gmane.comp.mathematics.reliable-computing

[Reliable Computing] FW: NAFIPS'09: Paper Submission Deadline Extended

Kreinovich, Vladik — 2008-12-04T18:18:09

Dear Friends, As you may remember, the organizers approved an interval session at NAFIPS that Weldon Lodwick and myself are organizing. Please submit to this session! **************************************************************************************** From: Asli Celikyilmaz [mailto:asli-aFE07iDfcCIb0cFwG/AQJIdd74u8MsAO< at >public.gmane.org] Dear colleagues, The deadline for the paper submission for NAFIPS'09 has been extended to January 10, 2009. We would greatly appreciate your contribution to the success of the conference. Please visit the NAFIPS 2009 website for recent updates and news (http://nafips2009.ewu.edu/). The paper submission website is now OPEN. Thank you very much in advance for your support. Best regards, ********************************************************** NAFIPS'09 -- 28th North American Fuzzy Information Processing Society Annual Conference University of Cincinnati, Cincinnati, Ohio, USA June 14-17, 2009 http://nafips2009.ewu.edu/ Dear Colleague, You

[Reliable Computing] UAI 2009: Preliminary call for papers

Andrew Y. Ng — 2008-12-03T03:44:36

The 25th Conference on Uncertainty in Artificial Intelligence (UAI 2009) June 18th - 21st, 2009, Montreal, Canada http://www.cs.mcgill.ca/~uai2009 Preliminary Call for papers The 25th Conference on Uncertainty in Artificial Intelligence (UAI 2009) will take place on June 19-21, 2009 in Montreal, Canada. UAI 2009 is co-located with ICML 2009 and COLT 2009, and will take place following ICML 2009 and concurrently with COLT 2009. There will also be a joint UAI/ICML/COLT workshop day on June 18th. We encourage submissions that report on theoretical or methodological advances in modeling, inference, learning and decision making under uncertainty. Submissions reporting on novel and insightful applications of these techniques within intelligent systems are also strongly encouraged. Examples of such application areas include, but are not limited to, computational biology, computer vision, speech processing, computational linguistics, information retrieval, medical systems, multi-agent

[Reliable Computing] Re: A simple interval challenge

Arnold Neumaier — 2008-12-04T08:19:49

The attachment contains the Third intermediate report (December 4, 2008) to my simple interval challenge from November 26, 2008. The current best public algorithm achieves 112.5% of the optimal width. Please report even now suboptimal solutions if they are within 150% of the optimal width and make use of ideas not yet present! Main changes compared to the second report: - algorithms by John Pryce and myself - symmetries: I had forgotten the continuous scaling symmetry - new comments on the problem of minimizing the coefficient of c that make it an interesting test problem for global optimization Arnold Neumaier A simple interval challenge (November 26, 2008) --------------------------- In discussions with Nate Hayes, he mentioned the following test problem (arising from a problem in computer vision) as an example for the potential efficiency of modal interval arithmetic. With his permission, I make the problem public, adding performance evaluation criteria for a public contest.

[Reliable Computing] Re: A simple interval challenge

Arnold Neumaier — 2008-12-01T16:23:48

The attachment contains the Second intermediate report (December 1, 2008) Based on requests, I made the rules a bit more precise; see the items in the challenge formulation. The current record is [-3.274656,8.832152] (110% of optimal width) by Nate Hayes, with an unknown algorithm. Arnold Neumaier A simple interval challenge (Nov. 26, 2008) --------------------------- In discussions with Nate Hayes, he mentioned the following test problem (arising from a problem in computer vision) as an example for the potential efficiency of modal interval arithmetic. With his permission, I make the problem public, adding performance evaluation criteria for a public contest. Second intermediate report (December 1, 2008) Based on requests, I made the rules a bit more precise; see the items in the challenge formulation below. Challenge (based on a problem posed by Nate Hayes) --------- Find a cheap and good enclosure for f := (a(w^2+x^2-y^2-z^2)+2b(xy-wz)+2c(xz+wy))/(w^2+x^2+y^

[Reliable Computing] Re: A simple interval challenge

Arnold Neumaier — 2008-12-01T07:42:38

Ray Moore schrieb: 1788 public.gmane.org> is the list of the future IEEE interval standard. Actually, you'd send it to both 1788 and interval public.gmane.org>, which will happen automatically as a reply-all to my first simple interval challenge mail. Arnold

