###LOGO_MOBILE###
###SOCIAL_NETWORKS###

SIGLE NEWS - DO NOT DELETE

Ομιλία Θ. Ρεκατσίνα, University of Maryland, 9/1/2012

Video της ομιλίας (~1.9GB)

Title: Exploiting Local Structure in Correlated Probabilistic Databases

Ομιλητής: Θεόδωρος Ρεκατσίνας, University of Maryland

Ημερομηνία/Ώρα: Δευτέρα 9 Ιανουαρίου 2012, 12:00μμ

Αίθουσα: Αμφιθέατρο Κτιρίου Επιστημών

 

Περίληψη:

Increasing numbers of real-world application domains are generating data that is inherently noisy, incomplete, and probabilistic in nature. Statistical analysis and probabilistic inference, widely used in those domains often introduce additional layers of uncertainty. Examples include data integration and information extraction on the Web, social network analysis and biomedical data management. Managing and queering such data requires us to combine the tools and the techniques from a variety of disciplines including databases, first-order logic, and probabilistic reasoning. Over the last few years, numerous approaches have been proposed, and several systems built, to integrate uncertainty into databases. Most approaches make restrictive independence assumptions concerning the types of uncertainties that can be represented and cannot easily model rich levels of abstractions, needed to handle large-scale uncertain datasets.

In this talk, I will begin by presenting the probabilistic data management system developed at the University of Maryland, called PrDB, aimed at supporting rich correlation structures often present in real-world uncertain datasets. I will present the PrDB representation model, which is based on probabilistic graphical models and show how these enable PrDB to support uncertainties specified at various abstraction levels, from schema-level uncertainties to tuple-specific uncertainties. Query evaluation in PrDB can be seen as equivalent to inference in graphical models, and I will present a novel technique we developed for efficient query evaluation and its relationship to factorized representations of probability distributions, such as Arithmetic Circuits and Ordered Binary Decision Diagrams. I will conclude with some of the open research challenges in the area.



Σύντομο Βιογραφικό:

Thodoris Rekatsinas is a second year Ph.D. student in Computer Science at the University of Maryland. He is a member of the DB Group, working with Prof. Amol Deshpande and a member of the LINQS Laboratory directed by Prof. Lise Getoor. His research interests are in the fields of machine learning, large scale probabilistic inference, probabilistic and temporal graph databases. Prior to joining UMD, he received his Diploma in ECE from the National Technical University of Athens.

© Σχολή Ηλεκτρολόγων Μηχανικών & Μηχανικών Υπολογιστών 2014
Πολυτεχνείο Κρήτης