School on Large Scale Problems in Machine Learning and Workshop on Common Concepts in Machine Learning and Statistical Physics (smr2361)
|20 - 31 August|
|H.J. Kappen, M. Opper, R. Zecchina. Local Organiser: M. Marsili|
|SNN Adaptive Intelligence, Radboud University of Nijmegen, The Netherlands|
|M. de Comelli|
|Machine learning is a branch of computer science that addresses the issue of automatically recognizing complex patterns in empirical data and make intelligent decision based on the representation of these patterns. Optimal solutions are the result of the interplay between reproducing empirical observations and providing answers that generalize optimally for yet unobserved data. This interplay is similar to that between energy and entropy in statistical mechanics. Many applications of machine learning deal with large problems with many variables, and empirical data (training set) is limited to a very small fraction of the space of possible inputs. In statistical mechanics jargon, the entropic contribution plays a dominant role.
Concepts and tools in statistical physics have been proved useful in applications to ensembles of problems in computer science, such as constraint satisfaction or error-correcting codes, by revealing the structure of the space of solutions in typical cases, assessing the performance of algorithms and suggesting new efficient heuristics. The School and Workshop aim at investigating the relevance of these concepts and techniques for the broader field of machine learning, exploring opportunities of mutual cross fertilization between the two communities.
The School provides an overview of the key concepts of Machine Learning which are central to large scale applications. Lectures will be delivered by experts in machine learning and target an audience of researchers and graduate students with a background in physics and/or mathematics. It will last for a week and a half, and be followed by a two days Workshop. Both the School and Workshop will have a particular emphasis on applications of machine learning to computational biology.
TOPICS COVERED BY THE SCHOOL:
- Graphical models
- Approximate inference
- Control theory and reinforcement learning
- Neural networks
- Non-Bayesian kernel methods
- Algorithms for Bayesian kernel machines on large scale problems
- Nonparametric Monte Carlo techniques and Dirichelet processes
- Machine learning in systems biology
Applicants willing to present a poster are invited to submit a short one-page abstract along with their online application form.
The Abdus Salam International Centre for Theoretical Physics (ICTP)
Strada Costiera 11 ?I-34151 Trieste
?Telefax: +39-040-224163? Telephone: +39-040-2240305
DEADLINE for receipt of applications: 20 APRIL 2012