A new evolutionary optimization-method for designing reconfigurable neural networks

This paper introduces a new hybrid evolutionary algorithm suitable for designing evolving neural networks. The purpose is to search the best network configuration for solving particular problems. In the proposed framework, for instance, the authors deal with a prediction problem, starting from a data time series and leading to a fast converging process of network optimization. The adopted hybrid approach guarantees that updating rules of a specific population result in a more natural evolution of the following generation step