A bounded exploration approach to constructive algorithms for recurrent neural networks

Abstract

When long-term dependencies are present in a time series, the approximation capabilities of recurrent neural networks are difficult to exploit by gradient descent algorithms. It is easier for such algorithms to find good solutions if one includes connections with time delays in the recurrent networks. One can choose the locations and delays for these connections by the heuristic presented. As shown on two benchmark problems, this heuristic produces very good results while keeping the total number of connections in the recurrent network to a minimum.