The Second Workshop on Management and mining Of UNcertain Data

 (MOUND 2010)
In conjunction with ICDE 2010

March 1, 2010,  Long Beach, California, USA

Speaker: Dan Olteanu (Univerity Lectuer, OUCL)

Title: A Toolbox of Query Evaluation Techniques for Probabilistic Databases

Abstract:

In this talk I will discuss the problem of query evaluation in probabilistic databases and survey some of the existing techniques proposed by the database community. Although this problem is subsumed by general probabilistic inference, two fundamental aspects of databases, that is, (i) the separation of (very large) data and (small and fixed) query, and (ii) the use of mature relational query engines, can lead to more scalable techniques. I survey both exact and approximate query evaluation techniques. In case of exact evaluation, I discuss syntactical restrictions of the language of conjunctive queries with inequalities, under which the queries become tractable (in general, the problem is #P-hard). Relational query plans extended with efficient aggregation operators can be effectively used to evaluate such tractable queries. In case of intractable queries, exact techniques decompose the data-query instance into a tractable subinstance, which is solved as before, and a (usually much smaller) intractable subinstance that can be solved using AI inference techniques. Alternatively, intractable queries can be evaluated using (deterministic or randomized) approximation techniques with error guarantees.

Constrained Frequent Itemset Mining from Uncertain Data Streams
Carson Leung (kleung@cs.umanitoba.ca)

Cleansing Uncertain Databases Leveraging Aggregate Constraints
Haiquan Chen (chenhai@auburn.edu)
Wei-Shinn Ku (weishinn@auburn.edu)
Haixun Wang (haixunw@microsoft.com)

U-DBSCAN : A Density-Based Clustering Algorithm for Uncertain Objects
Apinya Tepwankul (apinyate@ais.co.th)
Songrit Maneewongwattana (songrit@cpe.kmutt.ac.th)

 Recently, uncertain data management and mining has become a critical issue in many real applications, such as sensor data monitoring, location-based services, object identification, and moving object search. Unlike exact data, uncertain data are often represented as a set of discrete samples or a probability density function, which presents new challenges for analyzing, querying, and mining the uncertain data effectively and efficiently. Following the success of the First International workshop on Management and mining Of UNcertain Data (MOUND) 2009, the Second MOUND 2010 will continue to investigate key issues related to the data management and mining over uncertain data. Specifically, this forum welcomes contributions that explore uncertain data management issues such as data representation, various types of queries, and indexes. Additionally, the workshop hopes to attract work that studies the new data mining techniques of data cleaning, clustering, and classification over uncertain data including sensor data, location data, web and multimedia data.

Paper Submission

High quality research papers in the relevant areas are solicited. Original papers exploring new directions will receive especially careful consideration. Papers that have already been accepted or are currently under review for other conferences or journals will not be considered for MOUND10. Paper submissions should be limited to a maximum of 8 pages in the camera-ready format of ICDE10. All papers will be reviewed by the Program Committee on the basis of technical quality, relevance to data mining, originality, significance, and clarity. All accepted workshop papers will be included in a proceeding published by the IEEE Computer Society Press.

Please submit your papers using the following submission site: https://cmt.research.microsoft.com/MOUND2010

For any questions, please email to: shenht@itee.uq.edu.au or  graham@research.att.com.

Important Dates