[Reliable Computing] Fw: A simple interval challenge

Ray Moore — 2008-11-30T18:32:25

Colleagues, At the suggestion of Arnold Neumaier, I am sending a pdf file of a paper I was working on a couple of years ago. It is an unfinished piece of work, but may be relevant to the ongoing discussion of the topic of improving interval bounds on ranges of values of functions without using box splitting. I was attempting to see what can be done in the special case of multinomials. Ray Moore ----- Original Message ----- From: "Arnold Neumaier" public.gmane.org> To: "Ray Moore" public.gmane.org> Sent: Sunday, November 30, 2008 5:33 AM Subject: Re: A simple interval challenge

[Reliable Computing] simple interval challenge

2008-11-29T12:17:49

Dear Prof. Neumaier, Could you, please, be a little bit more specific about the “simple interval challenge” problem. Are the variables w, x, y and z intervals that have the properties: w<0, y>0 and 0 is in x, 0 is in z? How “constant” are the interval constants a, b and c: can they change their bounds keeping, however, the properties: a>0, b and c are symmetric? The answer to these questions makes the specificity of the problem considered clearer and can influence the efficiency of the solution suggested. I have had a first try at finding an enclosure of the range sought, using: a rearrangement in the form of a sum of two terms and application of G-intervals (affine arithmetic). The first (rearranged) term turned out to be the one suggested by Prof. Raymond Moore. The second term is enclosed using a simple rule for multiplication of G-intervals. My enclosure is [-3.8080, 9.3655] and is obtained using 13 standard interval operations and 9 G-interval operatio

[Reliable Computing] Re: A simple interval challenge

Arnold Neumaier — 2008-11-27T17:31:32

Yes. This is the ultimate purpose of the challenge. I'll provide as much detail as I can get from those submitting answers to the challenge. In particular, I'll try to reprogram the recipes in Intlab (if they are not yet in this format, which would be a significant help; see below for a template) to check the results before putting them online. It is likely that all tricks can be automatized to work not only for this example but in more general situations, though it may not always be easy to decide which trick should be applied when. So these choices should probably be made in a semi-automatic way based on statistics on trial inputs and perhaps also guided by interventions from a graphical user interface. The results will also show those not so familiar with interval methods but more concerned about low level implementations which issues really matter for efficiency in user applications. The standard should help expert programmers to exploit the potential of interval methods to the fullest, while not s

[Reliable Computing] Re: A simple interval challenge

Jean-Pierre Merlet — 2008-11-27T16:40:09

Ok, fine by me. But what will also be useful is a full description of the manipulation that led to the various bounds. This may be indeed interesting to see if we can somewhat automatize the process JPM

[Reliable Computing] Re: A simple interval challenge

Arnold Neumaier — 2008-11-27T13:55:54

Of course, the box over which the range needs to be enclosed changes from case to case; only the function is the same. So one needs an algorithm that applies uniformly over a wide range of input boxes. But for the evaluation in a contest, one must provide concrete data. The box chosen is a fairly difficult case for this function. Even when a method is fine-tuned to this paricular case, the computation must provide a proof of enclosure, hence is nontrivial. It will give insight into what sort of tricks are useful for creating a more complete routine that works well on all input boxes. And the same kinds of tricks can be applied or adapted to many other problems of this kind. To expose insights and tools are the final objectives in the contest. Arnold Neumaier

[Reliable Computing] Re: A simple interval challenge

Arnold Neumaier — 2008-11-27T12:25:01

Jean-Pierre Merlet schrieb: This is a misunderstanding. Intended by me for the contest was only category 1, to demonstrate how much naive techniques can be improved in particular cases that frequently occur as part of a bigger application, so that tuning it for optimal performance (and perhaps even putting it into hardware) is worthwhile. Category 2 amounts to using a package that does range bounding using branch and bound, which will yield very accurate ranges but at much higher costs, not tolerable when the enclosure is needed millions of time. So this contest is not about testing the performance of existing software but about testing (and making public) the bag of tricks experts have to squeeze the best out of a very limited budget for one range enclosure. I am not interested in enclosures that need more than 200 effective interval operations; this probably eliminates most branch and bound packages (quite apart from the fact that it is difficult to count there the number of operations). The results

[Reliable Computing] Re: A simple interval challenge

Jean-Pierre Merlet — 2008-11-27T11:29:40

Arnold, all, I think that there is a flaw in this test or at least that the results should be divided into 2 categories: 1- results provided by interval experts that will spend some time re-arranging expressions for this particular case to get the "best" result 2- results provided by general purpose interval software (the one that the end-user will use because they don't have the knowledge to do otherwise). In that case the data should be provided "as it" because the system and bound may be the result of some other calculation and are not known in advance. Best JPM

