基于SVM增量学习的用户适应性研究  

Study of SVM-Based Incremental Learning for User Adaptation

作　　者：彭彬彬[1] 金翔宇[1] 徐晓刚[1] 孙正兴[2] 

机构地区：[1]南京大学软件新技术国家重点实验室计算机科学与技术系,南京210093 [2]南京大学软件新技术国家重点实验室,计算机科学与技术系南京210093

出　　处：《计算机科学》2003年第3期75-79,共5页Computer Science

基　　金：国家自然科学基金(编号:69903006)

摘　　要：User adaptation is a critical and important problem. For users' specialization, such as Handwriting, Voice,Drawing Styles, the system is hard to adapt to all users. SVM-based incremental learning can find the most basic fea-ture of different users and cast away the special user's character, so this method can adapt the different users withoutover fitting. In this paper, the repetitive learning strategy and other two incremental learning algorithms are presentedfor comparison. Based on theoretical analysis and experimental results, we draw the conclusion that SVM-based incre-mental learning can solve the user conflict problem.User adaptation is a critical and important problem. For users' specialization, such as Handwriting, Voice, Drawing Styles, the system is hard to adapt to all users. SVM-based incremental learning can find the most basic feature of different users and cast away the special user's character, so this method can adapt the different users without over fitting. In this paper, the repetitive learning strategy and other two incremental learning algorithms are presented for comparison. Based on theoretical analysis and experimental results, we draw the conclusion that SVM-based incremental learning can solve the user conflict problem.

关 键 词：SVM 增量学习 用户适应性 图形识别 人机交互 支持向量机 机器学习 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP181[自动化与计算机技术—计算机科学与技术]