摘要
BP神经网络在目前的非线性系统中应用广泛,但是作为有导师的学习系统,BP神经网络必须要求提供相关的经验数据才能正常运行,这对一般系统来说是非常麻烦和不现实的。对此文章提出了一种基于神经网络集成的强化学习BP算法,通过强化学习体系来实现体统的自学习,通过网络集成来达到初始数据的预处理,提高系统的泛化能力,并在实际应用中取得较好的效果。
BP neural network has been used in nonlinear system controller widely.But as a supervised training algorithm,it requires experiential data to be trained.But in some system such data cannot be got.So this paper provides the optimization on a reinforcement leaming algorithm based on neural network ensemble. Reinforcement leaming is unsupervised and on-line.Neural network ensemble can significantly improve the generalization ability of leaming system. The method is tested and the expected results are obtained.
出处
《计算机工程与应用》
CSCD
北大核心
2006年第12期97-99,共3页
Computer Engineering and Applications
基金
燕山大学博士基金资助项目(编号:2004013)
关键词
神经网络集成
BP神经网络
强化学习
RBP模型
Neural Network ensemble,BP Neural Network,reinforcement learning, Reinforcement Baek-Propagation model
