The effect of the threshold function for design feed forward neural network for solving partial differential equations

Abstract EN

In this paper we disperse the outcome of threshold functions for designate feed forward neural network for solution partial differential equations”.

“We utility a multi-layer network having one hidden layer with 7 hidden units (neurons) and one linear output unit with different of threshold function of each unit are logsig , tansig, purelin, tribas and hardlim and use Levenberg – Marquardt (trainlm) training algorithmic rule”.

Finally the terminate of numerical experience are compare to with the true solution in illustrative examples to ratify the precision and effectiveness of the immediate plan.