[Compstats] Compstats seminar: Intro to Reinforcement Learning

Pascal Poupart ppoupart at cs.uwaterloo.ca
Mon Apr 12 14:45:47 EDT 2010

This Wednesday, I will give an intro to Reinforcement Learning.

Title: An Introduction to Reinforcement Learning

Speaker: Pascal Poupart
Date: Wed April 14
Time: 2:30pm
Location: AI seminar room (DC2306C)

In this talk, I will introduce the topic of reinforcement learning and 
relate it to other areas of machine learning.  While reinforcement 
learning was originally inspired by animal learning (i.e., how to train 
animals by reinforcements), today, it has evolved into one the most 
comprehensive and challenging form of machine learning. From a practical 
perspective, a wide variety of applications from robotic control and 
spoken dialogue management to game playing and ad online ad placement 
can be naturally modeled as reinforcement learning problems.  From a 
theoretical perspective, reinforcement learning includes active 
learning, online learning (i.e., sequential decision making) and 
arbitrary loss functions.  Furthermore, reinforcement learning is 
neither supervised nor unsupervised, but somewhere in between since 
there data provided is only correlated with the decisions.  Depending on 
time, I will try to cover some of the important results from learning 
theory and Bayesian learning.


Pascal Poupart
Associate Professor
David R. Cheriton School of Computer Science
University of Waterloo
200 University Avenue West
Waterloo, Ontario
Canada N2L 3G1
Web: http://www.cs.uwaterloo.ca/~ppoupart
Email: ppoupart at cs.uwaterloo.ca
Telephone: 1-519-888-4567x36239
Fax: 1-519-885-1208

More information about the Compstats mailing list