摘要
现有最小二乘支持向量机回归的训练和模型输出的计算需要较长的时间,不适合在线实时训练.对此,提出一种在线稀疏最小二乘支持向量机回归,其训练算法采用样本字典,减少了训练样本的计算量.训练样本采用序贯加入的方式,适合在线获取,并且该算法在理论上是收敛的.仿真结果表明,该算法具有较好的稀疏性和实时性,可进一步用于建模与实时控制等方面的研究.
Least square support vector machines regression without sparsity needs longer training time currently,and is not adapted to online real-time training.A better method of online sparse least square support vector machines regression(SVMR) is proposed.Less training time is needed by using sample dictionary.The online SVMR whose samples is available one by one by adding the samples sequentially is convergence theoretically.The simulation result shows that the algorithm has a better sparsity and real-time performance and could be applied in modeling and real-time control etc.
出处
《控制与决策》
EI
CSCD
北大核心
2007年第2期132-137,共6页
Control and Decision
基金
国家自然科学基金项目(60574083)
江苏省高校自然科学研究项目(06KJB210049)
关键词
在线
稀疏
最小二乘支持向量机
Online
Sparsity
Least square support vector machines