EFFICIENT TRAINING OF RECURRENT NEURAL-NETWORK WITH TIME DELAYS

Training recurrent neural networks to perform certain tasks is known t o be difficult. The possibility of adding synaptic delays to the netwo rk properties makes the training task more difficult. However, the dis advantage of tough training procedure is diminished by the improved ne twork performance. During our research of training neural networks wit h time delays we encountered a robust method for accomplishing the tra ining task. The method is based on adaptive simulated annealing algori thm (ASA) which was found to be superior to other training algorithms. It requires no tuning and is fast enough to enable training to be hel d on low end platforms such as personal computers. The implementation of the algorithm is presented over a set of typical benchmark tests of training recurrent neural networks with time delays. Copyright (C) 19 96 Elsevier Science Ltd.