论文部分内容阅读
本文将遗传算法升到对神经网络权重的优化，不必对B－P网络的参数进行调试，降低了其学习过程的计算量，也避免原算法陷入局部最优的可能性。得到参数影响遗传优化效果的规律，而且遗传算法在本问题中具有很强的鲁棒性。
In this paper, the genetic algorithm is raised to the optimization of neural network weights. It is not necessary to debug parameters of B-P network, which reduces the computational cost of learning process and avoids the possibility of the original algorithm falling into local optimum. The parameters that influence the effect of genetic optimization are obtained, and the genetic algorithm has strong robustness in this problem.