Analysis of non-Gaussian data using a neural network
Minglun Gong, Manry · 1989
Summary form only given, as follows. A neural net classifier is applied to non-Gaussian features calculated from numeric hand-printed (NHP) characters. A topological classifier is applied to the same data for comparison. Others have shown that neural nets can be optimal. A neural net is used to verify that the performance of the topological classifier is near optimal. A feature selection approach which utilizes the neural net is proposed and demonstrated as well as a method for fast learning. A reject category is developed so that bad characters are not classified, and the neural net is allowed to express uncertainty.>