论文部分内容阅读
针对核函数选择对最小二乘支持向量机回归模型泛化性的影响,提出一种新的基于l_p-范数约束的最小二乘支持向量机多核学习算法.该算法提供了两种求解方法,均通过两重循环进行求解,外循环用于更新核函数的权值,内循环用于求解最小二乘支持向量机的拉格朗日乘数,充分利用该多核学习算法,有效提高了最小二乘支持向量机的泛化能力,而且对惩罚参数的选择具有较强的鲁棒性.基于单变量和多变量函数的仿真实验表明了所提出算法的有效性.
Aiming at the effect of kernel function selection on the generalization of least-squares support vector machine regression model, a new multi-core learning algorithm based on l_p-norm constraint is proposed. The algorithm provides two solutions, Are solved by two loops. The outer loop is used to update the weight of the kernel function. The inner loop is used to solve the Lagrange multipliers of the least square support vector machine, and the multi-core learning algorithm is fully utilized to effectively increase the minimum two Multiplication by support vector machine, and has strong robustness to the selection of penalty parameters.The simulation based on univariate and multivariable functions shows the effectiveness of the proposed algorithm.