Efficient supervised learning of multilayer feedforward neural networks

Abstract

The paper presents the efficient training program of multilayer feedforward neural networks. It is based on the best second order optimization algorithms, including variable metric and conjugate gradient as well as application of directional minimization in each step. The method applies the signal flow graph approach for gradient generation. The results of standard numerical tests are given. The efficiency of the program tested on many examples, including symmetry, parity, dichotomy logistic and 2-spiral problems has shown considerable speed-up over the best, already known reported results.< >