elsman, DIKU and Brian Vinter, NBI.

Everyone is welcome! Beers, soda, and snacks will be served after the talk.

Automated Loan Credit Rating

2016-01-26T00:00:00+00:00

HIPERFIT Seminar Talk: Automated Loan Credit Rating

Presenter: Omri Ross, Post-doctorial Researcher, IMF/University of Copenhagen

Time: Tuesday, January 26, 2016, 15:15-16:00

Place: August Krogh Building (AKB), AUD 1. Universitetsparken 13, University of Copenhagen.

Abstract:

Biography

Omri Ross has done a PhD in mathematical finance at the university of Cambridge followed by a post-doc at DTU and KU. His main research interests are financial engineering and tactical asset allocation. The work presented here is part of a collaboration between KU and an industrial partner Densou Trading Desk.

Host: Martin Elsman and Rolf Poulsen

Status of the GPU-targeting Futhark compiler

2016-01-12T00:00:00+00:00

HIPERFIT Seminar Talk: Status of the GPU-targeting Futhark compiler

Presenter: Troels Henriksen, PhD. Student, DIKU/University of Copenhagen

Time: Tuesday, January 12, 2016, 15:15-16:00

Place: August Krogh Building (AKB), AUD 1, Universitetsparken 13, University of Copenhagen.

Abstract:

Futhark is a data-parallel programming language under development at HIPERFIT. The language supports nested data-parallelism and has an optimising compiler under development. This presentation gives a status update on recent Futhark development, with a particular focus on generation of efficient OpenCL GPU kernels and the steps automatically taken by the Futhark compiler to ensure efficient memory accesses. The presentation will always talk about various tradeoffs during code generation, and why it’s often not a good idea to exploit all the parallelism you have available.

Biography

Troels Henriksen is a PhD student at DIKU and works on the implementation and design of the HIPERFIT parallel-functional programming language Futhark.

Host: Martin Elsman and Cosmin Oancea

PhD Defence: Troels Blum

2016-01-04T00:00:00+00:00

Title: Automatic Parallelization of Scientific Applications

Presenter: Troels Blum, PhD. Student, NBI/University of Copenhagen

Time: Thursday, January 4, 2016, 10:00-13:00

Place: Aud A, NBI, Blegdamsvej 17

Committee:

Prof. Dr. Stig Skelboe, UCPH
Kevin Chalmers, Edinburgh Napier University, UK
Jan Bækgaard Pedersen, UNLV, USA

Abstract

In my PhD work I show that it is possible to run unmodified Python/NumPy code on modern GPUs. This is done by using the Bohrium runtime system to translate the NumPy array operations into an array based bytecode sequence. Executing these byte-codes on two GPUs from different vendors shows great performance gains.

Scientists working with computer simulations should be allowed to focus on their field of research and not spend excessive amounts of time learning exotic programming models and languages. We have with Bohrium achieved very promising results by starting out with a relatively simple approach. This has lead to more specialized methods as I have shown with the work done with both specialized, and parametrizied kernels. Both have their benefits and recognizable use cases. We achieved clear performance benefits without any significant negative impact on overall application performance. Even in the cases where we were not able to gain any performance boost by specialization, the added cost, for kernel generation and extra bookkeeping, is minimal.

Many of the lessons learned developing and optimizing the Bohrium GPU vector engine has proven to be valuable in a broader perspective, which has made it possible to generalize the developments and made them benefit the complete Bohrium project.

Biography

Troels Blum is a PhD student at NBI under supervision of Prof. Brian Vinter.

Host: NBI and HIPERFIT (Prof. Brian Vinter)

PhD Defence: Simon A F Lund

2015-12-10T00:00:00+00:00

Title: A High Performance Backend for Array-Oriented Programming on Next-Generation Processing Units

Presenter: Simon A F Lund, PhD. Student, NBI/University of Copenhagen

Time: Thursday, December 10, 2015, 10:00

Place: Aud D, NBI, Blegdamsvej 17

Committee:

Prof. Dr. Stig Skelboe, UCPH
Prof. Dr. Siegfried Benkner, University of Vienna
Associate Professor John Markus Bjørndalen, University of Tromsø

Abstract

The financial crisis, which started in 2008, spawned the HIPERFIT research center as a preventive measure against future financial crises. The goal of prevention is to be met by improving mathematical models for finance, the verifiable description of them in domain-specific languages and the efficient execution of them on high performance systems.

This work investigates the requirements for, and the implementation of, a high performance backend supporting these goals. This involves an outline of the hardware available today, in the near future and how to program it for high performance. The main challenge is to bridge the gaps between performance, productivity and portability.

A declarative high-level array-oriented programming model is explored to achieve this goal and a backend implemented to support it. Different strategies to the backend design and application of optimizations are analyzed and experimentally tested. Resulting in the design and implementation of Bohrium a runtime-system for transforming, scheduling and executing array-oriented programs.

Multiple interfaces for existing languages such as Python, C++, C#, and F# have been built which utilize the backend. A suite of benchmark applications, implemented in these languages, demonstrate the high-level declarative form of the programming model. Performance studies show that the high-level declarative programming model can be used to not only match but also exceed the performance of hand-coded implementations in low-level languages.

Biography

Simon A F Lund is a PhD student at NBI under supervision of Prof. Brian Vinter.

Host: NBI and HIPERFIT (Prof. Brian Vinter)

The HIPERmark Benchmark Toolkit

2015-11-10T00:00:00+00:00

HIPERFIT Seminar Talk: The HIPERmark Benchmark Toolkit

Presenter: Thorkil Værge, MSc student, DIKU/University of Copenhagen

Time: Tuesday, November 10, 2015, 15:15-16:00

Place: AUD 7 (HCØ, Universitetsparken 5)

Abstract:

In this talk, we present the HIPERmark Benchmark Toolkit, an infrastructure for comparing and visualising performance tradeoffs. The infrastructure can automatically run several different benchmark implementations with different datasets and with different compile time and runtime variables and compare the performances across many different dimensions. Using this tool, it becomes easy, for example, to see how the performance of an OpenMP implementation scales as a function of the number of threads used, and also to see how this scaling differs when computing with single precision vs. double precision. The tool automates visualisation of the results in user-configurable graphs. The user can also export the results and process the data in an external program such as MatLab.

Biography

Thorkil Værge is MSc. in physics from UCPH, now studying and employed at DIKU. He is a Bitcoin enthusiast and founder of Sirius Money, and Sirius Iberia.

Host: Martin Elsman/Cosmin Oancea

Analyzing the Swiss National Bank's euro exchange rate policy: A latent likelihood approach

2015-10-27T00:00:00+00:00

HIPERFIT Seminar Talk: Analyzing the Swiss National Bank’s euro exchange rate policy: A latent likelihood approach

Presenter: Rolf Poulsen, Professor, IMF/University of Copenhagen

Time: Tuesday, October 27, 2015, 15:15-16:00

Place: AUD 7 (HCØ, Universitetsparken 5)

Abstract:

Between September 6, 2011 and January 15, 2015 the Swiss National Bank (SNB’s) acted in order to enforce their guarantee that the CHFEUR exchange rate (i.e. the no. Swiss francs needed to buy one euro) would be above 1.20. In this paper we view that guarantee as a put option and apply a latent likelihood estimation approach to infer the market’s view of the credibility of the SNB’s guarantee, where the exchange rate would be without the guarantee as well CHFEUR-volatility. Our model tracks general market volatility well, is quite accurate in its prediction for where the exchange rate would jump on the removal date January 15, 2015 and shows that the market’s confidence in the SNB’s guarantee declined sharply in months up the removal. In short, the events of January 15 were unexpected, but not that unexpected.

Based on joint work with Michael Hanke and Alex Weissensteiner.

Biography

Rolf Poulsen is Professor at the Department of Mathematical Sciences.

Host: Martin Elsman

Talk on Optimal Hedge Tracking Portfolios in a Limit Order Book

2015-09-29T00:00:00+00:00

HIPERFIT Seminar Talk: Optimal Hedge Tracking Portfolios in a Limit Order Book.

Presenter: Simon Ellersgaard Nielsen, PhD student, IMF/University of Copenhagen

Time: Tuesday, September 29, 2015, 15:15-16:00

Place: AUD 7 (HCØ, Universitetsparken 5)

Abstract:

In this paper we develop a control theoretic solution to the manner in which a portfolio manager optimally should track a targeted delta, given that he wishes to hedge a short position in European call options the underlying of which is traded in a limit order book. Specifically, we are interested in the interplay between posting limit and market orders: when should the portfolio manager do what (and at what price)? To this end, we set up an Hamilton-Jacobi-Bellman quasi-variational inequality which we can solve numerically. Our scheme is shown to be monotone, stable and consistent. Finally, we provide a concrete numerical study, comparing our algorithm with more naive approaches to delta-hedging.

Biography

Simon Ellersgaard Nielsen is a graduate student in financial mathematics at the University of Copenhagen, working under the supervision of Rolf Poulsen. His primary research interests lie within optimal portfolio theory and foundational issues in asset pricing. Besides finance, he dabbles in questions pertaining to the fundamental nature of reality.

Host: Martin Elsman/Rolf Poulsen

Talk on Functional Array Streams

2015-08-28T00:00:00+00:00

HIPERFIT Talk: Functional Array Streams

Presenter: Frederik Meisner Madsen, PhD student, DIKU/University of Copenhagen

Time: Friday August 28, 2015, 13-14

Place: The APL Meeting Room (DIKU, HCØ, Universitetsparken 5; room 01-0-029)

Abstract:

Regular array languages for high performance computing based on aggregate operations provide a convenient parallel programming model, which enables the generation of efficient code for SIMD architectures, such as GPUs. However, the data sets that can be processed with current implementations are severely constrained by the limited amount of main memory available in these architectures.

In this paper, we propose an extension of the embedded array language Accelerate with a notion of sequences, resulting in a two level hierarchy which allows the programmer to specify a partitioning strategy which facilitates automatic resource allocation. Depending on the available memory, the runtime system processes the overall data set in streams of chunks appropriate to the hardware parameters.

In this paper, we present the language design for the sequence operations, as well as the compilation and runtime support, and demonstrate with a set of benchmarks the feasibility of this approach.

Biography

Frederik Meisner Madsen is a PhD student at the HIPERFIT research center where he is involved in programming language research. Particular areas of interest include functional array languages, nested data-parallelism and streaming.

Host: Martin Elsman

HIPERFIT Bachelor Projects

2015-08-25T00:00:00+00:00

We are now inviting Bachelor students at DIKU to consider doing a project within the HIPERFIT research center. Please consult the slides presenting the projects.

Contacts:

Danil Annenkov and Martin Elsman

Talk on Compiling APL to Accelerate

2015-06-09T00:00:00+00:00

HIPERFIT Seminar Talk: Compiling APL to Accelerate through a Typed Array Intermediate Language

Presenter: Michael Budde, MSc student, DIKU/University of Copenhagen

Time: Tuesday June 9, 2015, 15-16

Place: Aud 07 (HCØ, Universitetsparken 5)

Abstract:

In this talk, we present an approach for compiling a rich subset of APL into data-parallel programs that can be executed on GPUs. The compiler is based on the APLTail compiler, which compiles APL programs into a typed array intermediate language, called TAIL. We translate TAIL programs into Haskell source code, employing Accelerate, a Haskell-library for general purpose GPU-programming. We demonstrate the feasibility of the approach by presenting some encouraging results for a number of smaller benchmarks. We also outline some problems that we need to overcome in order for the approach to result in competitive code for larger benchmarks.

Biography

Michael Budde is an MSc student at University of Copenhagen, DIKU. The talk is based on joined work with Martin Dybdal and Martin Elsman, DIKU. The work is presented at the ARRAY’15 workshop in Portland, Oregon, later in June.

Host: Martin Elsman

Talk on kdb+ by Conor Tworny

2015-05-12T00:00:00+00:00

HIPERFIT lunch talk: An Introduction to kdb+

Presenter: Conor Twomey, First Derivatives/Kx Systems.

