Evolutionary Optimization of Recurrent Neural Network Topologies Along with Weights Considering Diversity of Searching Population

Abstract

The generation alternation model employed by one of the best neuroevolution methods, NEAT, has a drawback; it proceeds without intensively searching neighbors of parent individuals and hence it often fails in identifying solutions with minimal structures. We propose an alternative for the model and show its effectiveness through the application to the double pole balancing without velocities problem.