Abstract

   In this paper, a genetic algorithm (GA) has been proposed as an optimization method in order to evolve the artificial neural network with the aim of optimizing connection weights and network architecture for a pattern recognition and a classification tools. The proposed algorithm has been applied and tested for optimizing connection weights in recognition of a different group of patterns. Also, it has been studied to implement the optimization of the network architecture. Several training back-propagation algorithms are tested in an attempt to find an ideal artificial neural network (ANN) training algorithm in pattern recognition and classification. Tests have shown that the training optimized by GA method is faster than the BP training algorithm and produces a lower SSE in a smaller number of epochs.