SYNTHESIS OF OPTIMAL ARTIFICIAL NEURAL NETWORKS BY MODIFIED GENETIC ALGORITHM
Read the full article
Principles of neural networks synthesis are considered, shortcomings of approaches are shown when choosing
the structure and scales adjustment. Networks optimization is possible by evolutionary algorithms application
with advantages becoming apparent with multi coherent and multilayered networks operation. Simulation
experiments were made on concrete test function and confirm efficiency of the developed algorithms.
Keywords: artificial neural network, genetic algorithm, cluster analysis, neuron, chromosome, back error propagation, Branins test function.