GA-BP嵌套算法的理论及应用

The Theory and Application of GA-BP-NESTING Algorithm

分析了BP算法、遗传算法以及GA-BP-APARTING算法的特点,提出了GA-BP-NESTING算法.在人工神经网络的在线学习和离线学习方式下,分别对BP算法、GA算法、GA-BP-APARTING算法和GA-BP-NESTING算法进行了比较研究,研究发现:第一,网络初始权值的赋值对人工神经网络训练影响很大;第二,离线学习方式下GA-BP-NESTING算法效果最佳.

This paper analyses characters of BP algorithm, GA algorithm and GA-BP- APARTING algorithm. GA-BP-NESTING algorithm is proposed in this paper. Under the mode of on-line learning, off-line learning of neural network, using three algorithms to train the network respectively： Backpropagation （BP） algorithm,Genetic Algorithm （GA）, GA-BP- APARTING algorithm and GA-BP-NESTING algorithm. In the end of the paper, simulating a case under the two modes does a comparative study of the three algorithms. We can obtain the conclusion that initializing weights of networks is vital of training ANN and the GA-BP- NESTING algorithm under the mode of off-line learning is the best.