Title: Semistructured Probabilistic Databases

Date: Monday, 20 Aug 2001

Time: 4:00-5:00pm

Venue: Room 1505 (near lift nos. 25/26, opposite to Lecture Theater G) Hong Kong University of Science & Technology

Abstract:
The need to store and manipulate probabilistic information is widespread. It occurs today domains from survey and market research, logistics support, robotics, medical applications and many others. Although there is a wealth of research on probabilistic models of inference (Bayes Nets, Markov Decision Processes, Probabilistic Logic Programs, to name a few), there has been relatively little work on management of probabilistic data from the database perspective.

This work describes a new theoretical framework for representation and management of diverse probabilistic information. The unit of information in this framework, a Semistructured Probabilistic Object (SPO), is designed to store information about one probability distribution over a set of random variables together with the non-stochastic information that may be associated with that distribution. SPOs can store simple, joint and conditional probability distributions.

We have have defined the operations of selection, projection, cartesian product, join and conditionalization on SPOs and given efficient algorithms for computing them. The SPO data model and the query algebra provide a flexible solution for representing, storing and querying diverse probabilistic information.

This is joint work with Dr. Alex Dekhtyar and Sean Hawkes.

Biography:
Judy Goldsmith received her PhD in mathematics from the University of Wisconsin-Madison in 1988. Since then she has taught in math and computer science departments at Dartmouth College, Boston University, and the University of Manitoba. She is now at the University of Kentucky, where she is an associate professor.

Her work includes research in computational complexity, computational learning theory, control of stochastic processes (POMDPs and Bayes nets), and most recently, semistructured probabilistic databases.

In her (somewhat limited!) spare time, she likes folk dancing.