【摘要】 为了求解广义支持向量机(GSVM)的优化问题,将带有不等式约束的原始优化问题转化为无约束优化问题,由于此无约束优化问题的目标函数不光滑,所以引入一族多项式光滑函数进行逼近,实验中可以根据不同的精度要求选择不同的逼近函数。用BFGS算法求解。实验结果表明,该算法和已有的GSVM的求解算法相比,更快地获得了更高的测试精度,更适合大规模数据集的训练。因此给出的GSVM的求解算法是有效的。更多还原

【Abstract】 To solve the optimization problem of Generalized Support Vector Machine(GSVM),the primal optimization problem with inequality constraints is transformed into the unconstraint optimization problem,whose objective function is nonsmooth, therefore a series of polynomial smooth functions is introduced to approach the objective function.Different polynomial functions can be used according to the corresponding accuracy demand.The model is solved by the BFGS algorithm.Experimental results show,compared with the existing algorithms used for solving the optimization problem of GSVM,the proposed algorithm achieves higher testing accuracy more quickly and is useful for large-scale data.Therefore,the proposed algorithm is effective.更多还原