基于矩阵分解和数据流的推荐算法
首发时间:2017-12-04
摘要:协同过滤算法是推荐算法中应用最广泛的算法之一,目前基于矩阵分解的协同过滤算法往往是针对静态数据的,缺乏对动态数据的处理能力,导致模型更新周期长、推荐时效性差的问题。针对目前在分布式平台中采用参数服务器控制模型训练过程中存在的滞后梯度和掉队者问题,本文提出一种使用迭代数据流和对等参数交换网络来代替参数服务器的方法,并在经典的MovieLens-1m数据集上进行了实验。实验结果表明,算法能够降低近一半的模型训练通讯开销,同时提高系统的推荐时效性。
For information in English, please click here
Recommendation Algorithm based on Matrix Factorization and Data Stream
Abstract:Collaborative filtering algorithm is one of the most widely used methods in recommendation algorithm. However, recent studies based on matrix factorization focus on static data and lack the adaptability to dynamic data, resulting in long updating period of model and poor recommendation effectiveness. Aiming at the straggler and delayed-gradient problems in using parameter server to control model training in distributed platform, a new method of using peer-to-peer parameter exchange network is proposed and experiments on Movielens-1m in this work. Experimental results show that the algorithm can reduce the communication cost by half, while improving the systems\'s effectiveness of recommendation.
Keywords: matrix factorization stream computing peer-to-peer network recommendation system
引用
No.****
同行评议
共计0人参与
勘误表
基于矩阵分解和数据流的推荐算法
评论
全部评论0/1000