Time: Tuesday May 12, 2015

Place: The APL Meeting Room at DIKU (Universitetsparken 5, Building B, Ground Floor)

Abstract:

Kx offers kdb+, a high-performance column-store database with a built-in expressive query and programming language, q. Used as a central repository to store time-series data within an enterprise, kdb+ supports real-time analysis of billions of records and fast access to terabytes of historical data. It also:

provides seamless scalability;
runs on industry standard server platforms;
is top-ranked in third-party benchmark testing;
has an extremely small footprint, which makes installation and maintenance fast and straightforward;
easily accommodates available APIs for connectivity to major external systems and modules;
requires fewer developers, contributing to lower total cost of ownership.

Biography

As Vice President of Kx Services at First Derivatives plc, Conor’s area of expertise is in building bespoke trading and risk systems and managing near-shore development centres of excellence out of Newry, Northern Ireland. He has previously worked with Goldman Sachs (New York), Citigroup (London), Nomura (New York) and Barcap (Toyko). In 2011 he was shortlisted to the final eight for the GradIreland Graduate Employee of the Year award. Conor is the editor-in-chief of FD’s kdb+ whitepaper series, has had a paper published in the Journal of Physics, is the Dublin leader of a World Economic Forum initiative called the “Global Shapers” and is a power forward for the Newry Flyers basketball team.

Host: Fritz Henglein

Conference on Exchange Rate Stability

2015-05-08T00:00:00+00:00

On June 25, 2015, University of Copenhagen and HIPERFIT Researchers are hosting a conference on exchange rate stability.

Introduction

The high degree of instability and sudden movements in exchange rate markets have highlighted their flaws and resulted in a series of problems which range from private sector exposure to governmental issues. The conference will address these matters from various angles as well as introduce the Trade-Weighted Equilibrium Exchange Rate System from GCU. The conference is relevant for participants from the central bank communities, governmental agencies, and the financial sector, together with academic researchers and decision-makers within monetary affairs.

More information

NEWS: See the detailed program.

For more information about the venue, including information about speakers, please consult the formal invitation, which is available for download.

Talks on Streaming in Data Parallelism

2015-04-28T00:00:00+00:00

HIPERFIT Seminar Talks: Streaming in Data Parallelism

Presenters: Andrzej Filinski (Associate Professor, DIKU) and Frederik Meisner Madsen (PhD Student, DIKU)

Time: Tuesday, April 24, 15:00-16:00

Place: Aud 07 (HCØ, Universitetsparken 5)

Abstracts:

The presentation will consist of two shorter talks:

Streaming NESL (speaker: Andrzej Filinski, associate professor, DIKU)

Streaming NESL, under development in the HIPERFIT Center, is a refinement of the nested data-parallel functional language NESL. It extends NESL’s flattening-based implementation strategy and language-integrated cost model for time (in the form of work and step measures) to also achieve predictable (and optimistic) bounds for space usage. This is done by avoiding complete materialization of intermediate vector-typed values that do not actually require random access. Instead, such values are transparently materialized and processed in “chunks”, of size proportional to the available parallel computing resources - from SIMD instructions on single-core CPUs to large GPGPUs.

We give an overview of the status of the project, including a summary of the source language and cost model, a sketch of the implementation strategy in terms of a chunked-dataflow abstract machine (which may also be relevant to other language-technology projects within HIPERFIT), and an outline of some current challenges, both theoretical and practical.

Streaming Accelerate (speaker: Frederik Meisner Madsen, PhD student, DIKU)

The relatively mature embedded language Accelerate brings the power of hardware-accelerated, shape-polymorphic programming with regular arrays to Haskell. Extending Accelerate with streaming functionality, along the lines of our current research, will hypothetically improve its usability and performance for many important use-cases. This talk presents an extension to Accelerate that enables streaming by lifting a non-trivial subset of the language from working on arrays to working on sequences of arrays. The programmer obtains a sequence by slicing an array into a sequence of sub-arrays, and may return to ordinary Accelerate through reduction. Sequence expressions are evaluated efficiently in fixed-size chunks by reusing memory and compiled kernels, allowing sequences to scale in length without requiring additional computing resources except computation time.

Talk on The Fundamental Theorem of Derivative Trading

2015-04-14T00:00:00+00:00

HIPERFIT Seminar Talk: The Fundamental Theorem of Derivative Trading - Exposition, Extensions, and Experiments

Presenter: Rolf Poulsen, Professor, Math/University of Copenhagen

Time: Tuesday, April 14, 15:00-16:00

Place: Aud 07 (HCØ, Universitetsparken 5)

Abstract:

When estimated volatilities are not in perfect agreement with reality, delta hedged option portfolios will incur a non-zero profit-and-loss over time. There is, however, a surprisingly simple formula for the resulting hedge error, which has been known since the late 90s. We call this The Fundamental Theorem of Derivative Trading. This paper is a survey with twists of that result. We prove a more general version of it and discuss various extensions (including jumps) and applications (including deriving the Dupire-Gyöngy-Derman/Kani formula). We also consider its practical consequences both in simulation experiments and on empirical data thus demonstrating the benefits of hedging with implied volatility. The paper is available online.

Talk by Grigori Fursin on Collaborative and Reproducible Computer Engineering

2015-03-18T00:00:00+00:00

Collective Knowledge Project: Towards Collaborative and Reproducible Computer Engineering

Grigori Fursin, CTO, cTuning foundation.

Time and location: March 19, 15:30-16:30. The APL meeting room, DIKU, Universitetsparken 5, Building B, 2100 Copenhagen Ø.

To attend the talk, please signup by email to Martin Elsman (mael at di.ku.dk).

Abstract:

Designing novel computer systems and optimizing their software is becoming too tedious, ad-hoc, time consuming and error prone due to an enormous number of available design and optimization choices. Empirical auto-tuning combined with run-time adaptation and machine learning has been demonstrating some potential to address the above challenges for several decades but is still far from widespread production. Main reasons include unbearably long exploration and training times, ever changing tools and their interfaces, lack of a common experimental methodology, and lack of unified mechanisms for knowledge building and exchange apart from publications where reproducibility of results is often not even considered.

In this talk, I will present our community-driven approach and the 4th version of our public infrastructure and repository to preserve, systematize and share knowledge and experience about program optimization. Our framework helps to describe and preserve the whole experimental setup with all related artifacts (benchmarks, kernels, data sets, libraries, tools) in a reproducible way to automate and crowdsource optimization space exploration and learning. Any unexpected behavior is recorded and analyzed using shared data mining and predictive modeling plugins or is exposed to the community for collaborative explanation. Collected knowledge can be used to validate past research techniques or can be extrapolated to predict better optimizations, run-time adaptation scenarios and hardware designs. During the past 5 years, this approach has been extensively validated with our industrial partners, and helped initiate a new publication model where experiments and artifacts are validated and improved by the community. A beta version of the new framework is available at http://github.com/ctuning/ck.

Bio:

Grigori Fursin is a founder and CTO of the cTuning foundation. In the past, he was a tenured research scientist at INRIA Saclay and was co-founder of the Intel Exascale Lab in France. Grigori has an interdisciplinary background in computer engineering, physics, electronics, machine learning and mathematics. He obtained a PhD in Computer Science from the University of Edinburgh in 2004.

Grigori pioneered machine-learning based program auto-tuning and hardware co-design combined with crowdsourcing and run-time adaptation. In 2008, he established a public repository of optimizations knowledge (cTuning.org) to initiate collaborative and reproducible R&D in computer engineering. Since then, his techniques and tools have been used and extended in multiple industrial projects together with ARM, IBM, Intel, Synopsys and ST. In 2012, Grigori received an INRIA award and 4-year fellowship for “making an outstanding contribution to research”. He is leading an Artifact Evaluation initiative for CGO and PPoPP while developing an open research SDK to preserve, structure and reuse knowledge about program and architecture optimization: http://github.com/ctuning/ck.

Talks by Rolf Poulsen and Martin Elsman at GetF'IT

2014-11-26T00:00:00+00:00

On November 25, 2014, Rolf Poulsen and Martin Elsman participated with presentations at the monthly Get F’IT meeting. For more information, please consult the Get F’IT Web Page.

HIPERFIT Workshop

2014-11-04T00:00:00+00:00

2014 HIPERFIT Workshop for Partners and Faculty

On December 10, 2014, HIPERFIT is hosting a workshop for HIPERFIT partners and faculty. At the workshop, HIPERFIT faculty and PhD students will present the latest developments in the various HIPERFIT project developments. The workshop also offers HIPERFIT partners the possibility to present opportunities for involving HIPERFIT researchers in obstacles encountered by the partners with respect to high-performance computational problems in the finance domain. A separate session is therefore devoted to partners for presenting such encounters in the finance industry.

The Venue

Kurhotel Skodsborg, Skodsborg Strandvej 139, 2942 Skodsborg (45585800)

Schedule December 10, 2014

08:30

Breakfast served

08:55

Welcome (Fritz Henglein)

09:00

Compiling for Parallel Computations (Chair: Ken Friis Larsen)

Futhark - An Array Language for Data-parallel Execution (Troels Henriksen, DIKU)
Streaming Nested Data Parallelism (Frederik Meisner Madsen, DIKU)
Compiling APL for Data-Parallel Execution through a Typed Intermediate Language (Martin Dybdal, DIKU)

10:15

Break

10:30

Industry Input (Chair: Fritz Henglein)

The MC2 Challenge in Risk (Carl Balslev Clausen, SimCorp)
Big Data and Advanced Analytics in a Financial Institution (Nadeem Gulzar, Danske Bank) CANCELLED
"Carry that Load" — Risk-minimization in Electricity Markets (Martin Jönsson, IMF)

11:30

Break

11:45

Bohrium and Big-data (Chair: Brian Vinter)

Bohrium - Bridging High Performance and High Productivity (Kenneth Skovhede, NBI)
Effective Interoperability (Simon Lund, NBI)
Transactional Partitioning: A New Abstraction for Main-memory Databases (Vivek Shah, DIKU)

13:00

Lunch

14:00

Household Finance and Risk Management (Chair: Mogens Steffensen)

Household Finance Problems approached by Numerical Methods (Mai-Britt Nordfang, IMF)
Rethinking Exchange Rate Risk Management (Rolf Poulsen, IMF)
Calibration of the Local Volatility Function (Lykke Rasmussen, IMF) CANCELLED

15:15

Break

15:30

Financial Contracts and Valuation (Chair: Andrzej Filinski)

A Prototype Framework for Parallel Valuation and Risk Calculation (Martin Elsman, DIKU)
Certified Management of Financial Contracts (Patrick Bahr, DIKU)
Financial Benchmarks for GPGPU Compilation (Cosmin Oancea, DIKU)

16:45

Thanks - feedback

17:15

Chat and a drink

18:30

Dinner

A table of participants for the workshop on December 10, and for the discussion and planning sessions (HIPERFIT staff only) on December 11, appears below. The schedule for December 11 appears at the very end of the page.

Participation in the workshop is by invitation only, but please contact the HIPERFIT management if you think you should have received an invitation.

As always, please visit http://www.hiperfit.dk for news and information about published papers, our Tuesday lunch meetings, and news in general from the HIPERFIT research center.

Abstracts

09:00-10:15 Compiling for Parallel Computations

Futhark - An Array Language for Data-parallel Execution (Troels Henriksen)

Compiling real-world applications to efficient parallel code, especially when targeting many-core hardware, requires an optimiser and code generator that can exploit a combination of high-level invariants and low-level optimisations. We present a purely functional core language, named Futhark, intended as a compilation target for higher-level languages, that supports nested map-reduce parallelism on regular array, but also a set of “imperative” constructs, such as in-place updates and do-loops. We also report in-progress work on the Futhark optimizing compiler, which is thought to allow a gradual transformation of the program, within the same representation, towards low-level, imperative-like code, which can then be efficiently and straightforwardly translated to the parallel assembly languages of our time, such as CUDA and OpenMP.

Streaming Nested Data Parallelism (Frederik Meisner Madsen)

A clever compiler relieves the programmer from many technical details, facilitating rapid development, code re-use and hopefully high performance. If the compiler obeys a simple optimistic time-space cost model, the hope becomes a guarantee. This talk aims to identify two key challenges related to functional data-parallel language design and costing: Nested data-parallelism and excessive space usage when too few processors are available. To solve these challenges, a language based on first-class sequences and dataflow execution is presented. The language supports nested data parallelism, obeys a time-space cost model and can run on many different architectures including single-core machines and GPGPUs. Experiments show promising results compared to similar languages, and the ideas are currently being implemented in the next release of Accelerate, a Haskell library for data-parallel execution on GPGPUs.

