Special Course in Machine Learning: AI-Safety 2023/24 fall

Timetable

The video lecture are from David Silver's reinforcement learning course. Some homeworks are also from his course, but others are from Berkley deep reinforcement learning course.

Date	Video lecture	Assignment	Test presenter	Homework presenter
8.02.2016	Introduction to Reinforcement Learning (slides)	none
15.02.2016	Markov Decision Processes (slides) (test)	Implementation of Easy21	Zura Isakadze	Aqeel Labash
22.02.2016	Planning by Dynamic Programming (slides) (test)	Value Iteration and Policy Iteration	Kristjan Jansons	Lauri Tammeveski
29.02.2016	Model-Free Prediction (slides) (test)	Monte-Carlo Control in Easy21	Irene Teinemaa	Gagandeep Singh
7.03.2016	Model Free Control (slides) (test)	TD Learning in Easy21	Lauri Tammeveski	Kristjan Jansons
14.03.2016	Value Function Approximation (slides) (test)	Linear Function Approximation in Easy21	Aqeel Labash	Üllar Lindmaa
21.03.2016	Policy Gradient Methods (slides) (test)	Implement policy gradient method	Ardi Tampuu	Irene Teinemaa
28.03.2016	Integrating Learning and Planning (slides) (test)	Implement some enhancement or variation on policy gradient optimization algorithm	Ilya Kuzovkin	Zura Isakadze
4.04.2016	Exploration and Exploitation (slides) (test)	Experiment with Atari domain	Üllar Lindmaa	Ilya Kuzovkin
11.04.2016	Case Study: RL in Classic Games (slides) (test)	TORCS rally simulator and RL	Daniel Majoral	Ardi Tampuu
18.04.2016	Bonus video: Richard Sutton's introduction to reinforcement learning
25.04.2016	Bonus video: David Silver's lecture about deep reinforcement learning
2.05.2016	Bonus video: Nando De Freitas's lecture about reinforcement learning with direct policy search
9.05.2016	Bonus video: Nando De Freitas's lecture about reinforcement learning with action-value functions
16.05.2016	Torcs competition rehearsal
17.05.2016	Jaan Tallinn's lecture about AI control as reinforcement learning problem 15:15 in Paabel Torcs competition after Jaan Tallinn's lecture
23.05.2016	Mastering the game of Go with deep neural networks and tree search by Ilya Kuzovkin