Comprehensive evaluation of large-scale parallel matrix factorization algorithms

Xiaolong Liu, Ching Yu Ho, Sifei Zheng, Shyan Ming Yuan

研究成果: Conference contribution同行評審

摘要

Fast parallel stochastic gradient descent (FPSG) and non-locking stochastic multi-machine (NOMAD) are two representative matrix factorization algorithms for real-time accurate recommendation of large-scale recommender systems. To fairly ascertain the performance of FPSG and NOMAD matrix factorization algorithms, a comprehensive evaluation is presented in this paper. Streaming SIMD extensions (SSE) and different learning rate schedules are applied to evaluate the convergence performance of these two algorithms. The experimental results show that SSE can speed up the convergence performance of original NOMAD algorithm in Netflix dataset. In addition, no matter what learning rate schedules are used, FPSG algorithm has relatively good convergence results than NOMAD under the same parameters and experimental environments.

原文English
主出版物標題Proceedings - 21st IEEE International Conference on High Performance Computing and Communications, 17th IEEE International Conference on Smart City and 5th IEEE International Conference on Data Science and Systems, HPCC/SmartCity/DSS 2019
編輯Zheng Xiao, Laurence T. Yang, Pavan Balaji, Tao Li, Keqin Li, Albert Zomaya
發行者Institute of Electrical and Electronics Engineers Inc.
頁面811-818
頁數8
ISBN(電子)9781728120584
DOIs
出版狀態Published - 8月 2019
事件21st IEEE International Conference on High Performance Computing and Communications, 17th IEEE International Conference on Smart City and 5th IEEE International Conference on Data Science and Systems, HPCC/SmartCity/DSS 2019 - Zhangjiajie, China
持續時間: 10 8月 201912 8月 2019

出版系列

名字Proceedings - 21st IEEE International Conference on High Performance Computing and Communications, 17th IEEE International Conference on Smart City and 5th IEEE International Conference on Data Science and Systems, HPCC/SmartCity/DSS 2019

Conference

Conference21st IEEE International Conference on High Performance Computing and Communications, 17th IEEE International Conference on Smart City and 5th IEEE International Conference on Data Science and Systems, HPCC/SmartCity/DSS 2019
國家/地區China
城市Zhangjiajie
期間10/08/1912/08/19

指紋

深入研究「Comprehensive evaluation of large-scale parallel matrix factorization algorithms」主題。共同形成了獨特的指紋。

引用此