Compiling APL for Data-parallel Execution through a Typed Intermediate Language (Martin Dybdal)

We present a compiler for a subset of the classical array programming language APL. The compiler translates APL programs into a typed array intermediate language, called TAIL, which is equipped with a clear type system and an operational semantics. We demonstrate, for a few benchmarks, the feasibility of compiling TAIL further into code that can execute on parallel architectures, such as GPGPUs.

10:30-11:30 Industry Input

The MC2 Challenge in Risk (Carl Balslev Clausen, SimCorp)

Monte Carlo methods are required for precise valuation of contracts that have a certain level of complexity. Monte Carlo methods are also required for various Risk calculations when large shocks and movements are modelled. Combining Monte Carlo Risk with Monte Carlo valuation we encounter the double Monte Carlo challenge, aka the MC2 challenge. We report on specific use cases and benchmarks from the industry.

Big Data and Advanced Analytics in a Financial Institution (Nadeem Gulzar, Danske Bank) CANCELLED

In Danske Bank a lot of effort is being put into utilizing new sources of data (social, weblogs, etc.) to gain a better understanding of the customers and their behavior. To handle and process the extremely large amount of data, we require an equally large amount of processing power. There are many ways to address this ranging from using lots of cash to optimizing techniques, etc.

“Carry that Load” — Risk-minimization in Electricity Markets (Martin Jönsson, IMF)

In this talk we look at a specific hedging problem of a contract in the derivatives market for power contracts. We suggest a simple hedging approach that relies on expected loss minimisation of our hedged position, and to calculate our optimal hedge we suggest a simple model for the spot price and consumption of electricity. We test our approach on empirical data obtained from an energy trading company, and compare our strategy to the hedging approach currently used by the company.

11:45-13:00 Bohrium and Big-data

Bohrium - Bridging High Performance and High Productivity (Kenneth Skovhede)

In this talk I introduce the Bohrium runtime system and show how it simultaneously enables rapid development and high performance execution. I then summarize our current results with the Bohrium approach for CPU, GPGPU and cluster execution. Finally I present the state of my current work on FPGA execution in Bohrium.

Effective Interoperability (Simon Lund)

Programming languages and models are essential building blocks for computer-based problem-solving. The general challenge is about modeling concepts and abstractions for a given problem and expressing it programmatically to obtain a solution. A challenge that no single programming model or language can solve for the general case when put under the constraint of also efficiently utilizing hardware. However, many languages and models exist which solves classes of problems efficiently. Interoperability provides the means of combining them. In this talk, this approach is exemplified through the use of the emerging parallel programming language Chapel and the long-lived productivity language Python. The interoperability approach is exemplified by introducing Chapel and comparing strengths/weaknesses of the two languages and how to overcome them through the use of interoperability.

Transactional Partitioning: A New Abstraction for Main-memory Databases (Vivek Shah)

Online Transaction Processing (OLTP) applications are making new demands from OLTP databases on variety, performance and application cost frontiers. Existing programming models provided by current OLTP databases do not meet all the demands of these classes of applications. The talk presents a new programming model (transactional partitioning) that needs to be targetted by OLTP databases (especially main-memory databases) that would allow them to meet these new application demands. The talk also outlines the challenges and their possible solutions that are being investigated while building the prototype system.

14:00-15:15 Household Finance and Risk Management

Household Finance Problems approached by Numerical Methods (Maj-Britt Nordfang)

Closed-form solutions to mathematical financial problems can only be obtained when various strict assumptions are met. As a consequence, the problems where explicit closed-form solutions exist are usually highly stylized. In this talk I will present a particular problem related to the financial decisions of individual households that cannot be solved on closed form and discuss numerical approaches to finding the solution.

Rethinking Exchange Rate Risk Management (Rolf Poulsen)

We describe some theoretical and practical (from computational to legal) issues in the construction of an exchange-rate risk management system that is feasible for small to medium enterprises. This work is a collaboration with http://www.gcureports.com/ and you will hear how the system was almost endorsed by the Pope.

Calibration of the local volatility function (Lykke Rasmussen) CANCELLED

In this talk we present the results of a study and comparison of 5 different methods for calibrating local volatility functions.

15:30-16:45 Financial Contracts and Valuation

A Prototype Framework for Parallel Valuation and Risk Calculation (Martin Elsman)

We give an overview of a simple open architecture and framework for demonstrating parallel execution of contract and portfolio valuation, as well as risk calculations. The framework, which is open source, builds on a contract language for specifying and managing multi-party contracts and on an efficient parallel pricing engine.

Certified Management of Financial Contracts (Patrick Bahr)

We present a contract language for describing complex financial contracts. Contracts modelled in our language are analysed and transformed for management according to a precise cash-flow semantics, formalised and verified using the Coq proof assistant.

Financial Benchmarks for GPGPU Compilation (Cosmin Oancea)

We present three studies of array-based applications from the financial domain, all suitable for GPGPU execution. The studies have resulted in three concrete application benchmarks for which the available parallelism is described by nested map-reduce functional combinators. We also describe the influence of the invariants and code transformations that govern the main trade-offs of the rich, dataset-sensitive optimisation space. The application benchmarks targets both CPUs and GPGPUs and we provide useful insight into the language constructs and compiler infrastructure capable of expressing and optimising such applications.

Participants

Name	Dec 10	Dinner Dec 10	Dec 11	Room
Andrzej Filinski (DIKU)	1	1	1	1
Brian Vinter (NBI)	1	1	1	1
Cosmin Oancea (DIKU)	0	1	1	1
Danil Annenkov (DIKU)	1	1	1	1
Frederik Meisner Madsen (DIKU)	1	1	1	1
Fritz Henglein (DIKU)	1	1	1	1
James Avery (NBI)	1	1	1	1
Jost Berthold (DIKU)	1	1	1	1
Ken Friis Larsen (DIKU)	1	0	1	0
Kenneth Skovhede (NBI)	1	1	1	1
Lykke Rasmussen (IMF)	0	0	0	0
Mads R. B. Kristensen (NBI)	1	1	1	1
Maj-Britt Nordfang (IMF)	1	1	1	1
Marcos Vaz Salles (DIKU)	1	1	0	0
Martin Dybdal (DIKU)	1	1	1	1
Martin Elsman (DIKU)	1	1	1	1
Martin Jönsson (IMF)	1	1	0	0
Mogens Steffensen (IMF)	1	1	1	0
Oleksandr Shturmov (DIKU)	1	1	1	0
Patrick Bahr (DIKU)	1	1	1	1
Rolf Poulsen (IMF)	1	1	1	1
Simon Ellersgaard (IMF)	0	0	0	0
Simon Lund (NBI)	1	1	1	1
Thomas Jensen (DIKU)	1	0	0	0
Troels Blum (NBI)	0	0	0	0
Troels Henriksen (DIKU)	1	1	1	1
Vivek Shah (DIKU)	1	1	0	0
Anette Broløs (CFIR)	1	1	0	0
Anders Pall Skött (CFIR)	1	0	0	0
Christian Hjersing (SimCorp)	1	0	0	0
Jens Ehlers (SimCorp)	1	1	0	0
Carl Balslev Clausen (SimCorp)	1	1	0	0
Henrik Nygaard Jensen (Danske Bank)	1	1	0	0
Nadeem Gulzar (Danske Bank) ILL	0	0	0	0
Uwe Heissner (Nordea)	1	0	0	0
Manuel Torrealba (Nordea)	1	0	0	0
Total	31	26	20	17

Schedule December 11, 2014 (HIPERFIT staff only)

On December 11, 2014, HIPERFIT staff will meet and discuss feedback and plan future activities and projects within HIPERFIT. Here is a draft schedule.

08:30: Breakfast served
08:55: Outline for the day (Plenum)
09:00: Session 1 (Groups)
12:00: Lunch
13:00: Session 2 (Groups)
15:00: Break
15:15: Group Presentations (Plenum)
16:30: End

New HIPERFIT PhD Scholarship

2014-09-22T00:00:00+00:00

The HIPERFIT research center seeks one new excellent PhD student for working within the area of High-Performance Functional Programming.

For information about the application procedure, and the scholarship in general, please consult the general announcement. The application deadline is (was: October 10) October 17, 2014 (23:59, timezone: LINT = UTC+14).

Below follows some more detailed information about the scholarship.

PhD scholarship in High-Performance Functional Programming

The PhD student will be working in the Research Center for Functional High-Performance Computing for Financial Information Technology (www.hiperfit.dk). The PhD scholarship is funded by the Danish Council for Strategic Research and will be carried out at DIKU in formal collaboration with the Department of Mathematical Sciences and the Niels Bohr Institute, University of Copenhagen and several finance it-partners in the Copenhagen area. The objective of the PhD project is to develop, implement and evaluate techniques related to executing, efficiently and safely, high-level, functional specifications of both numeric and symbolic computations on modern, highly parallel computational platforms such as multi- and many-core CPUs, GPGPUs and FPGAs.

The candidate must – in addition to the qualifications listed below – have a solid academic background in the relevant areas of computer science, as demonstrated by academic courses and project/thesis work related to several of the following areas: Functional programming, programming language theory, and compilers. Experience with either high-performance computing, mathematical finance, domain-specific language or parallel algorithms is desirable, but not required.

The position is available from 1st November 2014 or as soon as possible thereafter. Enquiries about the position can be made to Professor Fritz Henglein, e-mail henglein@di.ku.dk, phone +45 30589576.

Talk by Amos Robinson on Fusion

2014-08-18T00:00:00+00:00

Fusing Filters with Integer Linear Programming

Amos Robinson, University of New South Wales, Sydney, Australia.

Time and location: August 26, 12:15-13:00. The APL meeting room, DIKU, Universitetsparken 5, Building B, 2100 Copenhagen Ø.

To attend the talk, please signup by email to Martin Elsman (mael at di.ku.dk).

Abstract:

The key to compiling functional, collection oriented array programs into efficient code is to minimise memory traffic. Greedily fusing array operations together can reduce some memory traffic, but finding the fusion clustering that produces the least amount of memory traffic turns out to be NP-hard. Integer linear programming (ILP) can be used to solve NP-hard problems, by finding the minimal solution for a given cost metric. Previous work has used ILP to find fusion clusterings for certain graphs, however these approaches cannot handle size-changing operations such as filter. In this talk, I will show how size-changing operations differ, and how we have extended previous work to support such operations.

Bio:

Amos Robinson is a PhD student at University of New South Wales, Sydney, Australia. His focus is on optimisation and fusion in purely functional programming languages such as Haskell. He was also the winner of the inaugural Sydney Coq Fight, a theorem proving tournament.

HIPERFIT Seminar on Automated Solution of Differential Equations

2014-04-23T00:00:00+00:00

Speaker: Anders Logg, Professor of Computational Mathematics at Chalmers University of Technology, Gothenburg.

Time: Tuesday, May 6, 2014, 15:00-16:00

Place: Auditorium 5, Building C, HCØ, Universitetsparken 5, 2100 Copenhagen Ø.

Note: Drinks and snacks will be served after the talk.

Talk Abstract

For more than two centuries, partial differential equations have been an indispensable tool for scientists and engineers for modeling and understanding a wide range of physical phenomena. The solution of a partial differential equation is often a very challenging and time-consuming task, involving mathematical analysis, numerical analysis, and the development of sophisticated computer programs.

In this talk, I demonstrate how the solution process can be completely automated, to the point where complicated systems of nonlinear partial differential equations can be solved with ease in a web browser.

This work is a joint effort between research groups at Chalmers, University of Chicago, University of Cambridge, KTH and Simula Research Laboratory (in order of appearance) as part of the FEniCS Project, a collaborative project for the development of innovative concepts and tools for automated computational modeling, with a particular focus on automated solution of differential equations by finite element methods, based on domain specific languages, code generation, and high-performance parallel computing.

Bio

Anders Logg is a professor of Computational Mathematics at Chalmers University of Technology in Gothenburg. His main research interest is in the numerical solution of partial differential equations, in particular adaptive finite element methods and the development of efficient computer programs for automated multiphysics simulation. He is one of the founders and core developers of FEniCS, an international project for the development of free software for automated solution of differential equations.

