Recognitron-a neural net model for character recognition

A neural network model called a recognitron is described. The recognitron uses a global mechanism for feature detection as compared to another net, called the neocognitron, which applied a local detection mechanism. The net consists of four layers and the output subnet. The Hamming net is used as the output subnet. The results of computer simulation of the recognitron are presented to show its ability for extracting and mapping features from noisy images of handwritten characters.<
>