Sie sind hier: Startseite Teaching SS2018 Hierarchical Reinforcement …

Hierarchical Reinforcement Learning

Hierarchical Reinforcement Learning (Seminar)

 

  • Organized by:
    • Neurorobotics Lab Members
  • Seminar
    • HISinOne number: 11LE13S-7325
  • Language:
    • English
  • First meeting:
    • Wednesday, April 18th 2018, 14:00-15:00, bldg. 080, seminar room ground floor
    • Topics: very short introduction, general organization
  • Second meeting:
    • Wednesday, May 2nd 2018, 14:00-15:00, bldg. 080, seminar room ground floor
    • Topics: advice on giving presentations, good scientific behavior, more organizational details
  • Third meeting:
  • Presentations: Monday, July 23rd and Tuesday, July 24th 2018, bldg. 080, seminar room ground floor

 

Overview:

We'll read and discuss a number of papers from the area of hierarchical reinforcement learning. This will include a survey on early approaches to give context and background information, as well as recent papers presented by participants.

 

Format:

We'll have presentations of 45 min (30-35 min talk + 10-15 min for Q&A) for each paper. In addition, students will prepare a 2-3 page summary of the paper they presented, and answer questions on the other papers in the seminar. Questions will be posted within the next few weeks.

 

Paper Selection:

 

Paper Questions
1. A Deep Hierarchical Approach to Lifelong Learning in Minecraft hrl-q1
2. Hierarchical Actor-Critic hrl-q2
3. FeUdal Networks for Hierarchical Reinforcement Learning hrl-q3
4. DDCO: Discovery of Deep Continuous Options for Robot Learning from Demonstrations  hrl-q4
5. Learning by Playing – Solving Sparse Reward Tasks from Scratch hrl-q5
6. The Option-Critic Architecture hrl-q6
7. Stochastic Neural Networks for Hierarchical Reinforcement Learning hrl-q7