二元裂解算子交替方向乘子法的核极限学习机  

Kernel Extreme Learning Machine Based on Alternating Direction Multiplier Method of Binary Splitting Operator

作　　者：苏一丹[1] 续嘉 覃华[1] SU Yidan;XU Jia;QIN Hua(College of Computer and Electronic Information,Guangxi University,Nanning 530004,China)

机构地区：[1]广西大学计算机与电子信息学院,南宁530004

出　　处：《电子与信息学报》2021年第9期2586-2593,共8页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(62073291)。

摘　　要：凸优化形式的核极限学习机(KELM)具有较高的分类准确率,但用迭代法训练凸优化核极限学习机要较传统核极限学习机的解线性方程法花费更长时间。针对此问题,该文提出一种2元裂解算子交替方向乘子法(BSADMM-KELM)来提高凸优化核极限学习机的训练速度。首先引入2元裂解算子,将求核极限学习机最优解的过程分裂为两个中间算子的优化过程,再通过中间算子的迭代计算而得到原问题的最优解。在22个UCI数据集上所提算法的训练时间较有效集法平均快29倍,较内点法平均快4倍,分类精度亦优于传统的核极限学习机;在大规模数据集上该文算法的训练时间优于传统核极限学习机。The Kernel Extreme Learning Machine(KELM) with convex optimization form has higher classification accuracy, but it takes longer time to train kelm with iterative method than solving linear equation method of traditional kelm. To solve this problem, an Alternating Direction Multiplier Method(ADMM) of Binary Splitting(BSADMM-KELM) is proposed to improve the training speed of convex optimization kernel extreme learning machine. Firstly, the process of finding the optimal solution of the kernel extreme learning machine is split into two intermediate operators by introducing a binary splitting operator, and then the optimal solution of the original problem is obtained through the iterative calculation of the intermediate operators. On22 UCI datasets, the training time of the proposed algorithm is 29 times faster than that of the effective set method and 4 times faster than that of the interior point method. The classification accuracy of the proposed algorithm is also better than that of the traditional kernel extreme learning machine. On large-scale datasets,the training time of the proposed algorithm is better than that of the traditional kernel extreme learning machine.

关 键 词：核极限学习机 2次规划模型 2元裂解算子 交替方向乘子法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]