CPSC 532I: Topics in AI: Reinforcement Learning

Reinforcement learning is the branch of machine learning that studies learning to act. Agents observe, predict, and act to change their environment. Reinforcement learning has notable success in learning to play games and control robots. In this course, we will cover fundamental concepts and algorithms, and introduce techniques that underlie many of the successes from reinforcement learning.

Instructor: Jason Peng
TA: TBD
Lectures: Wednesday 11:00am-12:30pm
Friday 11:00am-12:30pm

Grading

3 programming assignments (40%)

A1 (10%) - Due Oct 5
A2 (15%) - Due Oct 26
A3 (15%) - Due Nov 23

Course project (60%)

Info
Proposal (10%) - Due Oct 19
- Template
Presentation (25%)
- Info
- Signup Sheet
Report (25%) - Due Dec 7
- Info
- Template

Late days: You have 3 late days that you can use for any assignment. You can distribute the late days however you like, but they can only be applied to programming assignments. Once you run out of late days, any late assignments will no longer be accepted.

Syllabus

Sep 3

Introduction

Sep 8

MDP

Sep 10

Policy Evaluation

Sep 15

Policy Evaluation, Behavioral Cloning

Sep 17

No Class

Sep 22

Behavioral Cloning

Sep 24

Policy Search

Sep 29

Policy Gradient

Oct 1

No Class

Oct 6

Policy Gradient

Oct 8

Q-Learning

Oct 13

No Class - Thanksgiving

Oct 15

Q-Learning

Oct 20

Actor-Critic Algorithms

Oct 22

Model-Based RL

Oct 27

Model-Based RL, On-Policy vs Off-Policy Algorithms

Oct 29

Advance Policy Gradient

Nov 3

Advance Policy Gradient, Advance Q-Learning

Nov 5

Advance Q-Learning

Nov 10

Exploration

Nov 12

Domain Transfer

Nov 17

Domain Transfer, Project Presentations

Nov 19

Project Presentations

Nov 24

Project Presentations

Nov 26

Project Presentations

Dec 1

Project Presentations