摘要
针对支持向量机对大样本学习占用内存多、训练速度慢等不足,本文归纳总结出分治、约减训练集、增量学习、并行化等四种解决策略。四种策略基于两个改进方向:其它算法结合、改变支持向量机算法结构,最终目的是减少支持向量机训练占用内存,提高训练速度。
According to the shortages of Support vector machine in occupying much memory and slowly training speed,this paper summarizes four sorts of strategy： divide-and-conquer, reduced sets, incremental learning, parallel algorithm,These strategy are based on two methods： combining other algorithms and changing the SVM algorithm structure to improve the SVM training speed.
出处
《微计算机信息》
2012年第4期22-23,56,共3页
Control & Automation
基金
基金申请人:奉国和
项目名称:自动文本分类技术研究
基金颁发部门:全国哲学社会科学规划办公室(08CTQ003)
关键词
支持向量机
大样本训练
分类
学习策略
Support Vector Machine
Large Scale Samples Training
Categorization
Learning Strategy
