基于遗传程序设计优化神经网络激励函数问题研究

ON NEURAL NETWORK'S ACTIVATION FUNCTION OPTIMIZED BY GENETIC PROGRAMMING

近年来数据建模问题在数据挖掘、预测等领域得到广泛应用;神经网络由于其固有的许多优点,已成为解决很多问题的得力工具,对更深入探索非线性等现象起到了重大作用。如何根据问题建立一个好的神经网络是摆在我们面前最棘手的问题。利用遗传程序设计对神经网络激励函数进行优化,实验验证,通过此方法能更快学习到更适合问题解的神经网络。

In recent years, data modelling is widely applied in areas of data mining and forecasting process, etc. Neural network becomes an important tool in solving many problems due to its inherent good particularities and plays a significant role in further exploring the phenomenon such as nonlinear, etc. How to construct an excellent artificial neural network based on actual problems is a knotty problem for us. In the paper,Genetic Programming is used to optimize the activation function of neural network. In this way a neural network which is more suitable to problem solving can be learnt quicker,as proved by the experiment＇s result.