MSc Thesis defence on Multireduce and Multiscan on Modern GPUs

2014-04-09T00:00:00+00:00

MSc thesis defense by Marco Eilers.

Time: Friday, 11 April 2014, 14:00-15:00.

Place: Auditorium “Lille UP1”, Universitetsparken 1.

Abstract

With the introduction of platforms like CUDA and OpenCL, the superior computing power of modern GPUs compared to CPUs is used more and more often to accelerate general purpose computations. The multireduce and multiscan operations, generalizations of the ordinary reduce and scan, have immediate applications for common algorithmic problems and show potential to be used as simple, deterministic building blocks for parallel algorithms.

We first discuss sequential CPU algorithms for both of these operations, focusing especially on problems resulting from poor use of various caches, and point out ways to solve them. For GPUs, we systematically examine three main groups of algorithms: parallel adaptations of the sequential algorithm, GPU adaptations of existing PRAM algorithms, and sort-based conversions to simpler problems. For each of these possibilities, we discuss how the GPU memory hierarchy can be used for best performance, and which additional algorithmic improvements can be made for operators which are commutative as well as associative.

We conclude that multireduce can be implemented efficiently on GPUs and can therefore be recommended as a parallel primitive, whereas the multiscan cannot. For histogramming, a special case of multireduce, our algorithmic improvements for the multireduce can be used to achieve a 40% speedup over the best existing algorithm.

Supervisor: Andrzej Filinski

External examiner: Peter Sestoft, ITU

Two New HIPERFIT PhD Scholarships

2014-03-28T00:00:00+00:00

The HIPERFIT research center now seeks two new excellent PhD students for working within the area of High-Performance Functional Programming.

For information about the application procedure, and the scholarships in general, please consult the general announcement. The application deadline is April 22, 2014 (23:59, timezone: LINT = UTC+14).

Below follows some more detailed information about the scholarships.

PhD scholarship(s) in High-Performance Functional Programming

The objective of the PhD project is to develop, implement and evaluate techniques related to executing, efficiently and safely, high-level, functional specifications of both numeric and symbolic computations on modern, highly parallel computational platforms such as multi- and many-core CPUs, GPGPUs and FPGAs.

The candidate must - in addition to the qualifications listed above - have a solid academic background in the relevant areas of computer science, as demonstrated by academic courses and project/thesis work related to several of the following areas: Functional programming, programming language theory, and compilers. Experience with either high-performance computing, mathematical finance, domain-specific language or parallel algorithms is desirable, but not required.

The position is available from 1 July 2014 or as soon as possible thereafter.

HIPERFIT Fall Activities

2013-11-19T00:00:00+00:00

A series of activities are happening in the HIPERFIT research center in the fall. Since the PhD Workshop in August, HIPERFIT researchers have worked together with the involved partners on topics such as contract modeling, household finance, GPGPU execution of model calibration code, and scalable portfolio analysis.

In September, HIPERFIT researchers were presenting two strands of work on the 2nd ACM SIGPLAN Workshop on Functional High-Performance Computing, held in Boston, USA. The first paper, authored by Troels Henriksen and Cosmin E. Oancea, introduced a fusion technique for avoiding unnecessary storage of intermediate results in array languages. The second paper, authored by Frederik M. Madsen and Andrzej Filinski, introduced a streaming model for nested parallelism, which allows for splitting up the processing of streams of data in a way that allows for high utilization of parallelism. Another paper, authored by Martin Elsman and Anders Schack-Nielsen (SimCorp), has been accepted for publication, in January 2014, at the International Symposium on Practical Aspects of Declarative Languages (PADL’14). This papers introduces the concept of Typelets - A Rule-Based Evaluation Model for Dynamic, Statically Typed User Interfaces. The papers are available from the HIPERFIT Publication page.

In December, HIPERFIT researchers and students will be going on a retreat where they will work together on a series of topics, including credit value adjustments, household finance, fusion for streaming, and a backend for the HIPERFIT L language.

As a final note, we would like to point you to the event on using “F# in Finance”, which is held on November 25, 2013 and December 11, 2013 in London and New York, respectively. Please consult the web-site http://www.fsharpinfinance.com for more information about the events, which demonstrate the effective use of functional programming in the finance industry.

Internal HIPERFIT PhD Workshop

2013-07-31T00:00:00+00:00

See below for updated information for speakers.

The 2013 HIPERFIT PhD Workshop is held on August 13, 2013. The workshop is internal for IMF, DIKU, NBI employees.

Location: The workshop is held at DAC& Meeting center at Strandgade 27B, DK 1401, Copenhagen K. Breakfast and coffee is served from 9:30.

Program (preliminary)

09:30 Breakfast and coffee available
09:55 Welcome by Martin Elsman
10:00 Frederik Meisner Madsen (first-year). Supervisor: Andrzej Filinski.
10:45 Fritz Henglein. Towards a New Computational Model for Parallelism.
11:30 Break
11:45 Simon Lund (second year). Supervisor: Brian Vinter.
12:30 Lunch Break
13:30 Lykke Rasmussen (first year). Supervisor: Rolf Poulsen.
14:15 Troels Blum (second year). Supervisor: Brian Vinter.
15:00 Break
15:15 Vivek Shah (first year). Supervisor: Marcos Vaz Salles.
16:00 Maj-Britt Nordfang (first year). Supervisor: Mogens Steffensen.
16:45 Break
17:00 Cosmin Oancea (Postdoctoral researcher)
17:30 A drink in the bar
18:00 Dinner

Participants (preliminary)

James Avery, Jost Berthold, Troels Blum, Martin Elsman, Andrzej Filinski, Troels Henriksen, Mads Ruben Burgdorff Kristensen, Ken Friis Larsen, Fritz Henglein, Simon Lund, Frederik Meisner Madsen, Simon Ellersgaard Nielsen, Maj-Britt Nordfang, Cosmin Oancea, Rolf Poulsen, Lykke Rasmussen, Marcos Vaz Salles, Kenneth Skovhede, Mogens Steffensen, Brian Vinter

Instructions for Speakers

Each speaker should prepare a 30 minute presentation, including (1) an outline of the PhD topic (with some technical detail) and (2) an overview of already obtained results, research plans, plans for courses, traveling abroad, and so on. The 30 minute presentation is followed by a 15 minutes discussion session where participants can ask questions to the speaker and the speaker can ask questions to the participants.

PhD students should read Section 4.1 of the Rules and Guidelines for the PhD Programme and discuss with their respective supervisors how the PhD Workshop may help the students fulfill the PhD requirements.

HIPERFIT Seminar on July 2, 2013

2013-06-27T00:00:00+00:00

The HIPERFIT July 2013 seminar is planned for Tuesday, July 2 from 14:00 to 15:30 at the DIKU small auditorium (Lille Aud), where we will have two presentations by Troels Henriksen and Brian Vinter:

14:00-14:30. A T2 Graph-Reduction Approach To Fusion. (Troels Henriksen, HIPERFIT, DIKU).
14:30-15:00. BPU Simulator. (Brian Vinter, HIPERFIT, NBI).
15:00-15:30. A light refreshment, served by HIPERFIT.

Abstracts for the talks follow below.

We look forward to see you at the seminar!

Best Regards,

Martin Elsman

HIPERFIT Center Manager

A T2 Graph-Reduction Approach To Fusion

Troels Henriksen, HIPERFIT, DIKU

14:00-14:30 July 2, 2013

The (DIKU) small auditorium at Universitetsparken 1 (UP1)

Abstract: Fusion is one of the most important code transformations as it has the potential to substantially optimize both the memory hierarchy time overhead and (sometimes asymptotically) the space requirement. In imperative languages, the legality of loop-fusion is typically verified by dependency analysis on arrays applied at loop-nest level. Such analysis, however, has often been labeled as “heroic effort” and, if at all, is supported only in its simplest and most conservative form in industrial compilers. In functional languages, fusion is naturally and more easily derived as a producer-consumer relation between program constructs that expose a richer, higher-order algebra of program invariants, such as the map-reduce list homomorphisms. Related implementations in the functional context typically apply fusion only when the to-be-fused producer is used exactly once, i.e., in the consumer. This guarantees that the transformation is conservative: the resulting program does not duplicate computation. We show that the above restriction is more conservative than needed, and present a structural-analysis algorithm, inspired from the T1 -T2 transformation for reducible data flow, that enables fusion even in some cases when the producer is used in different consumers and without duplicating computation. We report an implementation of the fusion algorithm for a functional-core language, named L0 , which is intended to support nested parallelism across regular multi-dimensional arrays. We succinctly describe L0’s semantics and the compiler infrastructure on which the fusion transformation relies.

Troels Henriksen is an MSc student at DIKU, employed in the HIPERFIT research center, and working on the implementation and design of the HIPERFIT parallel-functional programming language L.

BPU Simulator

Brian Vinter, HIPERFIT, NBI

14:30-15:00 July 2, 2013

The (DIKU) small auditorium at Universitetsparken 1 (UP1)

Abstract: A number of scientific applications start their life as a Matlab prototype that is later re-implemented in a low level programming language, typically C++ or Fortran for the sake of performance. Bohrium is a project that seeks to eliminate both the cost and the potential errors introduced in that process. Our goal is to support all execution platforms, and in this work we introduce the Bohrium Processing Unit, BPU, which will be the FPGA backend for Bohrium. The BPU is modeled as a PyCSP application, and the clear advantages of using CSP for simulating a new CPU is described. The current PyCSP simulator is able to simulate 220 Monte Carlo simulations in less than 35 seconds in the smallest BPU simulation. Joined work with Martin Rehr and Kenneth Skovhede, Niels Bohr Institute.

Brian Vinter is professor at Niels Bohr Institute (NBI) and has many years of research experience within the areas of grid computing, supercomputing, multicore architectures, and metods for transparent utilization of parallellism.

Global Systems Science

2013-06-24T00:00:00+00:00

Martin Elsman gave a presentation titled Global Systems Science meets Programming Languages and Systems at the Global Systems Science conference in Brussels on June 11; see the abstract for the talk below.

After returning to DIKU, Martin gave an overview presentation of what Global Systems Science is all about. See also the Global Systems Science blog and the Global Systems Science Synthesis paper.

Global Systems Science meets Programming Languages and Systems

Martin Elsman

In this talk, we demonstrate how functional programming and domain specific languages, in particular, can be useful for effectively deriving performance efficient programs and systems. As an example, we outline a system for specifying financial contracts (used in practice by the financial industry) and demonstrate the effect of applying programming language technology to derive tools for pricing contracts efficiently on modern parallel hardware. We argue that research in managing and querying big data and efficiently performing big computations (simulations), as for instance carried out by the HIPERFIT research center, is a central ingredient of the development of a Global Systems Science. PDF.

Bohrium Day

2013-05-17T00:00:00+00:00

On Friday May 24, 2013, Brian Vinter’s group at the Niels Bohr Institute is arranging a Bohrium workshop called the “Bohrium Day”.

Place: at NBI - see Bohrium Day venue information.

Registration: Please send an email to “vinter at nbi.dk”.

Program

See the detailed program.

Backgound

Bohrium is a “Plug-and-play” high-performance backend targeting CPU clusters and other parallel architectures, such as GPGPUs. The goal of the Bohrium project is to provide a runtime environment for efficiently executing vectorized applications using a variety of programming languages (Python/NumPy, C#, F#) on a variety of platforms (Linux, Windows and MacOSX).

By collecting bytecode instructions at runtime and by generating (and caching) efficient GPU kernels, programmers may utilize end-user’s clusters of GPUs and multi-core CPUs without having to program these new architectures explicitly.

The Bohrium project is a part of the HIPERFIT research center.

HIPERFIT Get FIT Event

2013-05-13T00:00:00+00:00

HIPERFIT is on the agenda for the next Get F’IT meeting on May 28, 2013, where Associate Professor Martin Elsman will give an overview of the center activities and Professor Rolf Poulsen and Postdoc Cosmin E, Oancea will give presentations of current research.

Facilitator: Director Anette Broløs, CFIR.

Place: Finanssektorens Hus, Amaliegade 7, København K.

Date: Tuesday, may 28, 2013, kl. 16.00-18.00.

Price: Attendance is free, but registration is required.

