MEA for Designing Neural Network Weights and Structure Optimization

For artificial neural network application, its weights and structure optimization design is a key problem. The mind evolutionary algorithm (MEA) is a new evolutionary algorithm which simulates the process of human mind evolution, it has the powerful ability to find global optimum, and it also has much superiority for resolving the problem of numerical and non-numerical optimization. In this paper, a new typical MEA is presented based on the foundational MEA framework to optimize the neural network structure and weights, in which effective similar taxis and dissimilation operators of structure optimization are designed. Through similar taxis operators, the local optimum is found, then exceeding the restriction of local range through dissimilation operators, the global optimum is acquire in global solution space. Finally, simulation results show the effectiveness and correctness of the method.