论文部分内容阅读
提出了一种用于回归估计的最小二乘广义支持向量机.这种最小二乘广义支持向量机的核函数同标准的支持向量机相比没有或者只有很少的限制.将这种用于回归估计的最小二乘广义支持向量机表示成标准的二次规划(QP)问题,采用基于矩阵分裂的超松弛法同投影梯度法相结合的算法来解这一QP问题.根据超松弛法的特点,这一算法可以处理大量数据的情形.
A least-squares generalized support vector machine for regression estimation is proposed.The kernel function of this least-squares generalized support vector machine has no or very few restrictions compared with the standard support vector machine.This method is applied to Regression estimation of least squares support vector machines expressed as a standard quadratic programming (QP) problem, the use of matrix-based over-relaxation method with the projection gradient method to solve this problem of QP. According to the characteristics of the over-relaxation method , This algorithm can handle the case of large amounts of data.