| pubmed-article:8049788 | pubmed:abstractText | Stochastic learning for a kind of feature detectors is studied. The technique is applied to multilayer neural networks intended for T-C discrimination. After evaluating the generalization error function, the training and generalization curves are found by Monte Carlo simulations. Different phase transitions take place and the overall behaviour agrees with previous discussions concerning designs of the same type. | lld:pubmed |
