【摘要】 针对Suyken等人提出的最小二乘支持向量机的共轭梯度法在输入样本的个数较大时,需要求解高阶线性方程组这一缺陷,提出了一种新算法。该算法利用分块矩阵的思想将该高阶线性方程组系数矩阵降阶,为了提高收敛速度,克服数值的不稳定性,采用条件预优共轭梯度法求解低阶的线性方程组。通过仿真试验证明用本文方法训练最小二乘支持向量机比共轭梯度法的训练速度提高了将近一倍。更多还原

【Abstract】 In allusion to the shortcoming of a conjugate gradient method for training least squares SVM proposed by Suykens etc,a new algorithm is presented.Matrix partitioning is applied to reduce the system of equations.In order to improve the rate of convergence and overcome instability of numerical value,a preconditioning conjugate gradient method is presented for solving the reduced system of linear equations.Numerical results show that our approach is about twice as efficient as the conjugate gradient method for least squares SVM.更多还原