Abstract
Among the various neural networks proposed, the interpattern associative (IPA) neural network model1 is quite special because it pays more attention to the special features than to the common ones among the reference patterns, and therefore it is more powerful than other neural networks, e.g., the Hopfield associative neural network, in recognizing the patterns that are similar. Its interconnection matrix is defined by, according to some special logical relations, a three-level set {−1,0,1}. The positive (excitatory) connections are indicated by 1, the negative (inhibitory) connections by −1, and no connections by 0. There is a major problem in the implementation of such an optical neural network because the implementation of such a system is complicated: it requires the subtraction of light intensities at the input ports of the neurons, while with incoherent light only addition is possible.
© 1992 Optical Society of America
PDF ArticleMore Like This
Wenlu Wang, Shutian Liu, Jie Wu, and Chunfei Li
CMH7 Conference on Lasers and Electro-Optics (CLEO:S&I) 1992
Xiang Y. Yang, Taiwei Lu, and Francis T. S. Yu
MN6 OSA Annual Meeting (FIO) 1990
Taiwei Lu, Xin Xu, Shudong Wu, and Francis T. S. Yu
WU6 OSA Annual Meeting (FIO) 1989