HIPERFIT Talks at Events for the Danish Quant Network and CFIRs Company Forum

2013-04-22T00:00:00+00:00

Martin Elsman was giving a presentation about HIPERFIT activities on April 16 for the Danish Quant Network (under CFA Denmark) and a shorter version of the same presentation on April 17 for the CFIR Company Forum.

February Seminar

2013-01-30T00:00:00+00:00

The HIPERFIT February 2013 seminar is planned for Tuesday, February 5 from 15:00 to 16:30 at the DIKU small auditorium (Lille Aud), where Andreas Magnussen from Velocytech is giving a talk:

15:00-16:00. FPGA technology for High performance computing with low latency (Andreas Magnussen, CTO, Velocytech).
16:00-16:30. A light refreshment, served by HIPERFIT.

More information about the talk is given below.

We look forward to see you at the seminar!

Best Regards,

Martin Elsman

HIPERFIT Center Manager

FPGA technology for High performance computing with low latency

Andreas Magnussen, Velocytech

15:00-16:00 February 5, 2013

The (DIKU) small auditorium at Universitetsparken 1 (UP1)

Velocytech develops IP for high speed communication and connectivity in FPGA and ASIC. The FPGA sizes follows Mores law. The FPGA can be used to implement high performance computation logic and high speed and low latency communication. This opens up for a number of new applications which require high network performance.

The presentation will give an overview of the type of problems that can be addressed in FPGA systems today and what network and system performance improvements that can be achieved. We will also discuss design and test issues that relates to FPGA development.

Scientific host: Brian Vinter, HIPERFIT

Andreas Magnussen has 30 years of professional international working experience in the electronic industry (data communication and telecommunication), where he has held senior positions within engineering (hardware, software, ASIC and systems), marketing and management.

Education:

E.E. (elektronik mekaniker) in 1984 from Brüel & Kjær A/S.
M.Sc. E.E. (civilingeniør) in 1991 from DTU, Technical University of Denmark.
Ph.D. in 1997 from DTU (ATM switching system).
Executive MBA (MMT) in 2005 from DTU&TEM.

Key experience and credentials:

Tele and data networks, software systems, communication protocols, communication components, system design, marketing and management. Architecture, design, implemented and tested state of the art ATM layer chip 155Mbps with integrated switching and line functions (1995). Completed a high number of complex system designs in tele and data communications. Re-defined ASIC strategy and ASIC team rebuilding at Intel (2001). CEO at IPBlaze (2001). CTO at Velocytech (2012).

Brüel & Kjær (1982-1991)
DTU (1991-1992)
NKT Elektronik/DSC/Tellabs (1992-2000)
Intel (2000-2001)
IPBlaze A/S (2001-2011)
Velocytech APS (2012-now)

Talk by Robert Bernecky on APEX

2012-10-12T00:00:00+00:00

Robert Bernecky is visiting DIKU on October 22, 2012. Robert will give a COPLAS/HIPERFIT talk on compilation techniques in APEX, a compiler from APL to SaC. An abstract for the talk is given below.

COPLAS –o– HIPERFIT –o– TALK –o– COPLAS –o– HIPERFIT

The Three Beaars – Basically, Every Array Allocation Reduces Speed

Robert Bernecky, Snake Island Research Inc

Monday October 22, 10:30am-11:30am, DIKU Meeting Room A/B

Abstract

Functional array language compiler and interpreter designers try to reduce the number of arrays created during application execution, because the negative impact of arrays on performance is so dramatic. Just as The Three Bears had different requirements for their own satisfaction, so do differing array shapes have different requirements for their elimination. The problem itself is a bear: scalar operations are the baby bear, typified here by dynamic programming and the Floyd-Warshall algorithm; operations on small arrays, such as numerically intense computations on complex arrays, is the mama bear; operations on large arrays, typified by acoustic signal processing, is the papa bear. We compare interpreted to compiled APL performance for several applications with different array shapes, and give an overview of the various optimizations that enable those speedups, in both serial and parallel contexts.

Biography

Robert Bernecky has designed and developed APL systems since 1971. While at I.P. Sharp Associates Limited, he was one of the people responsible for the design and development of SHARP APL, a system that set the standard for performance of large-scale APL systems. He has authored papers on language design, algorithm design, and interpreter performance.

Bernecky is the CEO of Snake Island Research Inc, a consulting and research firm headquartered in Toronto.

Bernecky developed APEX – the APL Parallel Executor – a high-performance, retargetable APL compiler for serial and parallel computers.

Bernecky holds a BA in philosophy from SUNY at Buffalo, and an MSc in Computer Science from the University of Toronto.

September Seminar

2012-08-24T00:00:00+00:00

The HIPERFIT September 2012 seminar is planned for Friday, September 7 from 13:00 to 15:00 at the DIKU small auditorium (Lille Aud). The seminar is hosted together with DIKU and consists of two talks:

13:00-14:00. DBToaster: Higher-Order Delta Processing for Dynamic, Frequently Fresh Views. Yanif Ahmad, Johns Hopkins University, USA.
14:00-15:00. A Cluster of Languages for Mathematical Computing. Stephen M. Watt, Western University, Canada.

The talks give insights into how to query big data using small languages and the development of programming abstractions for mathematical programming in languages such as Maple and MathML. For more information about the talks and the authors, please consult the event page.

We look forward to see you at the seminar!

Best Regards,

Martin Elsman HIPERFIT Center Manager

ICFP 2012 in Copenhagen

2012-08-07T00:00:00+00:00

ICFP Is the leading international conference on functional programming (FP). Finance has internationally emerged as a strategic and rapidly growing application domain for FP. After Edinburgh, Baltimore and Tokyo in 2009-2011, ICFP takes place in Copenhagen this year, organized locally by the HIPERFIT team at DIKU. We’d like to draw your attention to specifically CUFP and FHPC (see below), which target industrial uses and users of FP. These events represent a unique opportunity of getting access to FP-based technology and innovation trends. Could you please forward the enclosed information to those of your colleagues who might be interested in taking advantage of this possibility. (Please feel free to direct questions to Fritz Henglein.

ICFP 2012 @ Copenhagen

Microsoft, Google, Facebook, Twitter, Credit Suisse, Standard Chartered, Jane Street Capital, SimCorp, Lexifi and many others are sponsoring ICFP 2012 in Copenhagen – why would leading IT and Finance companies do such a thing?

What is ICFP?

The International Conference on Functional Programming is the premier conference series for cutting-edge scientific and industrial work in functional programming. ICFP 2012 will take place in the heart of Copenhagen, September 9-15. It is arranged by the University of Copenhagen under the auspices of ACM, the world’s oldest and largest organization of computing researchers and professionals. With its 14 colocated workshops and tutorials, ranging from foundational computer science via high-profile industrial applications to hands-on tutorials by the international Who’s-Who of functional programming, ICFP 2012 will be the gathering ground of FP researchers and professionals from around the globe.

What is functional programming?

Functional programming is based on the notion of transformation rather than machine state update. It is solidly based on logic, has pioneered technologies that have subsequently seen widespread adoption in other languages (garbage collection, type inference, language-integrated querying and others), incorporates the most advanced compilation technology in a variety of industrial-strength programming languages (F#, OCaml, Haskell, Scala, etc.). It has found strategic applications in finance (such as quantitative finance and high-frequency trading), telecommunication (such as telephone switches), web systems (massively scalable web systems) and other domains due to its built-in support for domain-specific modeling, safety, scalability, and hardware independence.

Why participate in ICFP 2012?

You get to hear about research, development and practice trends from “the horse’s mouth” and mingle with the world-leading researchers and commercial FP practitioners. At the Commercial Users of Functional Programming (CUFP) tutorials you can learn hot FP languages and technologies such as F#, Scala, Haskell from the leading experts. You can hear about Finance IT applications both at the CUFP workshop and the Functional High-Performance Computing (FHPC) workshop. The latter has finance as its focus area and serves as the 4th HIPERFIT workshop.

What to do?

It is easy to register for participation at ICFP: Go to the Registration page at icfpconference.org/icfp2012 and follow instructions.

Important: Early registration deadline with discounted registration fees is August 9th, 2012.

For more information on ICFP, see http://icfpconference.org; for more on ICFP 2012, see http://icfpconference.org/icfp2012, for more on CUFP, see cufp.org, for more on ACM, see http://acm.org, for more on University of Copenhagen, see http://diku.dk (Department of Computer Science), and http://ku.dk.

On behalf of the DIKU HIPERFIT team,

Fritz Henglein

PhD positions in parallel FP

2012-07-06T00:00:00+00:00

Do you know promising parallel functional programmers? HIPERFIT currently seeks to employ two PhD scholars in the area of parallel functional programming.

The objective of the PhD projects is to develop, implement, and evaluate techniques related to efficiently executing high-level, functional specifications of both numeric and symbolic computations on modern, highly parallel computational platforms – such as large-scale grids and clusters, multi- and many-core CPUs, GPGPUs, and FPGAs. According to the overall goal of HIPERFIT, the results will be applied in the domain of financial computations, but are expected to be of more general interest.

Candidates are expected to have a solid academic background in relevant areas, including several of: programming language theory and implementation, parallel algorithms, high-performance computing, and large-scale functional-programming. Experience in financial math or stochastic modeling are helpful, but not essential or required. A full description of the positions and working conditions, along with requirements and a list of compulsory application material, can be found on the respective www.diku.dk page.

Applications have to be submitted via the online system of the university of Copenhagen (please read about the required documents before).

The deadline is July 30, 2012

Apart from that, we would like to wish you a good summer vacation. The next seminar will be held in late August and announced in a separate newsletter after the main holiday period.

Best regards from HIPERFIT

One Day Workshop on SaC and APL

2012-06-19T00:00:00+00:00

On June 28, 2012, DIKU and HIPERFIT host a one day workshop on Single assignment C (SaC) and APL programming. Participants will have a chance to try out both SaC and APL in practice through separate tutorial/exercise sessions. An overall theme of the workshop is to investigate the possibilities for using SaC as a target for realizing a subset of APL, so as to utilize SaC’s possibilities for parallel computations on modern hardware.

For registration and a detailed program, please consult the SaC/APL workshop page.

Best regards,

Martin Elsman, Associate Professor, PhD, HIPERFIT Center Manager

June Seminar

2012-05-22T00:00:00+00:00

Spring in Denmark is full of public holidays, and one of them (Grundlovsdag) is on the first Tuesday in June. Therefore, our next seminar will not be on Tuesday, but on Friday, June 8, and it starts already at 14:00.

The seminar features two talks from 14:00 to 15:30. After the talks, refreshments will be served.

The location is the usual one: Mødelokale A/B at DIKU, Universitetsparken 1.

First, we are happy to have Kostas Tzoumas from the Technical University of Berlin visiting us on Friday. Kostas Tzoumas is a senior researcher at the Technical University of Berlin. His research interests are data analytics, especially query processing and optimization in massively parallel environments. Kostas received his PhD from Aalborg University in 2011, and has also worked at the University of Maryland, and at Microsoft Research. He graduated in 2007 from the National Technical University of Athens.

In his seminar talk, Kostas will give us an introduction to “Big Data Analytics with Stratosphere”, a system that generalises Google’s MapReduce programming model and features a robust, cloud-enabled optimising execution engine.

In the second talk, Simon Lund will give us “A tutorial introduction to MiG”, a Grid middleware which can give HIPERFIT researchers access to high-performance hardware and shared storage, and includes virtual groups and collaboration software.

Simon Lund holds a degree in computer science from DIKU, and is currently a 4-year PhD student in HIPERFIT, supervised by Professor Brian Vinter. Simon worked with, and on, Minimum-intrusion-Grid on several occasions during his bachelor studies.

We look forward to seeing you at the seminar.

Best regards from HIPERFIT

Seminar in May and Martin Elsman

2012-04-30T00:00:00+00:00

Welcome to a new edition of our HIPERFIT newsletter, announcing the next seminar talks and an important new member.

Due to the public holiday tomorrow, we are again moving our seminar by one week.

The next seminar will be on Tuesday, May 8, 3pm, in Mødelokale A/B at DIKU, Universitetsparken 1.

We could win two external speakers who will talk to us about exciting research topics related to HIPERFIT.

