基于动态BP算法的非线性滞后系统辨识

鉴于很多实际问题都可以转化到多示例框架下求解,多示例学习越来越受到机器学习领域内学者们的关注.提出了一个基于Logistic回归模型的多示例学习算法.首先定义了一个新的似然函数来表示每个包的标签与其示例的隐含标签之间的关系,然后利用凝聚函数把该似然函数转化为一个光滑的凹函数,从而使问题可以用常用的无约束优化方法快速求解.在一些标准数据集和一个文本分类问题上的实验结果表明,所提算法要优于其他常用多示例学习算法.

Nonlinear time delay systems identification based on dynamic BP algorithm

The modeling of nonlinear dynamic system is an important problem in the domain of automatic control. For the problem, especially for the nonlinear time delay system, a novel dynamic BP (back propagation) neural network algorithm with adaptive time delay parameters is proposed. Based on the traditional multilayer perceptron neural network structure, the adaptive time delay parameters are employed in the first hidden layer and output layer. The gradient descent method helps to realize the parameter adjusting and time delay estimation. The simulation results show that the proposed method can not only identify the nonlinear system effectively, but also estimate time delay exactly.