.png){kind=icon}
返回
Pattern on demand in transactional distributed databases
.png){kind=icon}
OA
AI
DOI：10.1016/j.is.2021.101908.png)
.png){kind=icon}
.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Many applications rely on distributed databases like sensor networks or the Semantic Web. However, only few methods exist to extract patterns without centralizing the data by following the exhaustive extraction paradigm. Their principle is to extract a unique large collection of frequent patterns that will be used for all downstream applications. Unfortunately, the communication of this large collection from the different nodes is often more expensive than the database centralization. Furthermore, this rigid principle is not suited to modern data analysis where data and analyst needs change daily. It is both too expensive to repeat the exhaustive extraction for each change and it is not possible to build the ideal collection of patterns to meet all the needs. To circumvent this difficulty, this paper revisits the problem of pattern mining in distributed databases by adopting the Pattern-On-Demand paradigm. This principle consists in instantly extracting the patterns at the moment when the analyst needs them. Specifically, we propose a new pattern sampling algorithm, named DDSAMPLING, that randomly draws a pattern from a transactional distributed database with a probability proportional to its interest. We demonstrate the soundness of DDSAMPLING and analyze its time complexity. Finally, experiments on benchmark datasets highlight its low communication cost and its robustness against network and node failures. We also illustrate its interest on real-world data from the Semantic Web for detecting outlier entities in DBpedia and Wikidata. In addition, our output space sampling method is more parsimonious in terms of communication cost than a baseline relying on input space sampling. (C) 2021 Elsevier Ltd. All rights reserved.
Keyword:
Pattern mining
Knowledge base
Pattern on Demand
Pattern sampling
Outlier detection
AI总结
对已上传原文的论文进行重点信息的提取，主要内容包括：简要概述、研究摘要、背景介绍、关键亮点、图文解析、展望与总结。
期刊
IF:
3.9
论文数:
2.8K
被引数:
1.8K
机构
引用论文
The role of crosslinguistic differences in second language anticipatory processing: An event-related potentials study跨语言差异在第二语言预期处理中的作用：一项事件相关电位研究OAAI
OAAI分享收藏Polymerization, grafting and adsorption in the presence of inorganic substrates: Thermal polymerization of styrene with untreated and γ-irradiated silica gel as a case study.png)
Polymer
PolymerIF0
PREAI
分享收藏
查看更多内容