Variation Analysis of Hidden Neuron Incitements on Neural Network Performance

Abstract

The modern computational process is used to design and develop an integrated solution as per industrial and social revolution.  The mathematical approaches are used to design computing models to construct an optimum solution according to the nature and complexity of the problem. In general, computational problems are solved using iterative and optimization techniques. The research work represents Neural Network as an approach to compute optimum solution in addition to traditional mathematical algorithms. It’s a layer based approach to compute the desired output via calculating and predicting weight variations in the hidden neurons. The hidden neuron functions are dynamic in accordance to fatal errors of the Neural Network. These functions are identical one with another to produce desired results. The number of iterations to achieve the desired result differs based on sigmoid function and weight values at the hidden layers. As per the Neural Network architecture, the desired result depends on the hidden layer neurons cumulative functional values and its variation with actual output. These weight values highly influence the desired results and determine the iterations. Therefore, hidden neuronsas well as its variations are determining the number of iterations. This study is aimed to evaluate the co-relation between hidden neurons and iterations, variation weights of hidden neuron and overall performance of Neural Network to produce the desired output. The experimental results exhibit the relationship between incremental or decremented nodes of hidden neurons and its impact on number of iterations to achieve the desired results.