郑州大学学报（理学版）2016，Vol.48Issue(3)：51-56,6.DOI:10.13705/j.issn/1671-6841.2016097
一种基于分类问题的光滑极限学习机
A Smooth Extreme Learning Machine for Classification
摘要
Abstract
Extreme learning machine ( ELM) had a high learning speed and a good generalization ablity. Smoothing strategy was an important technology for non-smooth problems. By combining a smoothing technique with ELM, a smooth ELM ( SELM) framework was proposed. Moreover, the Newton-Armijo al-gorithm was used to solve the SELM, and resulting algorithm converged globally and quadratically. The proposed SELM had less decision variables and better abitities to deal with nonlinear problems than the existing smooth support vector machine. Numerical experiments demonstrated that the speed of SELM was much faster than that of the existing ELM algorithms based on optimization theory. Compared with other popular support vector machines, the proposed SELM achieved better or similar generalization. The re-sults demonstrated the feasibility and effectiveness of the proposed algorithm.关键词
极限学习机/光滑化方法/Newton-Armijo算法/神经网络Key words
extreme learning machine( ELM)/smooth approach/Newton-Armijo algorithm/neural net-works分类
信息技术与安全科学引用本文复制引用
杨丽明,张思韫,任卓..一种基于分类问题的光滑极限学习机[J].郑州大学学报（理学版）,2016,48(3):51-56,6.基金项目
国家自然科学基金资助项目(11471010) （11471010）
