基于平衡学习的CMAC神经网络非线性辨识算法

Nonlinear identification algorithm of the improved CMAC based on balanced learning

为提高小脑模型关节控制器(CMAC)神经网络在线学习的快速性和准确性,提出一种平衡学习的概念,并设计一种改进的CMAC学习算法.在常规的CMAC中,误差的校正值被平均地分配给所有激活存储单元,而不管这些存储单元的可信度;在改进的CMAC中,利用激活单元先前学习次数作为可信度,其误差校正值与激活单元先前学习次数的负k次方成比例.仿真结果表明,当k为一适当数值时,改进CMAC具有较快的学习速度和较高的精度,特别是在神经网络的初始学习阶段.

In order to improve online learning speed and accuracy of CMAC, an improved CMAC neural network model based on balanced learning concept is designed. In the conventional CMAC learning scheme, the correcting amounts of errors are equally distributed into all addressed hypercubes, regardless the credibility of those (hypercubes.) The proposed improved learning approach uses the learned times of the addressed hypercubes as the (credibility(confidence)) of the learned values. The correcting amounts of errors are proportional to the inverse of the k-th power of learned times. The method provides high learning speed when k is the optimal value in the early (learning) stage especially.