基于牛顿梯度优化的弹性多核学习

摘要

已有稀疏多核学习(MKL)模型在产生核函数权重稀疏解时容易导致信息丢失且泛化能力差,且基于梯度下降法的MKL在接近最优解时收敛速度慢.建立了基于支持向量机(SVM)的弹性多核学习(EMKL)模型并给出了一种基于牛顿梯度优化的EMKL(NO-EMKL).模型在MKL的目标函数中引入弹性项,并设计了基于二阶牛顿梯度下降法的优化算法.实验结果表明:算法不仅具有更好的分类精度,还具有较快的收敛速度.%Due to the present sparse multiple kernel learning (MKL )model may lead to discard useful information,and it's generalization ability is degenerated when produces the sparse weight of kernel function. And the MKL method which is based on gradient descent method has slow convergence speed when close to the optimal solution. Aiming at these problems,establish elastic MKL (EMKL ) model which is based on support vector machine(SVM)and propose an EMKL method based on Newton gradient optimization(NO-EMKL). In this model, the elastic item is brought in and the second order Newton gradient descent method is used. The experimental results show that the algorithm not only has better classification precision,but also has faster convergence speed.