From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Original-To: caml-list@sympa.inria.fr Delivered-To: caml-list@sympa.inria.fr Received: from mail3-relais-sop.national.inria.fr (mail3-relais-sop.national.inria.fr [192.134.164.104]) by sympa.inria.fr (Postfix) with ESMTPS id AC8F3820A1 for ; Fri, 6 Sep 2013 11:36:54 +0200 (CEST) Received-SPF: None (mail3-smtp-sop.national.inria.fr: no sender authenticity information available from domain of jeremy.gibbons@cs.ox.ac.uk) identity=pra; client-ip=163.1.2.162; receiver=mail3-smtp-sop.national.inria.fr; envelope-from="jeremy.gibbons@cs.ox.ac.uk"; x-sender="jeremy.gibbons@cs.ox.ac.uk"; x-conformance=sidf_compatible Received-SPF: None (mail3-smtp-sop.national.inria.fr: no sender authenticity information available from domain of jeremy.gibbons@cs.ox.ac.uk) identity=mailfrom; client-ip=163.1.2.162; receiver=mail3-smtp-sop.national.inria.fr; envelope-from="jeremy.gibbons@cs.ox.ac.uk"; x-sender="jeremy.gibbons@cs.ox.ac.uk"; x-conformance=sidf_compatible Received-SPF: None (mail3-smtp-sop.national.inria.fr: no sender authenticity information available from domain of postmaster@relay14.mail.ox.ac.uk) identity=helo; client-ip=163.1.2.162; receiver=mail3-smtp-sop.national.inria.fr; envelope-from="jeremy.gibbons@cs.ox.ac.uk"; x-sender="postmaster@relay14.mail.ox.ac.uk"; x-conformance=sidf_compatible X-IronPort-Anti-Spam-Filtered: true X-IronPort-Anti-Spam-Result: AoABACmhKVKjAQKinGdsb2JhbABbgzyGXLxoFg4BAQEBAQYNCQkUKIJIgiQJFIdcAw4BDJ1VmCBNh3mNBIEwBYRdgQADlhCOHIZqgWaBaAEHFwY X-IPAS-Result: AoABACmhKVKjAQKinGdsb2JhbABbgzyGXLxoFg4BAQEBAQYNCQkUKIJIgiQJFIdcAw4BDJ1VmCBNh3mNBIEwBYRdgQADlhCOHIZqgWaBaAEHFwY X-IronPort-AV: E=Sophos;i="4.90,853,1371074400"; d="scan'208,217";a="25935312" Received: from relay14.mail.ox.ac.uk ([163.1.2.162]) by mail3-smtp-sop.national.inria.fr with ESMTP; 06 Sep 2013 11:36:53 +0200 Received: from smtp1.mail.ox.ac.uk ([129.67.1.207]) by relay14.mail.ox.ac.uk with esmtp (Exim 4.80) (envelope-from ) id 1VHsTF-0001NE-l9 for caml-list@inria.fr; Fri, 06 Sep 2013 10:36:53 +0100 Received: from dhcp4-nat.cs.ox.ac.uk ([163.1.88.6] helo=[192.168.18.179]) by smtp1.mail.ox.ac.uk with esmtpsa (TLSv1:AES128-SHA:128) (Exim 4.69) (envelope-from ) id 1VHsTF-0003Ee-4n for caml-list@inria.fr; Fri, 06 Sep 2013 10:36:53 +0100 From: Jeremy Gibbons Content-Type: multipart/alternative; boundary=Apple-Mail-297-721054627 Date: Fri, 6 Sep 2013 10:36:52 +0100 Message-Id: <558B91CA-7098-49AC-876C-879227E80017@cs.ox.ac.uk> To: caml-list@inria.fr Mime-Version: 1.0 (Apple Message framework v1085) X-Mailer: Apple Mail (2.1085) X-Oxford-Username: kell0064 Subject: [Caml-list] Oberwolfach Seminar on Mathematics for Scientific Programming --Apple-Mail-297-721054627 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=iso-8859-1 Dear all, I'm lecturing in November at a week-long seminar on rigorous (ie functional= ) programming techniques for computational science. There is funding for be= d, board, and part of travel costs. If you're in contact with computational= scientists who might be interested, could you pass this on, please? Applic= ation deadline is 15th September.=20 Thanks, Jeremy * Call for Participation OBERWOLFACH SEMINAR ON=20 MATHEMATICS FOR SCIENTIFIC COMPUTATION Mathematisches Forschunginstitut Oberwolfach 24th to 30th November 2013 http://www.mfo.de/occasion/1348a MOTIVATION Computational science today depends crucially on simulations, which are typ= ically based on algorithms that have a sound mathematical justification. F= or example, an iterative procedure such as Newton's method is motivated by = appealing to the properties of twice continuously differentiable functions = and their Taylor expansion, which also yield convergence conditions and app= roximation estimates. These algorithms are then implemented on a computer, using a programming la= nguage such as Fortran or C++. Often, the implementation will introduce ne= w computational steps and otherwise modify the structure of the mathematica= l algorithm - for handling or reducing round-off errors, enabling more effi= cient memory access, exploiting parallelization, and so on. As a result, t= he final implementation usually looks very different from the mathematical = algorithm, and the justification given for the latter does not directly ext= end to the former. But if we are to ensure the correctness of simulations,= we need mathematical certainty for both. We aim to bring to the scientific programming community mathematical techni= ques that allow us to achieve the transition from mathematical algorithm to= efficient implementation in a principled manner, with each step motivated = by the application of a mathematical theorem. The intended participants are= students and researchers in computational science (including areas such as= engineering, biology, and economics), and any scientists dissatisfied with= state of the art in transforming mathematics into code. They will be equip= ped subsequently to make a significant contribution to increasing the corre= ctness of the simulations that play such an important role in current scien= tific activity. CONTENT This rigorous approach to programming is most easily presented in the frame= work of functional programming: program calculation can be reduced to strai= ghtforward equational reasoning, provided that all program variables are im= mutable. Accordingly, we will introduce the basic syntax and ideas using H= askell, currently the one of the most successful functional programming lan= guages. The emphasis is not on functional programming as such, and even le= ss so on a specific language such as Haskell; but rather, on the mathematic= s behind program development, which can then be transferred to other contex= ts, such as imperative programming, or parallel programming. This mathematical foundation lies in category theory, which unifies what co= uld otherwise appear as a large collection of "bite-sized" theorems for pro= gram development, too many for any developer to remember and use efficientl= y. Category theory is a broad subject: we will limit ourselves to what is = essential as a framework for datatypes and programs (functors, universal pr= operties, algebras, monads). The many examples, such as fusion (loop elimin= ation), optimal bracketing (important for non-associative operations such a= s those on floating-point numbers), or parallel programming skeletons (such= as Google's MapReduce), will be readily understandable and relevant to sci= entific computing practitioners. One of the most effective ways to counter floating-point errors and to obta= in validated results is to use interval analysis, which however requires mo= re complex data structures and algorithms than is common in other areas of = scientific computation. Extending a function on real or floating-point num= bers to one on intervals is a matter of symbolic computation, similar to th= e symbolic differentiation or integration that is performed by tools such a= s Mathematica. The problem of obtaining the best extension is complicated = by the fact that some familiar properties (such as that x-x=3D0 for any x, = and distributivity of multiplication over addition) do not apply to interva= ls, and is a good source of examples for calculational programming. Finally, we will present a larger application, a generic program for inter-= temporal optimization with dynamic programming. This kind of problem is ub= iquitous in economic modeling, and hence in many integrated assessment mode= ls, such as those aiming to compute costs of climate change. It has both al= gebraic aspects (the organization of the computation for backward induction= ), which can be tackled with the categorical methods presented, and numeric= al ones (the local optimization techniques), where interval analysis can be= used. The Seminar is organized by: * Paul Flondor, Professor of Mathematics at Politehnica University Bucharest (pflondor@yahoo.co.uk) * Jeremy Gibbons, Professor of Computing at the University of Oxford (jeremy.gibbons@cs.ox.ac.uk) * Cezar Ionescu, researcher at Potsdam Institute for Climate Impact Research (ionescu@pik-potsdam.de) HOW TO APPLY Applications to participate should include * full name and address, including e-mail address * short CV, present position, university * name of supervisor of PhD thesis * a short summary of previous work and interest and should be sent preferably by e-mail (pdf files) to: Prof. Dr. Dietmar Kr=F6ner Mathematisches Forschungsinstitut Oberwolfach Schwarzwaldstr. 9-11 77709 Oberwolfach-Walke Germany seminars@mfo.de The deadline for applications is 15th September, and the number of particip= ants is restricted to 25. The Institute covers accommodation and food; than= ks to support from the Carl Friedrich von Siemens Foundation, some contribu= tion may also be made towards travel expenses. For more information, contac= t the organizers or see the Institute's webpage: http://www.mfo.de/scientific-programme/meetings/oberwolfach-seminars Jeremy.Gibbons@cs.ox.ac.uk Oxford University Department of Computer Science, Wolfson Building, Parks Road, Oxford OX1 3QD, UK. +44 1865 283521 http://www.cs.ox.ac.uk/people/jeremy.gibbons/ --Apple-Mail-297-721054627 Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset=iso-8859-1
Dear all,

