基于M-estimator与可变遗忘因子的在线贯序超限学习机  

Online Sequential Extreme Learning Machine Based on M-estimator and Variable Forgetting Factor

作　　者：郭威 徐涛[2] 于建江[1] 汤克明 GUO Wei;XU Tao;YU Jianjiang;TANG Keming(College of Information Engineering, Yancheng Teachers University, Yancheng 224002, China;College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China)

机构地区：[1]盐城师范学院信息工程学院,盐城224002 [2]南京航空航天大学计算机科学与技术学院,南京210016

出　　处：《电子与信息学报》2018年第6期1360-1367,共8页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(61603326;61379064;61273106)~~

摘　　要：该文针对时变离群值环境下的在线学习问题,提出一种基于M-estimator与可变遗忘因子的在线贯序超限学习机算法(VFF-M-OSELM)。VFF-M-OSELM以在线贯序超限学习机模型为基础,通过引入一种更加鲁棒的M-estimator代价函数来替代传统的最小二乘代价函数,以提高模型对于离群值的在线处理能力和鲁棒性。同时VFF-M-OSELM通过融合使用一种新的可变遗忘因子方法进一步增强了其在时变环境下的动态跟踪能力和自适应性。仿真实例验证了所提算法的有效性。To solve the online learning problem under the scenario of time-varying and containing outliers,this paper proposes an M-estimator and Variable Forgetting Factor based Online Sequential Extreme Learning Machine（VFF-M-OSELM）.The VFF-M-OSELM is developed from the online sequential extreme learning machine algorithm and retains the same excellent sequential learning ability as it,it replaces the conventional Least-Squares（LS） cost function with a robust M-estimator based cost function to enhance the robustness of the learning model to outliers.Meanwhile,a new variable forgetting factor method is designed and incorporated in the VFF-MOSELM to enhance further the dynamic tracking ability and adaptivity of the algorithm to time-varying system.The simulation results verify the effectiveness of the proposed algorithm.

关 键 词：在线贯序超限学习机 M-估计 可变遗忘因子 鲁棒性 自适应性 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]