Under the title “Statically Typed Pensions”, David Christiansen from ITU will introduce us to a domain-specific language (DSL) developed in the context of project ACTULUS. Actulus is a research project in actuarial mathematics, involving Mogens Steffensen from IMF who is also leading the mathematical finance research in HIPERFIT. In ongoing work, David is developing a DSL for declaratively specifying life insurance and pension plans, as well as probabilistic models of the events upon which their valuation depends. He will present the language as it currently exists with special emphasis on the type system. The talk was originally planned for the cancelled April seminar.

Michael Christensen from ALOC (Odense) gives the second talk of the seminar:

“Does your organization support evaluation and decision-making?”

We might summarise it as “Idiots vs. Game Theorists”, but rather use Michal’s own words: The talk is based on a study [1] of the organization and procedures surrounding the evaluation of credit applications (in the order of 100 million dollars per year) in a Danish bank. The analysis relies on a screening formalism [2], which at the heart assumes individuals to be fallible, to examine the impact of organizational structure on organizational performance. Recent dynamic models explain the remarkable adaptability observed in the data at the level of the individual. This validation of the theory carries a promise of applicability: How do I design the optimal organization?

There will also be the opportunity for informal chats about the 3rd HIPERFIT workshop, which took place in Chicago on 21-23 April in cooperation with the Stevanovich Center for Financial Mathematics.

At the same time, the May seminar marks the debut of an important new member of staff in HIPERFIT. We are very excited to announce that Martin Elsman is joining the HIPERFIT team. Martin will fill the position of the HIPERFIT center manager, which is an associate professorship at the Department of Computer Science (DIKU).

From April 2008 to May 2012, Martin worked as Chief System Development Consultant and Team Leader in SimCorp, one of the HIPERFIT partners, so he is well-known to core HIPERFIT staff. We are very happy that we could win Martin Elsman for our academic staff at DIKU, as he can also build on substantial previous experience in the university context: He was associate professor at ITU before joining SimCorp, and obtained important research results in program analysis, type systems, module systems, as well as compilation and optimisations, in the context of Standard ML (MLKit).

A warm welcome to Martin Elsman!

We are looking forward to seeing you at our May seminar.

Best regards from HIPERFIT

[1] Michael Christensen and Thorbjørn Knudsen. The Architecture of Knowledge Organization. In Knowledge Governance: Processes and Perspectives (red. Nicolai J. Foss, Snejina Michailova; Oxford University Press):47-80, 2009.

[2] Michael Christensen and Thorbjørn Knudsen. Design of decision-making organizations. In Management Science, 56(1):71-89, January 2010.

Seminar today canceled

2012-04-24T00:00:00+00:00

Due to unexpected circumstances, we have to cancel today’s seminar.

The planned talks will be postponed to a later time.

Our next seminar is scheduled for the 8th of May, we will get back to you with the programme again.

Best regards from HIPERFIT

April Seminar and Workshop in Chicago

2012-03-30T00:00:00+00:00

The Easter period is approaching with rapid strides now, it is time for a quick update on HIPERFIT and upcoming activities.

Our next seminar will be held on Tuesday, April 24, 3pm, in the usual room Mødelokale A/B at DIKU, Universitetsparken

We are looking forward to the following two talks:

Martin Rehr will give us an introduction to MiG, a Grid middleware which can give HIPERFIT researchers access to high-performance hardware and shared storage, and includes virtual groups and collaboration software. Martin works on GPU parallelisations in Brian Vinter’s High-Performance Systems (HPS) group in HIPERFIT, and holds a PhD in Computer Science from 2010 awarded for his extensions to MiG.
In the other talk, David Christiansen will introduce us to a domain-specific language for computations in the life and pension business, developed in the context of project Actulus. Actulus is a research project in actuarial mathematics, involving Mogens Steffensen from IMF who is also leading the mathematical finance research in HIPERFIT.

Right before the seminar, from April 20 to April 22, several HIPERFIT researchers will attend the 3rd HIPERFIT workshop, “Functional Programming in Quantitative Finance”, in Chicago. This 3rd HIPERFIT workshop is organised in cooperation with the Stevanowich Center for Financial Mathematics.

We are looking forward to presenting our vision of domain-specific support for financial computation to international industry representatives.

The workshop information page with latest information can be found here.

At the seminar on April 24, there will be opportunities to hear about the event in retrospect in an informal setting.

Best regards from HIPERFIT

Seminar in March and additional funding

2012-03-01T00:00:00+00:00

Please find here a quick update and invitation to our HIPERFIT seminar in March, next Tuesday.

The seminar will be held on Tuesday, March 6, 3pm in Mødelokale A/B at DIKU.

This month, the seminar is dedicated to our additional SME funding.

The focus of the cooperation with HIPERFIT is parallel quantitative analysis of financial instruments, especially pricing and risk analysis. In particular, we want to accelerate such computations through using highly-parallel graphics processing units (GPGPUs).

Knowledge and software within financial computation will be contributed by SMEs – cooperation focuses on parallelisation, and to a lesser extent on the mathematical methods used.

In the seminar, Professor Rolf Poulsen will introduce the topic in more detail, and present and discuss suggested SME-collabration projects we received in response to the call:

FinE Analytics: Pricing Danish mortgage bonds in scenario trees in a parallel cloud environment using functional programming in Dyalog APL.
Bangert Research: Post-crisis yield curve estimation and risk management using parallel architectures

Interested parties and HIPERFIT partners are especially welcome.

Best regards from HIPERFIT

HIPERFIT seminar in February postponed

2012-02-03T00:00:00+00:00

It is time to inform about our HIPERFIT seminar in February. However, you still have more than one week, as we are again postponing the seminar by one week.

The seminar will be held on Tuesday, February 10, 3pm (moved by one week) in Mødelokale A/B at DIKU. It features two talks by faculty members:

Marcos Vaz Salles: Making Time-stepped Applications Tick in the Cloud
Cosmin Oancea, Jost Berthold: On Real-World Pricing Code and its Parallelisation

The first talk reports Marcos’ joint work with Cornell fellows (presented at SoCC 2011). The second talk is work in progress on essentials of, and insights into, real-world pricing code using a Monte-Carlo method. Abstracts are included below. We hope you will enjoy the talks!

Best regards from HIPERFIT

Abstracts

Marcos Vaz Salles: Making Time-stepped Applications Tick in the Cloud

Scientists are currently evaluating the cloud as a new platform. Many important scientific applications, however, perform poorly in the cloud. These applications proceed in highly parallel discrete time-steps or “ticks” using logical synchronization barriers at tick boundaries. We observe that network jitter in the cloud can severely increase the time required for communication in these applications, significantly increasing overall running time. In this paper, we propose a general parallel framework to process time-stepped applications in the cloud. Our framework exposes a high-level, data-centric programming model which represents application state as tables and dependencies between states as queries over these tables. We design a jitter-tolerant runtime that uses these data dependencies to absorb latency spikes by (1) carefully scheduling computation and (2) replicating data and computation. Our data-driven approach is transparent to the scientist and requires little additional code. Our experiments show that our methods improve performance up to a factor of three for several typical time-stepped applications.

Cosmin Oancea, Jost Berthold: On Real-World Pricing Code and its Parallelisation

HIPERFIT’s main mandate is to investigate language and compiler technology to allow financial software to be (i) elegantly and readily written and (ii) effectively optimized to exploit hardware parallelism. In this context, we report our insights from work in progress: evaluating a reasonably-complex kernel of a partner’s financial library, comprising about 6200 lines of low level C.

First, we experiment with parallelization solutions in the C-language domain and show that speed-ups as high as 22x and 34x can be achieved on a commodity GPU (Quadro2000M), for double and single precision floats, respectively. Second, we distinguish two main sources of optimizations: (i) classical transformations that adapt the memory usage to the hardware parameters, and (ii) the higher-level ones that exploit algorithmic invariants, which need to be synthesized at interface level. Finally, to support this higher-level perspective, we are in the process of rewriting the library in a higher-level language (Haskell) with good support for mathematical abstraction, higher-order functions, modularity and code reuse. Not surprisingly, the functional approach makes the data-parallel nature of the computation much more evident than the original C version.

We expect that the C-Haskell performance gap can be substantially reduced by using modern libraries for data-parallelism in our – so far sequential – code. Furthermore, our results indicate that the optimizations that were identified in the C code carry over to the Haskell domain and can even be partially automated (by means of rewrite rules). Our work in progress on extensions and optimizations is an important step towards a more generic codebase and domain-specific abstractions of pricing code.

Call for collaboration with SMEs

2012-01-12T00:00:00+00:00

Call for collaboration with small and medium enterprises (SMEs)

With great pleasure, we can announce today that small and medium enterprises (SMEs) now have the opportunity to collaborate with HIPERFIT in a special additional programme.

HIPERFIT’s funding agency, the Danish Council for Strategic Research (DSF), provides additional funds to existing big research projects, with the goal to expand ongoing research to small and medium enterprises. A particular goal is to involve “new” companies which have not had the opportunity for funded collaborations. HIPERFIT members have successfully applied for these additional funds with DSF, and we are now looking for companies interested in cooperating with us.

The topic and focus of this cooperation is parallel quantitative analysis of financial instruments, especially pricing and risk analysis. A particular focus is to accelerate such computations through using highly-parallel graphics processing units (GPGPUs). Participating SMEs are expected to contribute knowledge and software within quantitative analysis of financial instruments as a prerequisite - cooperation with HIPERFIT primarily focuses on parallelization, and to a lesser extent on the mathematical methods used.

A more detailed description (in Danish) of the project goals and organisation is available here. Please also consult the official call for collaboration at the Strategic Research Council for latest information, the description of the entire programme, and criteria for defining SMEs.

Interested SMEs are asked to contact Fritz Henglein or Rolf Poulsen, the coordinators of this initiative within HIPERFIT, or info@hiperfit.dk. The deadline for expressions of interest is March 1, 2012.

Happy New Year and come to our seminar

2012-01-05T00:00:00+00:00

First of all, we would like to wish you all a happy new year 2012!

Besides, this is just a quick and gentle reminder for our seminar next week:

On Tuesday, January 10, 3pm (moved by one week), the HIPERFIT seminar in Mødelokale A/B at DIKU will feature two talks by faculty members of HIPERFIT. Meanwhile, abstracts are available (also in the calendar), and here they come:

Christian Andreetta, new post-doctoral researcher in the DSL strand of HIPERFIT: On Bayesian Networks.

We will give a short overview of Monte Carlo strategies, probability calculus and directed graphical models like Dynamic Bayesian Networks. Examples of applications will be the use of mixture models to represent financial market data, and Markov Fields to propose a quick and general solution to the drunken sailor problem.
Mogens Steffensen, research area manager for the mathematical finance strand of HIPERFIT: Dynamic Consumption-Investment Decisions.

We review the classical consumption-portfolio choice problem by Merton and discuss the complications in connection with generalizations like uncertain lifetime, labor income, and a stochastic opportunity set. We discuss the assumption of time-additive utility and generalizations to recursive utility. This is related to the notion of consistency which we review in that context and in the context of mean variance optimization.

We hope to see you in the seminar next Tuesday!

Best regards from HIPERFIT

January Seminar and new Post Docs

2011-12-08T00:00:00+00:00

We can look back at a very successful 2nd HIPERFIT workshop last week. A large number of participants (100 subscriptions, and some walk-in participants) showed us that our selection of invited talks could indeed motivate a heterogenous audience and stimulate interdisciplinary dialog.

To all attendants and contributors, we say thank you for coming; we enjoyed the event very much and appreciated the inspiring conversations.

The slides of the talks are now available at the workshop web page, so that they can be read post-hoc.

For workshop number three, we have already arranged a cooperation with the Stevanovich Center for Financial Mathematics in Chicago. The 3rd HIPERFIT workshop will be a joint event, on “Functional Programming in Quantitative Finance” (April 20-22, 2012). At the moment, the organisers are looking for contributions (in particular, but not limited to, implementation of quantitative algorithms, trading applications or data analysis). Contact us if you are interested!

But now the Christmas break is coming with rapid strides. In view of the holiday period, we have moved our January seminar by one week. On January 10, the seminar features two talks by faculty members.

First, Christian Andreetta will give a talk “On Bayesian Networks”. He will introduce graphial models in general (of which Bayesian networks are a subset), and discuss the drunken sailor problem. As this might lead to a statistical physics angle on portfolio allocation, we have a nice transition to the following talk.

