Philip S. Thomas, M.S.

Graduate Student
Autonomous Learning Laboratory

Department of Computer Science
University of Massachusetts Amherst
140 Governors Drive
Amherst, MA 01003-9264 U.S.A.

pthomas [æt] cs [daat] umass [daat] edu
Adviser: Andrew Barto

The web Search PSThomas.com
Research Interests:

 My general area of interest is reinforcement learning (RL): creating machines that can learn when provided rewards relating to their performance. They learn in a way similar to the way dogs are trained: when they do something well, they are given a reward. When they do something wrong, they are punished. My interest in RL stems from these similarities to the ways in which animals (including humans) learn.

Previous RL research tends to focus heavily on solving an individual problem. However, learning a complex task, such as driving a car using visual inputs, cannot be done from scratch. Rather, the agent must first learn skills (hierarchical reinforcement learning [HRL], using intrinsic motivation) and features that are useful (feature creation/selection). To drive, these would include picking out lines, corners, shapes, and eventually objects from the visual input, and abilities such as grasping the steering wheel, turning the wheel, etc. Only once an agent has these features and skills can it have any hope of learning to drive.

My current goal for my dissertation is to develop an HRL agent that performs feature creation and selection and is capable of transferring these skills and features to new problems. This seems quite daunting, though there has been significant progress in HRL and feature creation/selection. I can therefore focus on the problem of knowledge transfer between problems. My initial goal will be to develop architectures that can transfer well between very simple tasks in an attempt to better understand the nuances of the field.

Publications

Resume

Useful C++ Stuff

Logic Puzzles

Links

Created: 6-24-2009. Last Modified: 9-1-2009. © Philip S. Thomas