光折变神经网络(英文)

PHOTOREFRACTIVE NEURAL NETWORKS

本文阐述了几种神经网络模型和模式两分法中感知算法的一种光学执行过程。利用光折变介质动态全息术来存入和修正光学神经网络的加权矩阵。该系统可用以完成模式分类。还介绍了光折变晶体LiNbO_3的一些实验结果。

Neural network models and an optical implementation of the Perception algorithm for pattern dichotomy are described.Dynamic halography in the photorefractive medium is utilized to store and update the weight matrix of the optical neural network.The system can be trained to perform pattern classification.Experimental results using a photorefractive LiNbOa crystal are ateo presented.