Into silicon: real time learning in a high density RBF neural network

The authors describe an artificial neural network (ANN) architecture that is able to model complex data distributions. This P-RCE network, employs radius-limited and inner-product perceptrons, in a three-layer feedforward architecture that can be trained with real-time speeds using a non-gradient descent, procedural learning algorithm. The authors discuss the use of this network for Parzen-windows estimation of probability density functions, for implementation of the probabilistic neural network (PNN), and for feature extraction in image processing. The authors highlight some of the features of an upcoming silicon implementation of the P-RCE network