基于相关向量机的非线性动态系统辨识  

Nonlinear Dynamic System Identification Based on Relevance Vector Machine

作　　者：朱世增[1] 党选举[1] 

机构地区：[1]桂林电子科技大学智能系统与工业控制研究室,广西桂林541004

出　　处：《计算机仿真》2008年第6期103-107,共5页Computer Simulation

基　　金：国家自然科学基金项目(60572055);广西科技创新能力与条件建设项目(桂科能063006-5G-1);"广西高校百名中青年学科带头人资助计划"资助项目(桂教人[2006]77号)

摘　　要：基于具有核函数不用满足Mercer条件、相关向量自动确定及核函数少特点的稀疏贝叶斯的相关向量机核学习方法,提出了平滑先验条件约束的相关向量机的学习方法,采用稀疏贝叶斯模型的最大边缘似然算法加快了求解相关向量机的向量,并采取交叉验证法确定其核参数提高了相关向量机辨识的泛化性。该方法避免了支持向量机的非线性系统辨识的模型结构难于确定的问题,与支持向量机辨识方法相比较,辨识的模型结构更简洁。仿真表明,该方法应用于非线性动态系统的辨识,具有良好的效果。Based on Relevance Vector Machine, Sparse Bayesian, a kind of kernel method, which has the advantages such as its kernel functions without the restriction of Mercer condition, the relevance vectors automatically determinated, and smaller kernel functions, the smoothness priors restriction on Relevance Vector Machine （RVM） is suggested. The algorithm＂fast marginal likelihood maximization for sparse Bayesian models＂is applied to solve rele- vance vectors effectively, and improving the generalization of identification is also improved. The Cross Validation is adopted to determine the kernel parameter. The suggested method avoids the problem of difficultly determining the model structure by the Support Vector Machine for nonlinear dynamic system identification. Comparing with the Support Vector Machine, a quite simpler model structure is obtained. The result shows that the Relevance Vector Machine applied for nonlinear dynamic system identification achieves a better performance.

关 键 词：稀疏贝叶斯 相关向量机 支持向量机 非线性系统辨识 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]