[Reliable Computing] Re: A simple interval challenge

Arnold Neumaier — 2008-11-27T09:16:38

First intermediate report (Nov. 27, 2008) [...] I had a mistake in my program. The corrected version gets only the enclosure [-3.6517,9.2223] using 12 real and 54 interval operations, hence 60 effective operations. This is 117% of optimal width, and has excess cost 1.7e-3. (The excess cost is very sensitive to the width.) Raymon Moore provided a pure rearrangement evaluation u=w^2+x^2;v=y^2+z^2; s=b*(x*y-w*z)+c*(x*z+w*y); f=a*(1-2/(u/v + 1)) + 2*s/(u+v); with 24 operations, giving the enclosure [-5.8080,11.3655], which is 157% of optimal width, and has excess cost 0.79. Note that the rearrangements s=(b*x+c*w)*y+(c*x-b*w)*z s=(b*y+c*z)*x+(c*y-b*z)*w of s lead to inferior results. Mihaly Markot calculated the global extrema to high accuracy. The global minimum -2.95607850118520?9 is attained at (a,b,c)=(9,1,1), (w,x,y,z)=(-0.6,-0.04181974?8,0.7,-0.2). The global maximum 8.00936984210601?8 is attained at (a,b,c)=(9,1,-1), (w,x,y,z)=(-0.9,0.2,0.3,0.04115383?9). Thu

[Reliable Computing] Re: pointers to typical and non-typicalIA applications

Chenyi Hu — 2008-11-26T15:53:57

C. Hu and B. Kearfott Interval Matrices in Knowledge Discovery in Knowledge Processing with Interval and Soft Computing, Eds. C. Hu, B. Kearfott, A. de Korvin, V. Kreinovich, pp. 99-118, Springer, 2008. C. Hu, L. He, and S. Xu, Interval Function Approximation and Applications, in Knowledge Processing with Interval and Soft Computing, Eds. C. Hu, B. Kearfott, A. de Korvin, V. Kreinovich, pp. 119-134, Springer, 2008. C. Hu Interval Rule Matrices for Decision Making, in Knowledge Processing with Interval and Soft Computing, Eds. C. Hu, B. Kearfott, A. de Korvin, V. Kreinovich pp. 135-146, , Springer, 2008. C. Hu, and P. Hu Interval-Weighted Graphs and Flow Networks, in Knowledge Processing with Interval and Soft Computing, Eds. C. Hu, R. B. Kearfott, A. de Korvin, V. Kreinovich, pp. 167-182, Springer, 2008. C. Hu Using interval function approximation to estimate uncertainty in Interval / Probabilistic Uncertainty and Non-Classical Logics, Advances in Soft Computing, 46, Eds. V. Huynh, Y. Nakamo

[Reliable Computing] Re: pointers to typical and non-typical IA applications

Arnold Neumaier — 2008-11-26T12:11:57

Arnold Neumaier schrieb: Applications to computer-assisted proofs can be found in A. Neumaier, Computer-assisted proofs, in: (W. Luther and W. Otten, eds.) Proc. 12th GAMM-IMACS Symp. Sci. Comp., (SCAN 2006). IEEE Computer Society, 2007. http://www.mat.univie.ac.at/~neum/papers.html#caps Applications to uncertain large-scale truss structures are in A. Neumaier and A. Pownuk, Linear systems with large uncertainties, with applications to truss structures, Reliable Computing 13 (2007), 149-172. http://www.mat.univie.ac.at/~neum/papers.html#linunc Applications to uncertain partial differential equations are in A. Neumaier, Certified error bounds for uncertain elliptic equations, J. Comput. Appl. Math. 218 (2008), 125-136. http://www.mat.univie.ac.at/~neum/papers.html#pdebounds Arnold Neumaier

[Reliable Computing] A simple interval challenge

Arnold Neumaier — 2008-11-26T10:24:17

In discussions with Nate Hayes, he mentioned the following test problem (arising from a problem in computer vision) as an example for the potential efficiency of modal interval arithmetic. With his permission, I make the problem public, adding performance evaluation criteria for a public contest. Please send results directly to me at public.gmane.org>; I'll post summaries when something interesting happens. Arnold Neumaier Challenge: --------- Find a cheap and good enclosure for f := (a(w^2+x^2-y^2-z^2)+2b(xy-wz)+2c(xz+wy))/(w^2+x^2+y^2+z^2) given the following bounds on the variables a,b,c,w,x,y,z: a in [7,9] b in [-1,1] c in [-1,1] w in [-0.9,-0.6] x in [-0.1,0.2] y in [0.3,0.7] z in [-0.2,0.1] The computation together with general theoretical results must constitute a proof that the result is a rigorously valid enclosure of the range. (Thus, the relevant cost is that of checking that some data provided are a certificate for

