Exploiting Structure in Relational Inference and Optimization

Speaker: Vaishak Belle , University of Edinburgh

Date: Friday, April 07, 2017

Time: 3:00 PM to 4:00 PM Note: all times are in the Eastern Time Zone

Public: Yes

Location: 32-G882 (Stata Center - Hewlett Room)

Event Type:

Room Description:

Host: Leslie Kaelbling, MIT

Contact: Leslie Pack Kaelbling, 8-9695, lpk@csail.mit.edu

Relevant URL: vaishakbelle.com

Speaker URL: None

Speaker Photo:
None

Reminders to: seminars@csail.mit.edu

Reminder Subject: TALK: Exploiting Structure in Relational Inference and Optimization

Exploiting Structure in Relational Inference and Optimization

Probabilistic inference and optimization are central problems in machine learning and AI. The advent of large-scale probabilistic knowledge bases, as well as areas such as natural language processing, game playing and automated planning, have generated enormous interest in relational representations for data management, often defined over complex constraints and dependencies. Unfortunately, developing solvers for such languages is very challenging, and algorithms often expect models to be painstakingly reduced to standard forms.

In this talk, we look at two recent results on the theory and implementation of general-purpose solvers for structured specifications. In the context of inference, we discuss results that show how statistical relational models in open-universes can be treated, while inheriting results on “lifted” algorithms. Open-universe settings frequently arise in robotics and Web mining where complete knowledge of relations and constants cannot be assumed a-priori. In the context of optimization, we consider an approach that allows and naturally exploits symbolic dependencies for convex linear and quadratic programs.

Part of this is joint work with Martin Mladenov and Kristian Kersting.

Bio: Vaishak Belle is an assistant professor and Chancellor’s Fellow at the School of Informatics, University of Edinburgh, UK. Vaishak’s research is in artificial intelligence. Previously, he was at KU Leuven, the University of Toronto, and the Aachen University of Technology. He has co-authored several articles in AI-related venues, and won the Microsoft best paper award at UAI, the Machine learning journal best student paper award at ECML-PKDD, and the Kurt Goedel silver medal.

Research Areas:

Impact Areas:

This event is not part of a series.

Created by Leslie Pack Kaelbling Email at Saturday, April 01, 2017 at 8:40 AM.