Showing papers on "Convolutional neural network published in 1992"

Visual Pattern Recognition with Neural Networks

Abstract: The “neocognitron” was first proposed as a hierarchical neural network model for the mechanism of visual pattern recognition in the brain. It is capable of deformation-resistant pattern recognition. Various experiments have demonstrated its powerful ability to recognize visual patterns. For example, the authors have designed several systems which recognize hand-written characters, such as, a system recognizing ten numerals, and a system recognizing alphanumeric characters. This paper also discusses recent advances in the neocognitron. The network has been modified to have an architecture closer to that of the real biological brain, and a new learning algorithm has been introduced.