Re: [Reliable Computing] pointers to typical and non-typical IA applications

Christian Jansson — 2008-11-23T18:53:49

Janos Pinter asked for recent surveys considering IA applications. There are some recent developments and surveys in optimization, where especially Janos may be interested. First, there is survey written by Arnold Neumaier A. Neumaier, Complete Search in Continuous Global Optimization and Constraint Satisfaction, pp. 271-369 in: Acta Numerica 2004 (A. Iserles, ed.), Cambridge University Press 2004. This survey covers the state of the art of techniques for solving general purpose constrained global optimization problems and continuous constraint satisfaction problems, with emphasis on complete techniques that provably find all solutions. Sections on interval arithmetic, constrained propagation and local optimization of important problem transformations follows, in particular of linear, convex, and semilinear (= mixed integer linear) relaxations that are important for handling larger problems. In practice, a user needs a reliable and robust software that works (in most cases) also for larger pro

RE: [Reliable Computing] FW: die Welt online

Kreinovich, Vladik — 2008-11-23T17:30:36

Dear Alexandre, Many thanks for your clarification. I am familiar with Arnold's and others papers, so I did not mean at all to imply that Banhelyi et al, were the first to apply intervals to chaos. There were several other related papers as well, some mentioned by Martine Berz in his SCAN'08 talk. Actually, Banhelyi et al cite several of prior papers in their work. After re-reading my original posting I agree with you that to those readers who are not familiar with this history this posting may have sounded this way, I apologize for the bad wording. The point of my posting was not to claim that this was the first time this is done (this is not), but to emphasize that this paper got a wide publicity. Specifically, Banhelyi et al. prove chaos-related properties of a specific (and very natural) system for which chaos was empirically found out in 1999 but not proven until 2008: forced damped pendulum $ml \ddot x+ b\cdot x+ mg \sin(x)=A\cos(wt)$ -- probably among the most natural and simple

[Reliable Computing] FW: die Welt online

Kreinovich, Vladik — 2008-11-22T05:13:43

Dear Friends, In their recent paper Balazs Banhelyi, Tibor Csendes, Barnabas M. Garay, and Laszlo Hatvani, "A computer-assisted proof of \Sigma_3-chaos in the forced damped pendulum equation, SIAM Journal on Applied Dynamical System, 2008, Vol. 7, pp. 843-867 The authors used interval computations to provide a proof of the chaotic character of a system -- and thus, to confirm empirical observations. This paper was cited among the major scientific achievements on the online site of die Welt, one of the major German-language newspapers http://www.welt.de/welt_print/article2752854/Wissenschaft.html Congratulations to the authors! Vladik P.S. The paper can be found on Tibor Csendes' webpage, as http://www.inf.u-szeged.hu/~csendes/fdpendul.pdf

Re: [Reliable Computing] pointers to typical and non-typicalIA applications

Chenyi Hu — 2008-11-21T22:31:57

Baker, et al, Thanks for mention our recent work on applying interval computing in computational finance and economics at SCAN 2008. Here are some of the results published (or accepted for publication) recently. 1. L. T. He, C. Hu, M. Casey, Prediction of Variability in Mortgage Rates: Interval-Computing Solutions, , J. Risk Finance, accepted for publication. 2. L. T. He, C. Hu,Impacts of Interval Computing on Stock Market Variability Forecasting, J. Computational Economics, online available at http://dx.doi.org/10.1007/s10614-008-9159-x, in printing. 3. L. T. He, C. Hu,Midpoint Method and Accuracy of Variability Forecasting, J. Empirical Economics, accepted for publication. 4. D. Collins, C. Hu, Studying Interval Valued Matrix Games with Fuzzy Logic, DOI 10.1007/s00500-007-0207-6, J. Soft Computing, 12(2), 147-155, 2008 5. A. Han, X. Chen, C. Hu, S. Xu, Interval Computing in Econometrics, J. Management Review, in Chinese, Vol. 20, No. 5, pp. 37-43, 2008 6. L. T. He, C. Hu,Impacts of Interval M

Re: [Reliable Computing] pointers to typical and non-typical IA applications

Noel Titus — 2008-11-21T20:40:02

Interval analysis has been applied to product design - the early stages of product design are characterized by uncertainty and ambiguity, which can be handled by interval methods - this is a recent application. It is also applied in concurrent engineering, where multiple stakeholders are involved in collaborative design of products. There is a design paradigm called set-based design developed by Ward et al.(Quantitiave Inference in a mechanical design "compiler") that also uses interval methods. Ward and his collaborators discovered that something analogous to set-based design is being used at Toyota. On Fri, Nov 21, 2008 at 6:50 AM, R. Baker Kearfott public.gmane.org> wrote: