Backpropagation Learning for Multilayer Feed-Forward Neural Networks Using the Conjugate Gradient Method

In many applications, the number of interconnects or weights in a neural network is so large that the learning time for the conventional backpropagation algorithm can become excessively long. Numerical optimization theory offers a rich and robust set of techniques which can be applied to neural networks to improve learning rates. In particular, the conjugate gradient method is easily adapted to the backpropagation learning problem. This paper describes the conjugate gradient method, its application to the backpropagation learning problem and presents results of numerical tests which compare conventional backpropagation, steepest descent and the conjugate gradient methods. For the parity problem, we find that the conjugate gradient method is an order of magnitude faster than conventional backpropagation with momentum.