Optimal sizing of feedforward neural networks: Case studies

Abstract

Feedforward neural networks with sigmoidal hidden layers can be used to approximate any continuous functions within allowable tolerances in accuracy. However no systematic rules are available for the determination of the optimal number of hidden nodes for the networks. An algorithm is proposed which can be employed to find the optimal number of hidden nodes in FNNs used for function approximation. The algorithm has advantages over the conventional trial and error method as the computational time will be reduced and there will be a lower probability of solutions getting stuck at local minima. Two case studies are made to investigate the performance of the algorithm yielding encouraging results.