OPTICAL NEURAL-NETWORK USING FRACTIONAL FOURIER-TRANSFORM, LOG-LIKELIHOOD, AND PARALLELISM

Optical neural networks based on the fractional Fourier transform (FRT ) are examined in connection with log-likelihood and parallelism. It i s found that a neural network using FRT and the mean square error clas sifies patterns far better than the one using Fourier transform and th e mean square error. However, the classification performance of this n eural network is limited. In order to speed up its learning convergenc e, the mean square error is replaced first with the log-likelihood. Th en, parallelism is introduced to the FRT neural network with the log-l ikelihood and its effect on the neural network is studied. Finally, it is demonstrated that the combination of FRT, log-likelihood, and para llelism significantly improves both the learning convergence and the r ecall rate of the neural network. (C) 1998 Elsevier Science B.V. All r ights reserved.