Course Schedule

Introduction and logistics. Basics of game theory, equilibria, zero-sum games, Minimax theorem.

Presentation: Lecture 1

Readings:

• Chapters 1-2 of AI, Games and Markets
• Lecture Notes

Basics of online learning algorithms and theoretical foundations.

Presentation: Lecture 2

Readings:

• Lecture Notes
• Lecture Notes, First Half

Further Reading

• Online Learning and Online Convex Optimization, Chapters 1-2

Presentation Python Tutorial

Solving zero-sum games using no-regret dynamics.

Presentation: Lecture 3

Readings:

• Lecture Notes

Applications of zero-sum games in machine learning and AI: boosting, adversarial robustness, distributional robustness, fairness, GANs, Imitation learning, Reinforcement learning from human feedback, NPIV (causal machine learning).

Presentation: Lecture 4

Readings:

• Game Theory, Online Prediction and Boosting
• Tutorial on GANs, Especially Sections 3.2, 5.1, 8.2
• Robust Optimization for Non-Convex Objectives
• Distributionally Robust Neural Networks
• Adversarial Training for Causal ML
• Zero-Sum Games for RLHF
• Zero-Sum Games for Imitation Learning
• Zero-Sum Games for Fair ML

Due: April 22nd

Extensive form games of complete and incomplete information.

Presentation: Lecture 5

Readings:

• Chapter 8-9, AI, Games and Markets

Videos:

• Noam Brown, Superhuman AI for heads-up no-limit poker: Libratus beats top professionals

Further Reading:

• Stratego AI
• Poker AI
• Chess, Shogi and Go AI
• Optimistic FTRL for Extensive Games

Presentation: Lecture 6

Readings:

• Chapter 8-9, AI, Games and Markets

Videos:

• Noam Brown, Superhuman AI for heads-up no-limit poker: Libratus beats top professionals

Further Reading

• Chapter 5, AI, Games and Markets
• Lectures 11-15, CMU Computational Game Solving Course

Further Videos:

• Noam Brown, Deep Learning and Search for Imperfect Information Games

Due: April 29th

No-regret learning in general games and coarse correlated equilibria. No-swap regret and correlated equilibria.

Presentation: Lecture 7

Readings:

• Lecture Notes

Further Readings:

• Quality of Learning Outcomes Lecture Note

Learning algorithms for correlated equilibria. Reduction from no-regret to no-swap regret.

Presentation: Lecture 8

Readings:

• Lecture Notes

Further Readings:

• From Internal to External Regret, Blum and Mansour
• Example of Correlated equilibria in recent multi-agent RL research
• Noah Golowich, Lecture Notes on Stochastic Games, Part I
• Grandmaster Level AI Agent for Starcraft

Due: May 7th

Basics (Bayes-Nash equilibrium, truthfulness, First-Price auction, Vickrey auction) and applications (GSP, GFP etc). Sponsored search.

Presentation: Lecture 9

Readings:

• Lecture Notes

Presentation: Lecture 10

Readings:

• Lecture Notes

Further Readings:

• Shai-Shalev Schwarz, OCO, Chapters 4.2 and 2.7-2.8

Presentation: Lecture 11

Readings:

• Lecture Notes

Presentation: Lecture 12

Readings:

• Lecture Notes
• Lecture Notes
• Lecture Notes

Presentation: Lecture 13

Readings:

• Lecture Notes
• Lecture Notes

Presentation: Lecture 14

Readings:

• Lecture Notes
• Lecture Notes

Presentation: Lecture 15

Readings:

• Lecture Notes on Estimation of CDFs
• Lecture Notes on Estimation from First Price Auctions

Presentation: Lecture 16

Readings:

• Lecture Notes on A/B Testing in Auctions
• A/B Testing of Auctions, Chawla, Hartline, Nekipelov
• A/B Testing in Auctions with Budgets from Meta
• A/B Testing in Two-Sided Markets

Presentation: Lecture 17 A/B Testing in Two-Sided Markets: Colab Notebook

Readings:

• A/B Testing in Auctions with Budgets from Meta
• A/B Testing in Two-Sided Markets