遗传前馈神经网络在函数逼近中的应用

人工神经网络具有高计算能力、泛化能力和非线性映射等特点，被成功应用于众多领域，但缺乏用于确定其网络拓扑结构、激活函数和训练方法的规则。该文提出利用遗传算法优化前馈神经网络的方法，将网络结构、激活函数和训练方法等编码作为个体，发现最优或次优解，针对特定问题设计较理想的前馈神经网络。介绍遗传算法的具体步骤，对非线性函数逼近进行实验，结果表明优化后前馈神经网络的性能优于由经验确定的前馈神经网络，验证了本文方法的有效性。

Application of Genetic Feedforward Neural Network in Function Approximation

Artificial neural network is successfully applied to solve actual problems in many areas because of its excellent computation ability, universality and nonlinear mapping. There is not a guided formula to specify the network structure, activation function and training method. This paper presents a method to optimize the feedforward neural network by Genetic Algorithm(GA), in which the network structure, activation function and training method are encoded as an individual. With optimum solution founded by GA, the feedforward neural network is satisfied. Steps of GA and an example of nonlinear function approximation are given. The experimental results of nonlinear function approach show that the performance of optimized network is better than that of experiential network and identifies validity of the method.