Syllabus

Reading should be done after the associated lecture. Demonstrations during the lecture will be done in the Julia language using IJulia notebooks.

Date Topic Reading Due
Sep 25 Overview DMU 1
Sep 27 Probabilistic Representations DMU 2.1
Oct 2 Probabilistic Inference DMU 2.2
Oct 4 Parameter Learning DMU 2.3
Oct Midterm 1 Problem Review Session, 930am - 1020am, @ Gates B03 (bring your questions!)
Oct 9 Midterm 1 @ T.B.D (in class time 130pm-250pm)
Oct 11 Structure Learning DMU 2.4 Project 0
Oct 16 Decision Theory and Games DMU 3.1-3.3
Oct 18 Markov Decision Processes DMU 4.1-4.4 Project 1
Oct 23 Approximate Dynamic Programming DMU 4.5-4.7
Oct 25 Exploration and Exploitation DMU 5.1
Oct 30 Model-Based Reinforcement Learning DMU 5.2-5.3
Nov 1 Model-Free Reinforcement Learning DMU 5.4-5.5
Nov 6 Deep Reinforcement Learning Slides
Nov 8 State Uncertainty DMU 6.1-6.2
Nov 10 Project 2
Nov 13 Exact POMDP Methods DMU 6.3
Nov 15 Offline POMDP Methods DMU 6.4 Proposal due at 1159pm
Nov 16 Online POMDP Methods (Take-Home Quiz Released After Lecture) DMU 6.5
Nov 20 Thanksgiving week
Nov 24 Thanksgiving week
Nov 27 Take-home Quiz
Nov Midterm 2 Problem Review Session, 930am - 1020am, @ Thornton 102 (bring your questions!)
Nov 29 Midterm 2 @ T.B.D (in class time 130pm-250pm)
Dec 4 Guest Lectures
Optimal Aircraft Rerouting during Commercial Space Launches [Rachael Tompa] Slides
The Value of Inferring the Internal State of Traffic Participants for Autonomous Freeway Driving [Zach Sunberg] Slides
MDP Policy Compression through Neural Networks [Kyle Julian] Slides
Dec 6 Guest Lectures
Cooperative Control Using Deep Reinforcement Learning [Max Egorov / Jayesh Gupta] Slides or folder
Modeling Human Driver Behavior with Imitation Learning [Jeremy Morton] Slides
Making POMDPs Great Again [Louis Dressel] Slides
Dec 11 No class Final Project Paper
Dec 13 No class Peer Review*

Note: The dates for the various topics are approximate. Some topics may span multiple lectures. However, the schedule for the midterms will stay fixed. *Peer review can be submitted remotely online.