The theory of on-line learning: a statistical physics approach

Abstract

In this paper we review recent theoretical approaches for analysing the dynamics of on-line learning in multilayer neural networks using methods adopted from statistical physics. The analysis is based on monitoring a set of macroscopic variables from which the generalisation error can be calculated. A closed set of dynamical equations for the macroscopic variables is derived analytically and solved numerically. The theoretical framework is then employed for defining optimal learning parameters and for analysing the incorporation of second order information into the learning process using natural gradient descent and matrix-momentum based methods. We will also briefly explain an extension of the original framework for analysing the case where training examples are sampled with repetition.

Divisions: Engineering & Applied Sciences > Mathematics
Engineering & Applied Sciences > Systems analytics research institute (SARI)
Additional Information: The original publication is available at www.springerlink.com
Event Title: Studies in Classification, Data Analysis and Knowledge Organization
Event Type: Other
Event Dates: 2003-01-01 - 2003-01-01
Uncontrolled Keywords: on-line learning,neural networks,statistical physics,natural gradient descent,matrix-momentum,repetition
ISBN: 9783540441830
Full Text Link:
Related URLs: http://www.spri ... detailsPage=toc (Publisher URL)
PURE Output Type: Chapter
Published Date: 2003
Authors: Saad, David ( 0000-0001-9821-2623)

Download

[img]

Version: Accepted Version


Export / Share Citation


Statistics

Additional statistics for this record