Mogens Steffensen will give us insight about “Dynamic Portfolio Optimisation” – the problem of describing how changing parameters over time affect decisions in portfolio management, and how such effects can be modelled. Furthermore, we are happy to officially welcome two new post-doctoral members in HIPERFIT:

Cosmin Oancea joined us on November 15 as a Post-Doc. He has vast experience in both runtime techniques and static analysis to extract and optimise parallelism. Cosmin’s particular field in HIPERFIT will be parallelisation and efficient execution.
Christian Andreetta joined us on December 1st as a Post-Doc. He has a background in computational biology and statistical methods, and he is already a familiar face in our seminars. Christian will concentrate on DSLs for statistical and probabilistic computations.

Welcome Cosmin and Christian!

Best regards from HIPERFIT

2nd HIPERFIT workshop December 1st and 2nd

2011-11-10T00:00:00+00:00

Preparations are under way at full steam, so it is time to announce our 2nd HIPERFIT workshop, which will take place in Copenhagen on December 1-2, 2011.

While our first workshop was addressing partners and faculty, this time we target a broader audience. We have invited members of our advisory board and other guests, gathering 10 invited presentations and two special talks (DIKU talk and COPLAS talk series). Our speakers are international researchers from the different HIPERFIT areas, ranging from from modern mathematical finance, via programming language technology and modern approaches to parallel programming, to high-performance systems.

Up-to-date information, the programme in detail, and a collection of abstracts can be found at the workshop web page, please have a look! You can also download the call for participation as pdf or as txt.

There will be no published proceedings, the meeting is intended to be an interdisciplinary forum for discussion and networking. The workshop is open to public, interested researchers and practicioners of all related fields are invited to participate, meet and discuss with our international speakers and guests. Participation is free of charge, but for organisational purposes, a workshop registration is required. Please fill out this form to register.

The 2nd HIPERFIT 2011 workshop will be held at the University of Copenhagen, North Campus, in the HCØ building next to the Department of Computer Science (DIKU) - a map can be found at the department pages. Please follow the signs saying “2nd HIPERFIT Workshop” when approaching the DIKU building.

We are looking forward to seeing you at the workshop.

Best regards from HIPERFIT

Kevin Hammond Visiting

2011-11-02T00:00:00+00:00

Kevin Hammond from the University of St.Andrews visited us from October 17 to 19, 2011.

Kevin is a Haskeller of the first hour, proficient in parallel Haskell for more than 20 years. We were discussing research directions in parallel functional programming with Kevin, and he also gave a talk at the department on Monday Oct 17, 3:00 pm, in the small auditorium.

For more information about the talk, see the DIKU announcement.

Kevin Hammond is a full Professor in Computer Science, in the School of Computer Science, at the University of St. Andrews in Scotland, where he leads the Functional Programming research group. He is also an Honorary Professor at Heriot-Watt University, Edinburgh, Scotland. His main interests are in cost modelling, parallelism and real-time and embedded systems.

November Seminar and 2nd workshop

2011-10-27T00:00:00+00:00

We hope you enjoyed our last events, the October seminar and the invited talk by Kevin Hammond about parallel Haskell and “Paraforming”. With this newsletter, we announce our next events.

After the internal talks in October, our November seminar features two external talks.

First, Mogens Høgh Jensen from NBI (KU) will tell us about “Complex Systems and Econophysics – Asymmetries in the Stock Market”. Econophysics is a relatively new area where complex systems modelling is applied to explain economic phenomena.

Second, Kourosh Marjani Rasmussen and Ekkart Kindler from IMM/DTU present recent results about “FXSL – a domain specific language for generating and testing FX strategies”. This domain-specific language enables foreign exchange (FX) traders and analysts to quickly develop trading strategies, and to test them on historic data using a suitable tool.

Please find the complete abstracts in our event calendar.

As usual, the seminar will be held on the first Tuesday of the month: November 1st, 15:00 in meeting room A+B (Foyer) at DIKU, Universitetsparken 1.

Further, we would like to announce the 2nd HIPERFIT workshop on December 1st and 2nd. A more elaborate call for participation will follow and the workshop programme is not entirely fixed yet, but it is a good idea to reserve some time for this exciting event. We are looking forward to a bouquet of international researchers who present their recent work in connection with HIPERFIT. Amongst others, we are looking forward to hearing about: Collateral flows and counterparty-risk-neutral swap rates, declarative array programming, high-throughput portfolio processing, a hybrid high-performance database system architecture, and other talks.

The workshop will be held on December 1st and 2nd in auditoriums of the HCØ institute (Universitetsparken 5). More details and the complete programme will follow.

We hope to see you at both of these events!

Best regards from HIPERFIT

Seminar, Guest, and Reading Material

2011-09-27T00:00:00+00:00

It is time for another newsletter with forthcoming events.

Our monthly seminar will be held on Tuesday October 4, 3pm (in Mødelokale A/B), featuring two talks by academic HIPERFIT members.

Professor Brian Vinter will introduce us to Vectorisation, with examples in NumPy (numeric Python). Brian’s group is working on a vectorising multi-platform backend for different languages. The second talk, by Ken Friis Larsen, is entitled “What shall we do with a drunken sailor?” Ken will present a case study of a “drunken sailor walk”.

Furthermore, we would like to announce another exciting HIPERFIT talk by a guest of ours. From October 17 to 19, we are happy to host Kevin Hammond, Professor of Computer Science at the University of St.Andrews, where he leads the Functional Programming research group. Kevin is a Haskeller of the first hour (he wrote the first Haskell compiler), and profiscient in parallel Haskell for more than 20 years. His research interests are in cost modelling, parallelism and real-time and embedded systems.

On October 17, 3:00pm, Kevin will give a talk on “ParaForming”, applying novel refactoring techniques in order to parallelise Haskell programs (also including a discussion of parallel patterns and skeletons). The talk will be held in the DIKU small auditorium (Lille UP1). Everybody is welcome, no registration is required.

So, we hope to see you all at these talks, but we also have reading material for you: The master students Michael Werk and Joakim Ahnfelt-Rønne have finished and defended their thesis about “Pricing composable contracts on the GP-GPU” last week. They did a great job and obtained the highest grade.

Congratulations Michael and Joakim!

The thesis is available online from the publication page on our HIPERFIT website.

Last but not least, a paper describing our HIPERFIT research for an audience of functional programming researchers has been accepted at the symposium “Trends in Functional Programming”. This symposium is post-refereed and was held already in May. We are now putting together the final version of the paper and will publish it on our publication page very soon (so Michael and Joakim do not feel lonely).

Best regards from HIPERFIT

Jeroen Weijers visiting September 5 7

2011-09-27T00:00:00+00:00

Between September 5 and 7, 2011, we were hosting Jeroen Weijers from Tübingen, Germany. Jeroen, who is actually from the Netherlands, works in the database group of Torsten Grust. His research includes applying functional programming techniques and code transformations to database queries.

Jeroen gave a short talk about ongoing research in the HIPERFIT seminar.

Jeroen is a research assistant and member of the Database Systems Group at the University of Tübingen. His research focusses on developing an efficient embedded query language for Haskell using the Flattening Transformation. In general he is interested in functional programming, programming language development and program correctness.

The Database Systems Group at the University of Tübingen (led by Torsten Grust) performs research into the compilation, optimisation, and evaluation of a variety of database languages, walking the fine line between database query and programming language technology.

Dave Cliff

2011-08-26T00:00:00+00:00

Dave Cliff, from the University of Bristol, visited DIKU on May 5, 2011. He gave an exciting talk titled “The Flash Crash of May 6th, 2010: WTF?”. “Flash Crash” refers to a sudden drop and rise of the US stock market on May 6, 2010. Because of algorithmic high-frequency trading, the Dow Jones index lost around 9 percent - and recovered within minutes! Dave Cliff gave us his view on “large-scale socio-technical systems”, argueing that society needs “principled evaluation of systemic risk”.

Dave Cliff is a Professor of Computer Science at the University of Bristol. He has more than 20 years of experience as a researcher in computer science and complex adaptive systems. He has previously worked in academic faculty posts at the University of Sussex, at the MIT Artificial Intelligence Lab, and at the University of Southampton. He also spent seven years working in industry: initially as a senior research scientist at Hewlett-Packard Research Labs where he founded and led HP’s Complex Adaptive Systems Research Group; then as a Director in Deutsche Bank’s London Foreign-Exchange Complex Risk Group. His research for HP included early work, in the mid-to-late 1990s, on novel decentralized management systems for utility-scale (“cloud”) computing systems; as part of that work he invented the Zero-Intelligence Plus (ZIP) adaptive automated trading strategy. In 2001 a team of researchers at IBM showed that ZIP algorithmic traders consistently outperform human traders. In October 2005, Cliff was appointed Director of the £10m EPSRC-funded five-year UK national research and training initiative in the science and engineering of Large-Scale Complex IT Systems (the LSCITS Initiative). He is author or co-author on over 70 academic publications, and inventor or co-inventor on 15 patents; he has undertaken advisory and consultancy work for a number of major companies and for various departments of the UK Government; and he has given around 200 keynote lectures and invited seminars.

Center Manager Position

2011-08-23T00:00:00+00:00

In cooperation with our host institution DIKU, the HIPERFIT research center announces the following position:

1 Assistant or Associate Professor, as the HIPERFIT Center Manager

This is a permanent position with the Department of Computer Science (DIKU) at the University of Copenhagen, including research, supervision and training of research fellows, teaching, and other tasks requested by the Department. In the first 5 years of the appointment, the candidate will - apart from performing research, teaching and developing his/her own field - collaborate with the center director and the research area managers to attend to the center obligations: manage formulation, follow-up and achievement of project plans, organise center activities, communicate with interested parties, interact with partners, and potentially acquire additional external funding of research activities.

The successful applicant will have a PhD in computer science, mathematics, statistics, physics or similar, and documented scientific expertise in at least one of the following HIPERFIT research areas: domain-specific languages (DSLs), functional programming (FP) and high-performance systems (HPS), and a scientific and professional interest in the other research areas. In addition, applicants should have research project management skills, preferably documented by previous experience with project and/or center administration/management. Furthermore, the candidate must have good English language skills and should be fluent, or within a half year become fluent, in Danish.

The application must be submitted by e-mail to science-position@science.ku.dk (Subject: Your last name, 211-0324) and be received no later than September 6, 2011 at 12.00 noon local Danish time. Please see the full announcement at www.offentlige-stillinger.dk for the required application material and more details. In case of questions, please contact the center director, Professor Fritz Henglein, at henglein@diku.dk.

Next two seminar talks

2011-08-15T00:00:00+00:00

Welcome to a short post-holiday newsletter, to remind you of our next seminar sessions which were already announced in a previous newsletter, including small changes in the schedule.

The next seminar session is already tomorrow, August 16. We are happy to announce two contributed practicioner’s talks on domain-specific languages for financial contracts.

Dirk Bangert (Bangert Research): “DSLs for Financial Contracts in Valuation and Risk Management”.
Carl Balslev Clausen (SimCorp): “XpressInstruments: a modelling framework for contract management and valuation”.

The following seminar session has been moved away from August 29 (previously announced), to September 6, so we can include a talk by our guest Jeroen Weijers from Tübingen, Germany. Jeroen (who is actually from the Netherlands) works in the database group of Torsten Grust, his research includes applying functional programming techniques and code transformations to database queries. An abstract and more details can soon be found in the HIPERFIT calendar.

As usual, time and location for seminars is 3:00pm in Mødelokale A+B, DIKU. We hope to see you for both seminar sessions!

Best regards from HIPERFIT

Simcorp position 2011

2011-07-20T00:00:00+00:00

Opportunity for a part-time job and Master’s thesis project

Our HIPERFIT partner SimCorp is looking for a Computer Science student specialised in DSLs and functional programming. Specifically, the job is in the Financial Instruments unit, and the area of work is automated testing using an internal DSL implemented in F#. Apart from developing actual tests with the QA Analysts, the job includes further development of the test DSL, and may result in an interesting Master’s thesis project.

For more information, please see the full job description at the SimCorp web site.

Applications will be processed when they arrive, no deadline is given. Don’t miss this exciting opportunity if you are interested in functional programming languages and testing!

