Least Square Method Based Evolutionary Neural Learning Algorithm

In this paper, we present a new idea of evolving weights for the Artijicial Neural Networks (ANNs). We propose a novel hybrid learning approach for the training of a feed-forward ANN. The approach combines evolutionary algorithms with matrix solution methods such as GramSchmidt etc., to achieve optimum weights for hidden and output layers. Our hybrid method is to apply the evolutionary algorithm in the first layer and the least square method (U) in the second layer of the ANN. A two-layer network is considered. The hidden layer weights are evolved using the evolutionary algorithm (EA). When a certain number of generation or error goal in t e m of RMYClass error is reached, the training stops. We start with a small number of hidden neurons, and then the number is increased gradually. We have applied our algorithm for XOR, IO-bit odd parity and handwritten segmented characters recognition problems. The implementation of the algorithm was done in MATLAB & C. Experiments show some promising results when compared with other evolutionary based algorithm only in terms of results in classification rate and time complexity.