利用进化规划和逐步二次规划实现前馈神经网络的结构优化

Structure Optimization for Feed-forward Neural Networks Based on Evolutionary Programming and Sequential Quadratic Programming

用进化规划与逐步二次规划来实现前馈神经网络的结构优化问题 ,并提出了一个相应的学习算法 .针对进化规划与逐步二次规划各自的特点 ,进行了组合 ,使算法不仅具有随机全局搜索能力 ,而且还具有更好的全局收敛能力 ,并与环境有更强的自适应能力 .最后通过仿真和应用实验证实了算法的有效性.

In this paper, when evolutionary programming and sequential quadratic programming are applied to the structure optimization of feed-forward neural networks, a learning algorithm is proposed. The new algorithm retains the ability of stochastic global searching. It has better global convergence and very strong self-adaptive ability with environment. The efficiency of research work mentioned above has been shown by simulation and applications.