Seminar

Hierarchical Reinforcement Learning
Dr. Joschka Boedecker

Seminar:
- HISinOne Number: 11LE13S-7308

Time/Venue:
- This will be a block seminar held shortly after lectures have ended. A first meeting for planning purposes will take place on Monday, June 13th, 4:00 PM – 5:00 PM in room SR 00 019 (kitchen/seminar room on the upper floor) of building 079.

Language:
- English

Overview:

We'll read and discuss a number of papers from the area of hierarchical reinforcement learning. This will include papers on early approaches, new developments, and papers to give context and background information.

Papers:

1. Reinforcement Learning with Hierarchies Of Machines

2. The MAXQ Method for Hierarchical Reinforcement Learning

3. Between MDPs and semi-MDPs: A framework for temporal abstraction in reinforcement learning

4. Discovering Hierarchy in Reinforcement Learning with HEXQ

5. Learning Options in Reinforcement Learning

6. PolicyBlocks: An Algorithm for Creating Useful Macro-Actions in Reinforcement Learning

7. Hierarchical Reinforcement Learning Based on Subgoal Discovery and Subpolicy Specialization

8. Acquisition of stand-up behavior by a real robot using hierarchical reinforcement learning

9. Scalable Reinforcement Learning through Hierarchical Decompositions for Weakly-Coupled Problems

10. Hierarchical Relative Entropy Policy Search

11. The Option-Critic Architecture

12. Hierarchical Deep Reinforcement Learning: Integrating Temporal Abstraction and Intrinsic Motivation

Format:

We'll have presentations of 45 min (30 min talk + 15 min for Q&A) for each paper. In addition, students will prepare a 3-page summary of the paper they presented.