Optimum weights and biases for feed forward neural network by particle swarm optimisation

Abstract

This paper introduces particle swarm intelligence (PSI) in feed forward neural network (FFNN) with backpropagation for finding initial weights and biases of the feed forward neural network. The combination of particle swarm optimisation (PSO) and FFNN greatly help in fast convergence of FFNN in classification and prediction to various benchmark problems by overcoming the disadvantage of backpropagation of getting stuck at local minima or local maxima. The benchmarking databases for neural network contain various datasets from various different domains. All datasets represent realistic problems which could be called diagnosis tasks and all but one consist of real world data. Two such benchmarking problems are selected in this paper for comparison and the performance of PSO with FFNN for finding weights and biases is implemented and compared with random initialisation of weights and biases with normal FFNN. The result shows that using PSO minimises the prediction error.