论文部分内容阅读
一般的支持向量分类机需要求解二次规划问题,最小二乘支持向量机只需求解一个线性方程组,但其缺乏稀疏性.为了改进最小二乘支持向量分类机,本文结合中心距离比值及增量学习的思想提出一种基于预选、筛选支持向量的稀疏最小二乘支持向量机.该方法既能弥补最小二乘向量机的稀疏性,减少计算机的存储量和计算量,加快最小二乘支持向量机的训练速度和决策速度,又能对非均衡训练数据造成的分类面的偏移进行纠正,还不影响最小二乘支持向量机的分类能力.3组实验结果也证实了这一点.
General support vector classification machine needs to solve the quadratic programming problem, the least square support vector machine only needs to solve a linear system, but it lacks the sparsity.In order to improve the least square support vector classification machine, This paper proposes a sparse least squares support vector machine based on preselection and screening support vector which can not only make up the sparsity of least square vector machine but also reduce the amount of storage and computation of the computer and accelerate the least squares support The training speed and decision-making speed of vector machine can correct the offset of classification surface caused by non-equilibrium training data and does not affect the classification ability of least-squares support vector machine.The experimental results also confirm this.