现代电子技术2018，Vol.41Issue(3)：146-150,156,6.DOI:10.16652/j.issn.1004-373x.2018.03.034
改进学习率的一种高效SVD++算法
An efficient SVD++ algorithm for learning rate improvement
摘要
Abstract
The core of the recommendation system includes the recommendation algorithm and big data. The efficient com-putation of the recommendation algorithm is the essential requirement to realize the real-time human-machine interaction. Among various recommendation algorithms,the SVD++ algorithm is widely used because its special advantages. However,in the envi-ronment of big data,the SVD++ recommendation algorithm has the prominent problem of low computing efficiency,and is diffi-cult to satisfy the requirement of real-time human-machine interaction. In order to solve this problem,a new method to improve the computation efficiency of the SVD++ recommendation algorithm is proposed,whose kernel is to optimize the indicator of the target function with new learning rate function. In combination with the changing rate characteristics of the exponential function and linear function,the learning rate function has the characteristics of high initial value,low medium-term descend speed and later value,and slow change toward. The effectiveness of the method was verified with simulation experiment.关键词
推荐系统/推荐算法/大数据/SVD++/计算效率/学习率Key words
recommendation system/recommendation algorithm/big data/SVD++/computation efficiency/learning rate分类
信息技术与安全科学引用本文复制引用
王燕,李凤莲,张雪英,田玉楚..改进学习率的一种高效SVD++算法[J].现代电子技术,2018,41(3):146-150,156,6.基金项目
山西省国际合作项目(2015081007) （2015081007）
2016年太原理工大学教改项目(24) （24）
山西省优秀人才科技创新项目(201605D211021) Project Supported by Shanxi International Cooperation Project(2015081007),Teaching Reform Project of Taiyuan University of Technology in 2016 (24),Science and Technology Innovation Project of Excellent Talents in Shanxi(201605D211021) （201605D211021）
