A Connectionist Visual Field Analyzer
暂无分享,去创建一个
Abstract
We have built an artificial neural network to analyze visual field maps. The program uses the back-propagation method of learning and has been trained with 62 different types of classical visual field defects. When tested against 18 unknowns, it was able to correctly classify 17 of them. Connectionist networks are thus shown to be capable of efficiently recognizing two-dimensional geometric patterns.