Field Theoretical Analysis of OnLine Learning of Probability Distributions
Abstract
Online learning of probability distributions is analyzed from the field theoretical point of view. We can obtain an optimal online learning algorithm, since a renormalization group enables us to control the number of degrees of freedom of a system according to the number of examples. We do not learn parameters of a model, but probability distributions themselves. Therefore, the algorithm requires no a priori knowledge of a model.
 Publication:

Physical Review Letters
 Pub Date:
 October 1999
 DOI:
 10.1103/PhysRevLett.83.3554
 arXiv:
 arXiv:condmat/9911474
 Bibcode:
 1999PhRvL..83.3554A
 Keywords:

 Condensed Matter  Disordered Systems and Neural Networks;
 High Energy Physics  Theory;
 Nonlinear Sciences  Adaptation and SelfOrganizing Systems;
 Physics  Data Analysis;
 Statistics and Probability
 EPrint:
 4 pages, 1 figure, RevTex