A Pseudo-Parallelism Genetic Algorithm Framework to Optimization of Neural Networks

This paper present a new approach, combined pseudo- parallelism evolution technique based on sub-population competition with parent mutation mechanism, for automatic topology optimization of multilayer feedforward neural networks. It allows that two networks with different number of individuals can be crossed to a new valid "child" network. The calculation result of an example shows that PPGA is able to get the real-time information of population diversity during the process of evolution and has some improvements in both global converging velocity and searching precision.