Recognition of Spatiotemporal Patterns Using a Nonmonotone Neural Network With Hidden Neurons

Satoshi Murakami, Masahiko Morita and Naoto Sakamoto

Abstract: It has been shown that a nonmonotone neural network model can recognize spatiotemporal patterns without expanding them into spatial patterns. We improve the recognition ability of this model by introducing hidden neurons. We also show a simple method of training the hidden neurons. Computer simulation shows that this model can recognize complicated spatiotemporal patterns.

Keywords: Recognition, Spatiotemporal pattern, Nonmonotone neural network, Hidden neuron

Reprint [ PDF (287KB) / gzipped PostScript (180KB) ] is available