Advanced Statistical
Reinforcement Learning
Class Overview
Deep Reinforcement learning is a promising area in modern artificial intelligence research. Its recent development based on computation power leads to enormous applications, including robotics, finance, healthcare, and business administration. This course will deal with deep reinforcement learning, and students will learn about the mathematical elements and algorithmic approaches. At the end of this course, students will understand core theories, algorithms, and recent research topics in deep reinforcement learning research.
Class Time / Classroom
Friday, 14:00-16:30 / College of Political Science and Economics Bldg. #501
Instructor: Sungbin Lim
TA E-mail: stat436.ta@gmail.com
If you have any question about attendance, exam notice, etc., please email this address.
Office Hour: Tuesday, 15:30 - 16:30, Woodang Hall (우당교양관), #530
Notice
(new) This course will be online via recorded lecture, due to the professor's business trip.
- No attendance check.
- The recorded lecture will not be shared.
- Students can watch the recorded lecture on this [Zoom, password: 0000] on regular course time starting at 14:00.
- Questions on chat will be delivered to the professor.[Paper list for final presentation]
Breaking the Deadly Triad with a Target Network, Zheng et al., ICML 2021 [링크]
Distributed Prioritized Experience Replay, Horgan et al., ICLR 2016 [링크]
Dueling Network Architectures for Deep Reinforcement Learning, Wang et al., ICML 2016 [링크]
Stabilizing Off-Policy Q-Learning via Bootstrapping Error Reduction, Kumar et al., NeurIPS 2019 [링크]
VIREL: A Variational Inference Framework for Reinforcement Learning, Fellows et al., NeurIPS 2019 [링크]
Conservative Q-Learning for Offline Reinforcement Learning, Kumar et al., NeurIPS 2020 [링크]
Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixels, Yarats et al., ICLR 2021 [링크]
Decision Transformer: Reinforcement Learning via Sequence Modeling, Chen et al., NeurIPS 2021 [링크]
On the courses on 2023.05.12
- This course is canceled due to the wave of COVID-19 infection and the NeurIPS submission period.
- Recorded lectures for 05.05 and 05.12 will be uploaded on 06.09.For the final assignment, a paper list for presentation will be noticed on 05.19 on this homepage.
- Students should pick a paper from the list.
- Prepare individual presentation recordings for around 20 minutes.
- Submit by 06.16 23:59 via Blackboard assignment.For the 13-15th week, classes are scheduled to take place instead of the previously planned offline presentations.
For the course on 2023.04.28
- This course will cover code implementation (offline online) [Zoom link].
- No attendance check.
- The recorded lecture will be uploaded on this course homepage.
- We apologize for the sudden change due to the professor's personal circumstances.For the course on 2023.05.05
- This course will cover code implementation (only online via recording, which will be uploaded later).
- Of course, no attendance check.The Course on 2023.04.21 will be closed due to the school's midterm exam period.