PDF] The Method of Steepest Descent for Feedforward Artificial Neural Networks

Por um escritor misterioso

Descrição

This paper implements the method of Steepest Descent in single and multilayer feedforward artificial neural networks and calculates the three new update weight equations for taking different activation function in different processing unit separately. In this paper, we implement the method of Steepest Descent in single and multilayer feedforward artificial neural networks. In all previous works, all the update weight equations for single or multilayer feedforward artificial neural networks has been calculated by choosing a single activation function for various processing unit in the network. We, at first, calculate the total error function separately for single and multilayer feedforward artificial neural networks and then calculate the three new update weight equations for taking different activation function in different processing unit separately single and multilayer feedforward artificial neural networks. An example is given to show usefulness of this implementation.