2016 High Performance Graph Data Management and Processing Workshop (HPGDMP)最新文献

筛选

英文中文

An Efficient Parallel Trust-Based Recommendation Method on Multicores 一种基于信任的多核并行推荐方法

2016 High Performance Graph Data Management and Processing Workshop (HPGDMP) Pub Date : 2016-11-13 DOI: 10.1109/HPGDMP.2016.6

Huafeng Liu, L. Jing, Miaomiao Cheng

{"title":"An Efficient Parallel Trust-Based Recommendation Method on Multicores","authors":"Huafeng Liu, L. Jing, Miaomiao Cheng","doi":"10.1109/HPGDMP.2016.6","DOIUrl":"https://doi.org/10.1109/HPGDMP.2016.6","url":null,"abstract":"With the significant advances in online social networks which provide precious knowledge for personalized recommendation, it is necessary to design effective and efficient method to deal with such data. In this paper, we focus on the recommendation system to integrate the preference information and trust relations among users, following the trust-based recommendation model (TbRM) [1] which considers both the user's reputation and his/her nearest neighbors in the social network. Our main contribution is to present a fast algorithm to solve TbRM model with the aid of graph vertex programming (GVP) in parallel. The idea is to represent the preference and trust information as a graph with both user and item vertices, and its edges contain the preference and trust information. In this case, the recommendation problem becomes a graph analysis procedure which iteratively updates the vertices' states in parallel with the aid of predefined vertex state function and edge information. A series of experiments on four real-world recommendation datasets (Ciao, Epinions, Douban and Flixster) have shown that the graph parallel operations obviously speed up the recommendation procedure, e.g., GVP performs 1.1–2.3× faster than the existing popular distributed stochastic gradient descent algorithm with different number of cores.","PeriodicalId":189262,"journal":{"name":"2016 High Performance Graph Data Management and Processing Workshop (HPGDMP)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115398454","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Power-Efficient Breadth-First Search with DRAM Row Buffer Locality-Aware Address Mapping 具有DRAM行缓冲区位置感知地址映射的高能效宽度优先搜索

2016 High Performance Graph Data Management and Processing Workshop (HPGDMP) Pub Date : 2016-11-13 DOI: 10.1109/HPGDMP.2016.7

S. Imamura, Yuichiro Yasui, Koji Inoue, Takatsugu Ono, Hiroshi Sasaki, K. Fujisawa

{"title":"Power-Efficient Breadth-First Search with DRAM Row Buffer Locality-Aware Address Mapping","authors":"S. Imamura, Yuichiro Yasui, Koji Inoue, Takatsugu Ono, Hiroshi Sasaki, K. Fujisawa","doi":"10.1109/HPGDMP.2016.7","DOIUrl":"https://doi.org/10.1109/HPGDMP.2016.7","url":null,"abstract":"Graph analysis applications have been widely used in real services such as road-traffic analysis and social network services. Breadth-first search (BFS) is one of the most representative algorithms for such applications; therefore, many researchers have tuned it to maximize performance. On the other hand, owing to the strict power constraints of modern HPC systems, it is necessary to improve power efficiency (i.e., performance per watt) when executing BFS. In this work, we focus on the power efficiency of DRAM and investigate the memory access pattern of a state-of-the-art BFS implementation using a cycle-accurate processor simulator. The results reveal that the conventional address mapping schemes of modern memory controllers do not efficiently exploit row buffers in DRAM. Thus, we propose a new scheme called per-row channel interleaving and improve the DRAM power efficiency by 30.3% compared to a conventional scheme for a certain simulator setting. Moreover, we demonstrate that this proposed scheme is effective for various configurations of memory controllers.","PeriodicalId":189262,"journal":{"name":"2016 High Performance Graph Data Management and Processing Workshop (HPGDMP)","volume":"107 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123626827","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 3

Distributed Multithreaded Breadth-First Search on Large Graphs Using DXGraph 基于DXGraph的大型图的分布式多线程广度优先搜索

2016 High Performance Graph Data Management and Processing Workshop (HPGDMP) Pub Date : 2016-11-13 DOI: 10.1109/HPGDMP.2016.5

Stefan Nothaas, Kevin Beineke, M. Schöttner

引用次数: 3