Adaptive strategy for neural network synthesis constant estimation

Abstract

Neural Network Synthesis is a new innovative method for an artificial neural network learning and structural optimization. It is based on two other already very successful algorithms: Analytic Programming and Self-Organizing Migration Algorithm (SOMA). The method already recorded several theoretical as well as industrial application to prove itself as a useful tool of modelling and simulation. This paper explores promising possibility to farther improve the method by application of an adaptive strategy for SOMA. The new idea of adaptive strategy is explained here and tested on a theoretical experimental case for the first time. Obtained data are statistically evaluated and ability of adaptive strategy to improve neural network synthesis is proved in conclusion.