Enhanced reliability of multilayer perceptron networks through controlled pattern rejection
Germano Crispim Vasconcelos, M.C. Fairhurst, D.L. Bisset · Electronics Letters · 1993
Backpropagation networks are studied in the context of the recognition of handwritten and machine printed characters and, specifically, the problem of the recognition of ‘false—positive’ patterns is investigated. The idea is proposed of integrating into the processing architecture an independent set of ‘guard units’ which act as basic matchers, rejecting patterns not belonging to the training classes.