Neural networks for exact matching of functions on a discrete domain
Y. Shrivastava, Subrata Dasgupta · 1990
In many pattern classification problems it is desirable to realize functions defined on a discrete domain. The authors consider the exact matching of such functions through a three-layered perceptron-based neural network. It is shown that, for a function defined at N distinct points, exact matching can be accomplished through a network with N-1 hidden layer nodes. A construction procedure involving a matrix inversion is also presented.>