I'm = lecturing in November at a week-long seminar on rigorous (ie functional) pr= ogramming techniques for computational science. There is funding for bed, b= oard, and part of travel costs. If you're in contact with computational sci= entists who might be interested, could you pass this on, please? Applicatio= n deadline is 15th September. 

Thanks,
Jeremy

 &= nbsp;*

Call for Participation

OBERWOLF= ACH SEMINAR ON 
MATHEMATICS FOR SCIENTIFIC COMPUTATION
=

Mathematisches Forschunginstitut Oberwolfach
= 24th to 30th November 2013


MOTIVATION

Computational science today de= pends crucially on simulations, which are typically based on algorithms tha= t have a sound mathematical justification.  For example, an iterative = procedure such as Newton's method is motivated by appealing to the properti= es of twice continuously differentiable functions and their Taylor expansio= n, which also yield convergence conditions and approximation estimates.

These algorithms are then implemented on a computer, = using a programming language such as Fortran or C++.  Often, the imple= mentation will introduce new computational steps and otherwise modify the s= tructure of the mathematical algorithm - for handling or reducing round-off= errors, enabling more efficient memory access, exploiting parallelization,= and so on.  As a result, the final implementation usually looks very = different from the mathematical algorithm, and the justification given for = the latter does not directly extend to the former.  But if we are to e= nsure the correctness of simulations, we need mathematical certainty for bo= th.

We aim to bring to the scientific programming = community mathematical techniques that allow us to achieve the transition f= rom mathematical algorithm to efficient implementation in a principled mann= er, with each step motivated by the application of a mathematical theorem. = The intended participants are students and researchers in computational sci= ence (including areas such as engineering, biology, and economics), and any= scientists dissatisfied with state of the art in transforming mathematics = into code. They will be equipped subsequently to make a significant contrib= ution to increasing the correctness of the simulations that play such an im= portant role in current scientific activity.


<= /div>
CONTENT

This rigorous approach to progra= mming is most easily presented in the framework of functional programming: = program calculation can be reduced to straightforward equational reasoning,= provided that all program variables are immutable.  Accordingly, we w= ill introduce the basic syntax and ideas using Haskell, currently the one o= f the most successful functional programming languages.  The emphasis = is not on functional programming as such, and even less so on a specific la= nguage such as Haskell; but rather, on the mathematics behind program devel= opment, which can then be transferred to other contexts, such as imperative= programming, or parallel programming.

This mathem= atical foundation lies in category theory, which unifies what could otherwi= se appear as a large collection of "bite-sized" theorems for program develo= pment, too many for any developer to remember and use efficiently.  Ca= tegory theory is a broad subject: we will limit ourselves to what is essent= ial as a framework for datatypes and programs (functors, universal properti= es, algebras, monads). The many examples, such as fusion (loop elimination)= , optimal bracketing (important for non-associative operations such as thos= e on floating-point numbers), or parallel programming skeletons (such as Go= ogle's MapReduce), will be readily understandable and relevant to scientifi= c computing practitioners.

One of the most effecti= ve ways to counter floating-point errors and to obtain validated results is= to use interval analysis, which however requires more complex data structu= res and algorithms than is common in other areas of scientific computation.=  Extending a function on real or floating-point numbers to one on int= ervals is a matter of symbolic computation, similar to the symbolic differe= ntiation or integration that is performed by tools such as Mathematica. &nb= sp;The problem of obtaining the best extension is complicated by the fact t= hat some familiar properties (such as that x-x=3D0 for any x, and distribut= ivity of multiplication over addition) do not apply to intervals, and is a = good source of examples for calculational programming.

=
Finally, we will present a larger application, a generic program for i= nter-temporal optimization with dynamic programming.  This kind of pro= blem is ubiquitous in economic modeling, and hence in many integrated asses= sment models, such as those aiming to compute costs of climate change. It h= as both algebraic aspects (the organization of the computation for backward= induction), which can be tackled with the categorical methods presented, a= nd numerical ones (the local optimization techniques), where interval analy= sis can be used.

The Seminar is organized by:

* Paul Flondor, Professor of Mathematics at Politehnic= a University Bucharest
* Jeremy Gibbons, Professor of Co= mputing at the University of Oxford
* Cezar = Ionescu, researcher at Potsdam Institute for Climate Impact Research
<= div>  (ionescu@pik-potsdam.d= e)


HOW TO APPLY

<= /div>
Applications to participate should include

* full name and address, including e-mail address
* short CV, = present position, university
* name of supervisor of PhD thesis
* a short summary of previous work and interest

and should be sent preferably by e-mail (pdf files) to:
  Prof. Dr. Dietmar Kr=F6ner
  Mathematisc= hes Forschungsinstitut Oberwolfach
  Schwarzwaldstr. 9-11
  77709 Oberwolfach-Walke
  Germany
&= nbsp; seminars@mfo.de

The deadline for applications is 15th September, and the nu= mber of participants is restricted to 25. The Institute covers accommodatio= n and food; thanks to support from the Carl Friedrich von Siemens Foundatio= n, some contribution may also be made towards travel expenses. For more inf= ormation, contact the organizers or see the Institute's webpage:
=

<= span class=3D"Apple-style-span" style=3D"font-family: 'Lucida Grande'; ">Jeremy.Gibbons@cs.ox.ac.uk<= /span>Oxford University Department of Computer Science,
<= /span>+44 1865 283521


= --Apple-Mail-297-721054627--