摘要
首先介绍了遗传算法和标准BP算法及其改进形式 ,并指出遗传算法和BP算法各自的优缺点 ,然后着重讨论了如何采用遗传算法和梯度BP算法相结合的方法来训练前馈神经网络 ,从而提高神经网络的收敛速度和收敛质量。最后进行了仿真实验 ,结果表明 ,该方法不仅收敛速度快 ,而且易达到最优解 。
Genetic algorithm, standard back propagation and its improved methods are first introduced. Their merits and defects are also presented, then the combination of genetic algorithm and back propagation algorithm for training neural network is describled in detail. It can improve the search efficiency and realize global optimization. The simulation result shows this method has high convergent speed, easily-oriented global optimization and practical value.
出处
《计算机仿真》
CSCD
2005年第2期150-151,166,共3页
Computer Simulation
关键词
前馈神经网络
网络训练
反向传播算法
遗传算法
Feed forward neural network
Network training
Back propagation algorithm
Genetic algorithm
