Optimization algorithm solving Lagrangian multipliers is the key of training SVM,determining the perfor-mance of SVM ,affecting practical applications of SVM in various fields widely. Some kinds of optimization algori...Optimization algorithm solving Lagrangian multipliers is the key of training SVM,determining the perfor-mance of SVM ,affecting practical applications of SVM in various fields widely. Some kinds of optimization algorithmsin SVM of overseas are introduced. We classify the optimization algorithms into two kinds: 1. the algorithms based onOsuna's decomposition strategy; 2. The iterative algorithms based on the changes of SVM formulation proposed byO. L. Mangasarian. We also analyze the characteristics of various optimization algorithms in SVM ,and predicting thetrend of research on optimization algorithm in SVM.展开更多
基于结构风险最小的最小二乘支持向量机(least squares support vector machine,LSSVM)为标准支持向量机(SVM)的约简,训练简易,性能良好。其模型精度受超参数影响,常规的网络搜索法很难搜得最佳超参数。在快速留一法的基础上,以全样本...基于结构风险最小的最小二乘支持向量机(least squares support vector machine,LSSVM)为标准支持向量机(SVM)的约简,训练简易,性能良好。其模型精度受超参数影响,常规的网络搜索法很难搜得最佳超参数。在快速留一法的基础上,以全样本留一预测误差平方和最小化为目标,导出基于梯度的最优化算法,用以优选为LSSVM超参数,进而构建G-LSSVM模型。以柠檬酸发酵过程为算例对G-LSSVM进行检验,结果表明G-LSSVM的超参数选取耗时少,模型稳定性良好,且拟合和预报性能都优于标准SVM和神经网络。有望适用于机理不明、高度非线性、小样本的化工过程建模。展开更多
文摘Optimization algorithm solving Lagrangian multipliers is the key of training SVM,determining the perfor-mance of SVM ,affecting practical applications of SVM in various fields widely. Some kinds of optimization algorithmsin SVM of overseas are introduced. We classify the optimization algorithms into two kinds: 1. the algorithms based onOsuna's decomposition strategy; 2. The iterative algorithms based on the changes of SVM formulation proposed byO. L. Mangasarian. We also analyze the characteristics of various optimization algorithms in SVM ,and predicting thetrend of research on optimization algorithm in SVM.
文摘基于结构风险最小的最小二乘支持向量机(least squares support vector machine,LSSVM)为标准支持向量机(SVM)的约简,训练简易,性能良好。其模型精度受超参数影响,常规的网络搜索法很难搜得最佳超参数。在快速留一法的基础上,以全样本留一预测误差平方和最小化为目标,导出基于梯度的最优化算法,用以优选为LSSVM超参数,进而构建G-LSSVM模型。以柠檬酸发酵过程为算例对G-LSSVM进行检验,结果表明G-LSSVM的超参数选取耗时少,模型稳定性良好,且拟合和预报性能都优于标准SVM和神经网络。有望适用于机理不明、高度非线性、小样本的化工过程建模。
