This paper presents a new approach to the evolution of neural networks. A linear chromosome combined with a grid-based representation of the network and a new crossover operator allow the evolution of the architecture and the weights simultaneously. There is no need for a separate weight optimization procedure and networks with more tham one type of activation function can be evolved. This paper describes the representation, the crossover operator, and reports on results of the application of the method to evolve a neural controller for the pole-balancing problem.
展开▼
