Improved Nelder-Mead Optimization Method in Learning Phase of Artificial Neural Network

Abstract

It is difficult to find the optimum weight values of artificial neural networks for optimization problem. In this study, Nelder-Mead optimization method [17] has been improved and used for determining the optimal values of weights. The results of the proposed improved Nelder-Mead method are compared with results of the standard Nelder-Mead method which is used in ANNs learning algorithm. The most common data sets are taken from UCI machine learning repository. According to the experimental results, in this study better results are achieved in terms of speed and performance.