A new approach to the evolution of neural networks is presented. Alinear chromosome combined with a grid-based representation of thenetwork, and a new crossover operator, allow the evolution of thearchitecture and the weights simultaneously. In the approach there is noneed for a separate weight optimization procedure and networks with morethan one type of activation function can be evolved. A pruning strategyis also introduced, which leads to the generation of solutions withvarying degrees of complexity. Results of the application of the methodto several binary classification problems are reported
展开▼