Logo ČVUT
CZECH TECHNICAL UNIVERSITY IN PRAGUE
STUDY PLANS
2018/2019

Games and reinforcement learning

The course is not on the list Without time-table
Code Completion Credits Range Language
MI-GLR Z,ZK 4 2P+2C
Lecturer:
Juan Pablo Maldonado Lopez (guarantor)
Tutor:
Juan Pablo Maldonado Lopez (guarantor)
Supervisor:
Department of Applied Mathematics
Synopsis:

The field of reinforcement learning is very hot recently, because of advances in deep learning, recurrent neural networks and general artificial intelligence. This course is intended to give you both theoretical and practical background so you can participate in related research activities.

Presented in English.

Requirements:
Syllabus of lectures:

Algorithmic game theory

1. Sealed-bid combinatorial auctions

2. Iterative combinatorial auctions

3. Stable matching

4. Congestion games. Selfish routing and the price of anarchy

5. Potential games. Network cost-sharing games

6. Best response dynamics. No-regret dynamics.

Introduction to Reinforcement Learning

7. Multiarmed Bandit Algorithms.

8. Finite Markov Decision Processes

9. Dynamic Programming

10. Montecarlo methods

11. Temporal-Difference learning

12. Multi-step bootstrapping

13. Planning and learning with tabular methods

Syllabus of tutorials:

Algorithmic game theory

1. Mechanism design basics. Auctions of physical goods.

2. Sponsored search auctions (online advertising).

3. Congestion games. Selfish routing and the price of anarchy

4. Traffic assignment in networks.

5. Best response dynamics. No-regret dynamics.

6. Rock, paper, scissors.

Introduction to Reinforcement Learning

7. Multiarmed Bandit Algorithms.

8. Markov chains and MDP's.

9. Algorithms: Q-learning, TD

10. Playing tic-tac-toe, checkers.

11. Tensorflow introduction.

12. Case studies: TD-gammon, Atari games, Go playing.

13. OpenAI Gym. Policy gradient algorithm.

Study Objective:

Teach theoretical and practical aspects of the game theory and reinforcement learning.

Study materials:

Reinforcement Learning: An introduction, Sutton and Barto, 2nd edition draft, 2017.

Algorithmic Game Theory, Roughgarden, Tardos, Vazirani and Nisan, 2007.

Note:
Further information:
No time-table has been prepared for this course
The course is a part of the following study plans:
Data valid to 2019-06-20
For updated information see http://bilakniha.cvut.cz/en/predmet5107306.html