摘要
双胞参数间隔支持向量机在模式识别上拥有优秀的分类能力。然而,原始的TPMSVM模型仅针对二分类问题,并不能处理回归学习任务。为此,文章提出了一种新的正则双胞参数间隔支持向量回归机模型(RTPMSVR)。RTPMSVR模型的最终回归输出函数是间接通过寻找一对最优的非平行上界和下界参数间隔函数来构建的。通过继承TPMSVM模型的损失函数,RTPMSVR模型分别为上界和下界的参数间隔函数构建二次规划优化模型。此外,为提高模型的泛化能力,引入额外的正则项,进而保障模型解的唯一性。根据对偶理论,构建模型求解的最优KKT条件,并将RTPMSVR模型的原问题转换为对偶问题来求解。最后,通过对比实验,验证了该方法的有效性。
The twin parametric-margin support vector machine(TPMSVM)has excellent classification ability in pattern recognition.However,the original TPMSVM model is only for binary classification problems,and cannot handle regression learning tasks.To this end,a new regular twin parametric-margin support vector machine model(RTPMSVR)is proposed.The final regression output function of the RTPMSVR model is constructed indirectly by finding a pair of optimal non-parallel upper and lower parameter interval functions.By inheriting the loss function of the TPMSVM model,the RTPMSVR model constructs a quadratic programming optimization model for the parameter interval function of the upper bound and the lower bound respectively.n addition,in order to improve the generalization ability of the model,additional regularization terms are introduced to ensure the uniqueness of the model solution.According to the duality theory,the optimal KKT condition for model solving is constructed,and the original problem of the RTPMSVR model is converted into a dual problem to solve.Finally,through comparative experiments,the effectiveness of the method is verified.
作者
叶玲节
杨云露
冯昊
YE Lingjie;YANG Yunlu;FENG Hao(Shenyang Institute of Computing Technology,Chinese Academy of Sciences,Shenyang 110004,China)
出处
《电工技术》
2020年第24期71-73,77,共4页
Electric Engineering
