BYU CS Logo
Computing That Serves

Model Tolerant Agents

Date: 

Thursday, September 30, 2010 - 11:00am

Speaker: 

Daniel Bryce
Assistant Professor
Dept. of Computer Science
Utah State University

Assistant Professor

 

Utah State University

 

Dept. of Computer Science

Host: 

Mike Goodrich

Autonomous agents commonly rely on models of their environment to predict, plan, and otherwise reason about their actions. However, the task of designing effective models is fraught with peril from either the top-down, knowledge engineering perspective, or the bottom-up machine learning approach.  This talk will describe work on striking a middle-ground where agents can accept as much knowledge as given and learn the task-relevant knowledge on demand.  By combining planning with incomplete knowledge and model-based reinforcement learning, agents can exploit the knowledge given and identify learning goals that can be achieved by designing experiments or querying human domain experts.

Biography: 

Daniel Bryce is an assistant professor in the Computer Science Department at Utah State University.  Dr. Bryce studies Artificial Intelligence, focusing on automated planning and agents, and their application to defense, biology, and manufacturing.  Dr. Bryce was the recipient of the 2009 International Conference on Automated Planning and Scheduling Outstanding Dissertation award for his work on scaling planning under uncertainty.  Prior to joining USU, Dr. Bryce worked at SRI International and lectured at Stanford University, was awarded research fellowships at Honeywell Labs and NASA Ames Research Center, and received his PhD from Arizona State University.




Academics