Information theory, statistical learning
and pattern recognition,
CIRM, Marseille, France,
7-11 Dec. 1998
Théorie de l'information, statistiques adaptatives
et reconnaissance des formes

Soutenu par le Programme du CNRS ``Modélisation et simulation numérique''
et par le GDR ``Mathématiques des systèmes perceptifs et cognitifs''

List of organizers:
Organisé par:

Robert Azencott	CMLA ENS Cachan
Lucien Birgé	Université Paris VI
Olivier Catoni	Université Paris VI
Marie Duflo	Université de Marne-la-Vallée
Christine Graffigne	Prisme, Université Paris V
Marie-Anne Gruet	INRA Biométrie
Pascal Massart	Université Paris XI
Alain Trouvé	Université Paris XIII

(Names are linked to e-mail addresses, institutions are linked to web pages.)

Aim
Intention

The aim of this workshop is to review and develop interactions between Information Theory, Concentration of measure inequalities, Adaptive statistics and Pattern analysis. Plenary talks will illustrate these four connected fields:
(Le but de ce colloque est de recenser et de développer les interactions entre Théorie de l'information, Inégalités de concentration de la mesure, Statistiques adaptatives et Reconnaissance des formes. Les conférences plénières illustreront ces quatre thèmes:)

Invited speakers
(Conférenciers invités:)

(Names of speakers are linked to e-mail addresses, while institutions are linked to web pages.)

Andrew Barron Yale University
Information Theory in Probability and Statistics ( abstract ).

Amir Dembo Stanford University
Asymptotics of waiting times: large deviations and source coding. ( abstract , joint work with Yiannis Kontoyiannis ).

Donald Geman University of Massachusetts
Pattern Learning and Microimage Coding ( abstract ).

Stuart Geman Brown University
Invariance and Hierarchy in Statistical Image Models. ( abstract ).

Katalin Marton University of Budapest
Measure concentration inequalities for dependent random variables. ( abstract).

David Mumford Brown University
Defining the entropy of natural images
and can we ever hope to calculate it?

Boris Ryabko University of Novosibirsk
Hausdorff dimension, Kolmogorov complexity and Shannon information as the tools for the predicting problem (abstract).

Tjalling Tjalkens Eindhoven University of Technology
Data Compression using the Context-Tree Wighting method (joint work with Frans Willems) (abstract, slides).

Ofer Zeitouni Technion
Conditioning, propagation of chaos and large subsystems behaviour via entropy and sharp large deviation. ( abstract ).

Song Chun Zhu Ohio State University
Learning Probability Models for Texture and Shape By Minimax Entropy. ( abstract ).


Andrew Barron	Yale University
Information Theory in Probability and Statistics ( abstract ).

Amir Dembo	Stanford University
Asymptotics of waiting times: large deviations and source coding. ( abstract , joint work with Yiannis Kontoyiannis ).

Donald Geman	University of Massachusetts
Pattern Learning and Microimage Coding ( abstract ).

Stuart Geman	Brown University
Invariance and Hierarchy in Statistical Image Models. ( abstract ).

Katalin Marton	University of Budapest
Measure concentration inequalities for dependent random variables. ( abstract).

David Mumford	Brown University
Defining the entropy of natural images and can we ever hope to calculate it?

Boris Ryabko	University of Novosibirsk
Hausdorff dimension, Kolmogorov complexity and Shannon information as the tools for the predicting problem (abstract).

Tjalling Tjalkens	Eindhoven University of Technology
Data Compression using the Context-Tree Wighting method (joint work with Frans Willems) (abstract, slides).

Ofer Zeitouni	Technion
Conditioning, propagation of chaos and large subsystems behaviour via entropy and sharp large deviation. ( abstract ).

Song Chun Zhu	Ohio State University
Learning Probability Models for Texture and Shape By Minimax Entropy. ( abstract ).

Propositions of short communications
(proposition de communications courtes)

You can send propositions of short communications, or you can propose a communication on the first day of the workshop, when the schedule of participants talks will be discussed.

We have received the following proposals:

Peter Johansen (University of Copenhagen) Pattern recognition and universal data compression (abstract)
Nicolas Vayatis (CMLA, ENS Cachan) Large deviations bounds for empirical processes (abstract)
Yannick Baraud Variable selection and adaptive estimation in the regression framework.
Flemming Topsoe Instances of exact predictions. Inequalities for information divergence. Game theoretical framework for compression.( abstract).

Discussion session
(table ronde)

Two themes have been retained for the discussion session:

What is the difference between a parameter and a hidden variable ?
Should we use the same models for simulations, pattern recognition and data compression ?

Program
(Programme)

Scientific report
(Rapport scientifique)

For more information, contact
(Pour plus d'informations, contacter)

Olivier Catoni
Processus Stochastiques et Applications
U.M.R. 7599 du C.N.R.S.
Laboratoire de Probabilités,
Case 188,
Université Paris 6,
Tour 46-56, 3ième étage,
4, place Jussieu,
F-75 252 Paris Cedex 05,
tel: (33) (1) 44 27 70 41
e-mail: catoni@ccr.jussieu.fr

Information theory, statistical learning and pattern recognition, CIRM, Marseille, France, 7-11 Dec. 1998 Théorie de l'information, statistiques adaptatives et reconnaissance des formes

List of organizers: Organisé par:

Aim Intention

Invited speakers (Conférenciers invités:)

Propositions of short communications (proposition de communications courtes)

Discussion session (table ronde)

Program (Programme)

Scientific report (Rapport scientifique)