GENERALIZATION TRANSITIONS IN A LAYERED NEURAL NETWORK FOR AN ORDER 3DISTINCTION TASK

Authors
DURANY GF ROMEO A
Citation
Gf. Durany et A. Romeo, GENERALIZATION TRANSITIONS IN A LAYERED NEURAL NETWORK FOR AN ORDER 3DISTINCTION TASK, Physics letters. A, 175(5), 1993, pp. 282-288
Citations number
14
Categorie Soggetti
Physics
Journal title
Physics letters. A → ACNP
ISSN journal
03759601
Volume
175
Issue
5
Year of publication
1993
Pages
282 - 288
Database
ISI
SICI code
0375-9601(1993)175:5<282:GTIALN>2.0.ZU;2-Z
Abstract
We investigate stochastic learning processes in a hidden-layer network trained for a task of order three based on the T-C problem. Our model has a phase transition to perfect generalization for a finite number of examples. The annealed theory approach, which involves an Ising mod el with external field, is in good agreement with Monte Carlo simulati ons.