3rd IASC world conference on
Computational Statistics & Data Analysis
Amathus Beach Hotel, Limassol, Cyprus, 28-31 October, 2005
 
Title: Model Selection, Computational Methods, and Optimization Heuristics

Description:

In recent years, the statistical literature has placed more and more emphasis on model selection criteria. The problem is to choose the best approximating model among a class of competing models by a suitable model selection criterion given a finite data set. That model which optimizes the criterion is chosen to be the best model. Therefore, the goal of model selection is to optimize the quality of inference, and it advocates good modeling. The emphasis on model selection can be traced back to the pioneering work of Akaike and his classic: Akaike's (1973) information criterion (AIC). Since then, a number of other model selection criteria such as Bayesian Model Selection, Minimum Description Length, Information Complexity, and Minimum Message Length criteria, and others have been proposed, each with its own rationale and point of view. The developments and the introduction of model selection criteria and techniques and recent advancements in personal computers have marked the beginning of an era of systematic approach to model evaluation and selection. These techniques have become popular in many cross-disciplinary fields of science such as economics, biology, bio-informatics, data mining, ecology, engineering, medicine, psychology, the social and behavioral sciences, to mention a few.

The statistical techniques for model selection increasingly rely on demanding and sophisticated numerical and computational methods. Estimation and modeling as well as model selection raise many optimization problems which turn out to be intractable by standard numerical methods in high dimensional data. One way to deal with such a situation consists in simplifying models and/or applying new and fast algorithms and optimization heuristics such as local search methods (Simulated Annealing, Threshold Accepting, Tabu Search), Neural Networks, Genetic Algorithms, and hybrid methods and many others.

This track on "Model Selection, Computational Methods, and Optimization Heuristics" invites proposals for sessions covering on one hand statistical and computational aspects of model selection as well as real applications in cross-disciplinary fields; and on the other hand contributions which address optimization problems using model selection as the fitness function by making use of modern algorithms and heuristics.

If you are interested in organizing a session, please contact one of the track co-chairs: Prof. Bozdogan, or Gilli, or Winker. Details will be communicated with each session organizers.

Co-Chairs:

Hamparsum Bozdogan,
McKenzie Professor in Business
Department of Statistics, 336 SMC
The University of Tennessee
Knoxville, TN 37996-0532 USA
Tel: + 865 974-2556
Fax: + 865 974-2490
WWW: http://web.utk.edu/~bozdogan
E-mail: bozdogan@utk.edu

Manfred Gilli
Department of Econometrics
University of Geneva
40, Bd du Pont d'Arve
1211 Geneve 4
Switzerland
Tel: +41 22 37 98222
Fax: +41 22 37 98299
E-mail Manfred.Gilli@metri.unige.ch
Peter Winker
Faculty of Economics, Law and Social Sciences
University of Erfurt
P.O. Box 900221
D-99105 Erfurt, Germany
Tel: +49-361-737-4591
Fax: +49-361-737-4599
E-mail: Peter.Winker@uni-erfurt.de

ml>