Fast learning process of multi-layer neural nets using recursive least squares technique

Summary form only given, as follows. A novel approach for a learning process of multilayer perceptron neural networks using the recursive-least-squares (RLS) technique is proposed. This method minimizes the sum of the square of the errors between the actual and the desired output values recursively. The weights in the net are updated upon the arrival of a new training sample by solving a system of normal equations using the matrix inversion lemma. To determine the desired target in the hidden layers an analog of the backpropagation strategy used in the conventional learning algorithms is developed. This permits the application of the learning procedure to all the other layers. Simulation results on an exclusive-OR example are obtained which indicate significant (an order of magnitude) reduction in the total number of iterations when compared with those of conventional techniques.<>