摘要
针对传统BP算法训练速度慢、易陷入局部最优等缺点,该文提出了一种采用免疫遗传算法设计前向神经网络的方法。为解决神经网络权值随机初始化带来的问题,介绍了一种基于免疫的多样性模拟退火法(SAND算法)来进行神经网络权值初始化。仿真结果表明,该算法比混合遗传算法有更高的性能。
Aimed at the shortcoming of the traditional BP algorithm, such as the slow training speed, easy to be trapped into the local optimums. etc, a method to design the multi-layer feed-forward neural network based on immune genetic algorithm is proposed. In order to solve the problem of the random initial weights, the strategy of simulated annealing for diversity based on immunity is used to initialize the weight vectors. The simulation results show that the proposed algorithm displays a better performance than the hybrid genetic algorithm does.
出处
《计算机工程》
EI
CAS
CSCD
北大核心
2006年第15期179-180,183,共3页
Computer Engineering
关键词
免疫遗传算法
多样性模拟退火法
遗传算法
Immune genetic algorithm
Simulated annealing for diversity algorithm
Genetic algorithm
