进化神经网络的编码方法与应用

Encoding Methods and Application of Evolutionary Neural Networks

进化神经网络将进化算法与人工神经网络进行了有机结合,进化算法的参与使神经网络系统在进化发育过程中可自适应的进行网络结构与连接权值的调整,改善了神经网络在模拟仿真过程中自主智能化不足的缺陷,提高了神经网络系统的生物真实性。随着研究的深入,大量不同类型的进化神经网络相继出现,根据基因编码方式的不同,可将进化神经网络分为直接编码型和间接编码型两类。本文对神经网络中基因的编码方式进行了阐述分析,最后总结了间接编码方法的应用领域。

Evolutionary neural networks are the combination of evolutionary algorithms and artificial neural net-works. The participation of evolutionary algorithm can supply self-adaptive adjustment of network structure and con-nection weights in the network’s evolution and development. It greatly improved the lack of autonomous intelligence in the process of simulation and made neural networks more biological authenticity. With the deepening of research, a large number of different types of the evolutionary neural networks have emerged. According to the different coding strategies of gene, evolutionary neural networks can be divided into two categories--direct coding and indirect coding. This paper expounded the coding strategies of gene and then summarized the application fields of the indirect coding strategies in the evolutionary neural networks.