Dynamics of Supervised Learning with Restricted Training Sets
Disordered Systems and Neural Networks
2009-09-25 v1
Abstract
We study the dynamics of supervised learning in layered neural networks, in the regime where the size of the training set is proportional to the number of inputs. Here the local fields are no longer described by Gaussian probability distributions. We show how dynamical replica theory can be used to predict the evolution of macroscopic observables, including the relevant performance measures, incorporating the old formalism in the limit as a special case. For simplicity we restrict ourselves to single-layer networks and realizable tasks.
Cite
@article{arxiv.cond-mat/9803062,
title = {Dynamics of Supervised Learning with Restricted Training Sets},
author = {A. C. C. Coolen and D. Saad},
journal= {arXiv preprint arXiv:cond-mat/9803062},
year = {2009}
}
Comments
36 pages, latex2e, 12 eps figures (to be publ in: Proc Newton Inst Workshop on On-Line Learning '97)