Directly Searching Algorithm for Multi-Layer Feed-forward Artificial Neural Networks

This paper presents a new learning algorithm for multi-layer feed-for-ward artificial neural networks. The subject transforms into an optimal value ofthe multivariate function through taking the error square sum of all output unit onall samples as objective function, that is multivariate function of weighting coeffi-cient.Using batch method of the output error and step-changing directly adjustingmethod of the weight,the algorithm has fast learning speed and high precision.