Manuel Chakravarty

2011-07-15T00:00:00+00:00

We were happy to host Manuel Chakravarty, Professor of Computer Science in the University of New South Wales (Sydney), between July 11 and 13, 2011. Manuel has been working on Data-Parallel Haskell for more than 10 years and is a driving force behind several strands of development in parallel array processing in Haskell.

During his stay, Manuel gave a DIKU talk on data parallelism using Haskell, motivating both the use of functional programming as well as the appeal of data parallelism, and illustrating how data parallel Haskell programs look like in different variants. After this overview talk, we proceeded with a more technical talk about the GPU-enabled Haskell array library Data.Array.Accelerate, as a special issue of the HIPERFIT seminar.

Manuel M T Chakravarty is an Associate Professor at the University of New South Wales, Sydney. His main research interests are in functional programming languages, novel compiler technology, and parallel programming. He graduated from the University of Karlsruhe and received a doctoral degree from the Technical University of Berlin.

He contributed to Haskell’s foreign function interface, the theory and implementation of type families, and the design and realisation of Data Parallel Haskell, an implementation of nested data parallelism in the Glasgow Haskell Compiler. He receives his inspiration from combining theory with practice.

Nordea Computer Science student 2011

2011-06-29T00:00:00+00:00

One of our partners in HIPERFIT, Nordea, announces an open position in relation to HIPERFIT.

But you have to be quick, the deadline is June 26

Use your knowledge of domain-specific languages and high-performance computing in a demanding trading environment

Computer Science Student for Model Development, Copenhagen

Why wait for a job opening till the end of summer? Here is a unique opportunity to start your career already in June!

In Model Development you will be responsible for maintaining and optimizing our domain specific language Loke that is describing all our financial contracts. These contracts are passed to our model library for market values and risk number calculation on our in house computing grid, so part of the job will also be to help optimize the usage of our computing grid.

Extremely technical and communicative at the same time

Your background would be a computer science student having obtained your B.Sc. and are planning to continue your studies with a focus on domain specific languages or high performance computing.

The areas of work on our trade description language and with our high performance computing grid fits with the research areas in the HIPERFIT project, so this position could lead to suitable subjects for a Master’s thesis project together, and possible also a Ph.D. project. It is thus also preferred that your main interest is within the HIPERFIT research areas.

The position is for approximately 15 hours per week, with flexibility around holidays and exams, and the preferred candidate can start right away, possible with some weeks of full time work during the summer.

You can contact Jens Lund on +45 33 33 55 54 for further information.

Read more about the position and apply on http://nordea.com/job (use this direct link to Job 91580, or job-id 91580).

DIKU talk and Summer Seminar

2011-06-29T00:00:00+00:00

The main holiday period is approaching, maybe some of you are enjoying beach and sun in the south.

However, while those will find this message late, others have the chance to meet someone who lives far more in the south than that - in Australia. We are happy to host Manuel Chakravarty, Professor of Computer Science in the University of New South Wales (Sydney). Manuel has been working on Data-Parallel Haskell for more than 10 years and is a driving force behind several strands of development in parallel array processing in Haskell.

On July 12, 2:00pm, Manuel will give a DIKU talk on data parallelism using Haskell, motivating both the use of functional programming as well as the appeal of data parallelism, and illustrating how data parallel Haskell programs look like in different variants.

After this overview talk, we will proceed with a more technical talk about the GPU-enabled Haskell array library Data.Array.Accelerate. This second talk is an extra issue of the HIPERFIT seminar.

About the seminar in general: We are running a “HIPERFIT summer seminar” during the summer period, with two HIPERFIT seminars in July and August.

The first session is next Tuesday, July 4; with speakers Rolf Poulsen (about American Options) and Brian Vinter (about Vectorisation).
The second session is the aforementioned talk by Manuel Chakravarty, on Tuesday, July 12.
We restart (after our own holiday period) on Tuesday, August 16, with a “DSL day”: Dirk Bangert talks about domain-specific languages for contract valuation. We hope to win additional speakers, so we can discuss more DSL variants (to be confirmed).
On August 30, we will have talks about programming languages for parallelism, by Fritz Henglein (about APL and SETL) and Andrzej Filinski (about NESL).

Time and location for all seminars is 3:00pm in Mødelokale A+B, DIKU. As usual, you can find these events in the HIPERFIT calendar.

And finally: The application deadline for some open positions within HIPERFIT is approaching. We wish to employ a total of six PhD scholars and postdoctoral researchers, and a center manager; the latter is a faculty position as an assistant or associate professor. Interested candidates are invited to apply until Monday, July 4.

Best regards from HIPERFIT, Jost Berthold

Open Positions and Events in June

2011-06-01T00:00:00+00:00

We can look back at a very successful workshop last Wednesday. We were positively surprised by the number of subscriptions (62) and lucky to arrange for a larger room than originally planned. To all attendants, we say thank you for coming; it was a great day with inspiring conversations.

Aside from the workshop, we also ran our first steering committee meeting and discussed the start of our center with our funding source DSF. We were quite satisfied with the results of both meetings.

A number of open positions within HIPERFIT are now announced on the university pages and on the HIPERFIT web page. We wish to employ a total of six PhD scholars and postdoctoral researchers, and a center manager. The latter position will be a permanent appointment with the department of computer science DIKU as an assistant or associate professor. If you know of interested candidates, check out the description on the HIPERFIT website or the DIKU web site.

Next Tuesday, it is again time for our monthly seminar. Interested students, researchers and partner members are welcome. This time, Sinan Gabel from Nordea will talk about “Selected concrete challenges and ideas for further research”. In the second half, we will plan and schedule a series of overview talks about programming languages for parallelism and concurrency.

Furthermore, HIPERFIT contributes to a workshop on domain-specific languages (DSLs) for modeling on June 16 and 17, which is organised by Chalmers University. World-class researchers meet and discuss how to effectively apply DSLs and modeling techniques for societal tasks in environment protection, climate research, and in the financial domain. You can find details on the DSL4EE workshop pages.

Best regards from HIPERFIT

Center Manager Position and PhD and Postdoc Positions

2011-05-26T00:00:00+00:00

In cooperation with our host institution DIKU, the HIPERFIT research center is hiring.

We wish to fill the following positions until September 2011:

6 PhD scholars or Postdoctoral researchers

These positions are 3 year research positions within our research center. PhD project duration can be 4 years if the candidate holds a bachelor degree instead of a master’s degree. The announcement is deliberately flexible in order to encourage a larger group of applicants, e.g. positions are divided into the four research areas and into postdoctoral and PhD positions following the quality and preference of the received applications.

The PhD projects and postdoctoral research will be carried out at either DIKU, IMF or NBI under the supervision of senior HIPERFIT researchers. Successful candidates must have a solid academic background within at least one of the 4 research areas of HIPERFIT, as demonstrated by graduate level courses and/or project and thesis work in respective relevant fields, and interdisciplinary interest in (at least) one of the other research areas in question.

Please find more details about the positions on the department’s web pages.

Applications must be submitted by e-mail to suntonn@diku.dk (Subject: PhD Scholarship/Postdoctoral Research Position, HIPERFIT research area) and be received no later than July 4, 2011 at 12.00 local Danish time.

In case of questions, please contact the center director, Professor Fritz Henglein, at henglein@diku.dk, or one of our research area managers, as stated in the full announcement.

1 Assistant or Associate Professor, as the HIPERFIT Center Manager

The successful applicant will have a PhD in computer science, mathematics, statistics, physics or similar, and documented scientific expertise in at least one of the following HIPERFIT research areas: domain-specific languages (DSLs), functional programming (FP) and high-performance systems (HPS), and a scientific and professional interest in the other research areas. In addition, applicants should have research project management skills, preferrably documented by previous experience with project and/or center administration/management. Furthermore, the candidate must have good English language skills and should be fluent, or within a half year become fluent, in Danish.

Please find more details for these announcements on the department’s web pages.

The application must be submitted by e-mail to science-position@science.ku.dk (Subject: Your last name, 211-0324) and be received no later than July 4, 2011 at 12.00 noon local Danish time.

In case of questions, please contact the center director, Professor Fritz Henglein, at henglein@diku.dk, or one of our research area managers, as stated in the full announcement.

1st HIPERFIT workshop next week

2011-05-18T00:00:00+00:00

Earlier this month, we announced several events which we organise in HIPERFIT, including our first HIPERFIT workshop. This workshop will take place next Wednesday, 25th of May 2011, at the Department of Computer Science.

We have now completed the agenda and can promise a number of interesting talks from both researchers and industrial partners, in HIPERFIT related topics.

08:30-09:00: Arrival and Registration
09:00-09:10: Welcome and introduction
09:10-09:45: Invited talk: Nikolaj Bjørner (MSR Redmond)
09:45-10:00: Break
10:00-11:05: Session 1: Overview, Modeling (3 contributed talks)
11:05-11:20: Break
11:20-13:00: Session 2: Parallel programming, Performance (4 contributed talks)
13:00-14:00: Wrap up, followed by catered buffet lunch and networking
14:00: End of the workshop

Please find additional details and abstracts on the related HIPERFIT web page.

If you wish to attend the workshop, please take a moment to register (by May 23, 2011 at 12am) on the workshop registration page.

We are looking forward to meeting you at the workshop.

Best regards from HIPERFIT

Events And Positions

2011-04-27T00:00:00+00:00

Right after the Easter holiday, it is time for a quick update on HIPERFIT activities.

In particular, we would like to announce several events.

We are starting our monthly HIPERFIT Seminar on May 3. This semi-public recurring event will be held every first Tuesday of the month, and features talks for information interchange, reporting results, and generating new ideas.
On May 5, HIPERFIT proudly presents Dave Cliff, Professor of Computer Science at the University of Bristol, speaking about the “2010 Flash Crash”. No, this is not the latest security vulnerability of Adobe’’s ubiquitous multimedia plugin :-). “Flash Crash” refers to a sudden drop and rise of the US stock market on May 6 2010. Because of algorithmic high-frequency trading, the Dow Jones index lost around 9% - and recovered within minutes! Dave Cliff will give us his view on “large-scale socio-technical systems” and argue that society needs “principled evaluation of systemic risk”.

Both events are announced in our HIPERFIT event calendar, which you can find on the hiperfit web site.

Furthermore, we will hold the first HIPERFIT workshop on May 25, which is partially open to students and guest researchers. The agenda is being established at the moment; more details will follow as a calendar entry soon.
Last but not least, we can already announce a joint event of HIPERFIT and the EU project “Global Systems Dynamics and Policy” GSDP. Chalmers University of Technology is holding the workshop on “Domain Specific Languages for Economical and Environmental Modelling” DSL4EE near Gothenburg on June 16 and 17. A selection of world-class researchers will speak about domain-specific language technology and modelling, including Fritz Henglein about “DSLs for the financial sector”.

In other news, we are going to announce HIPERFIT open positions in early May. These include a DIKU professorship as the HIPERFIT Center Manager, with an interesting mix of research, administrative management and networking. Furthermore, a total of six PhD scholarships or post-doctoral positions in the four HIPERFIT research areas will be offered. More details can be found soon on the HIPERFIT web page.

Best regards from HIPERFIT

Starting

2011-03-01T00:00:00+00:00

This is the first HIPERFIT newsletter.

We can happily announce the initial version of our website

http://hiperfit.dk

The site gets in shape and is ready to provide you with information about the research center now. We describe our research approach and project organisation, as well as the research themes we are working on. In the future, we are going to use it for publishing scientific results and for announcing related events.

There is a calendar for upcoming HIPERFIT events, which we hope to populate with related conference and workshop events, a HIPERFIT seminar, and generally events where one can meet HIPERFIT fellows.

We will keep you posted.

In general, our project gets up to speed: Partners at DIKU and NBI have discussed a number of potential and concrete “cases”, or “projects”, as we prefer to say, with the industry partners. Work on some of them (mainly through study projects) has started, others are on the way.

In other news, HIPERFIT was described in “Orientering” on the Danish Radio in mid February, including an interview with Fritz Henglein. The accompanying text online mentions “insurance mathematics” as one partner - almost right :-) You can find the link, together with pointers to other articles and online material about HIPERFIT, under “About. In the Press” – Enjoy!

Best regards,

from HIPERFIT