Evolving Artificial Neural Networks Using GA and Momentum

摘要

Neural network learning methods provide a robust approach to approximating real-valued, discrete-valued and vector-valued target functions. Artificial neural networks are among the most effective learning methods currently known for certain types of problems. But BP training algorithm is based on the error gradient descent mechanism that the weight inevitably fall into the local minimum points. Genetic Algorithms (GAs) is good at global searching, and search for precision appears to be partial capacity inadequate. So, in this paper, the genetic operators were carefully designed to optimize the neural network, avoiding premature convergence and permutation problems. And with the momentum to solve the slow convergence problem of BP algorithm. To evaluate the performance of the genetic algorithm-based neural network, BP neural network was also involved for a comparison purpose. The results indicated that Gas and with momentum were successful in evolving ANNs.