Title: Learning in the presence of too much data

Speaker: Jonathan Mugan, University of Texas at Austin

Date/Time: Wednesday, June 30, 2010, 9-10 AM      

Location: CSRI Building/Room 279 (Sandia NM)

Brief Abstract: Sensors often provide a flood of information about the environment, and we desire methods that can distill this flood down to important events and models that allow us to make predictions and decisions. How an undifferentiated stream of information is broken up into discrete events determines what models can be learned. But interestingly, the current set of models can also help us to identify important new events. This talk presents a method for exploiting this synergy by simultaneously learning a set of events and a set of models in continuous environments. This method also enables an agent to autonomously convert the learned models into plans to bring about desired events in the world.

CSRI POC: Brett Bader, (505) 845-0514