Distributed Processing in Continuous Optical Media

通过采用各种“连接主义”模型提出的技术，光学神经网络领域得到了发展。该光学实现首先由Psaltis和Farhat于1985年（Opt. Lett. 10,98(1985)）演示，随后由Anderson于1986年（Opt. Lett. 11,59 (1986)）， Soffer, Dunning， Owechko和Marom （Opt. Lett. 11,118(1986)）以及Yariv和Kwong （Opt. Lett. 11,186(1986)）演示了直接光学联想记忆。从那时起，该领域发展迅速，人们考虑了许多见解、问题和方法。最有趣的是由Psaltis和同事演示的中英三字符识别和翻译，由Anderson和同事演示的光学新颖跟踪滤波器，以及n.r.l.的Lischer和Lee最近的工作。一些需要解决的问题包括多图像全息存储，实现学习，并加快非线性晶体的展示效果。

The area of optical neural networks has evolved by adopting the techniques proposed by various "connectionist" models. The optical implementation was first demonstrated by Psaltis and Farhat in 1985 Opt. Lett. 10, 98 (1985) and was followed by a direct optical associative memory demonstration by Anderson in 1986 Opt. Lett. 11, 59 (1986) and by Soffer, Dunning, Owechko and Marom Opt. Lett. 11, 118 (1986) as well as Yariv and Kwong Opt. Lett. 11, 186 (1986). Since this time the field has evolved rapidly and many insights, problems, and approaches have been considered. Most interesting has been Chinese to English three character recognition and translation demonstrated by Psaltis and co-workers, an optical novelty tracking filter by Anderson and co-workers, and recent work by Lischer and Lee at N. R. L. Some of the problems to be addressed include multiple image holographic storage, implementation of learning, and faster nonlinear crystals exhibiting large effects.