State-of-the-art Research of Deep Reinforcement-learning

OpenAI research, DeepMind research, Google research, Microsoft research

Hello I am Nitsan Soffair, a Deep RL researcher at BGU.

What you’ll learn

• Get state-of-the-art knowledge of deep reinforcement-learning research.
• Be able to start deep reinforcement-learning research.
• Be able to get engineering job on deep reinforcement-learning.
• Be able to get research job on deep reinforcement-learning.

Course Content

• OpenAI research –> 6 lectures • 8min.
• DeepMind research –> 5 lectures • 6min.
• Google research –> 8 lectures • 13min.
• Microsoft research –> 1 lecture • 2min.

Requirements

Hello I am Nitsan Soffair, a Deep RL researcher at BGU.

In my State-of-the-art Research of Deep Reinforcement-learning course you will get the newest state-of-the-art Deep reinforcement-learning research knowledge.

You will do the following

1. Get state-of-the-art research knowledge regarding
   1. OpenAI research
   2. DeepMind research
   3. Google research
   4. Microsoft research
2. Validate your knowledge by answering short quizzes of each lecture.
3. Be able to complete the course by ~2 hours.

Topics

1. Advanced exploration methods
2. Chatbot based Deep RL
3. Evaluation strategies
4. Advanced RL metrics
5. Navigating robot get human language instructions
6. Merging on-policy and off-policy gradient estimation
7. Hierarchical RL
8. More advanced topics

Syllabus

1. OpenAI research
   1. Emergent Tool Use from Multi-Agent Interaction
   2. Learning Dexterity
   3. Emergent Complexity via Multi-Agent Competition
   4. Competitive Self-Play Better Exploration with Parameter Noise
   5. Proximal Policy Optimization
   6. Evolution Strategies as a Scalable Alternative to Reinforcement Learning
2. DeepMind research
   1. Recurrent Experience Reply in distributed Reinforcement-learning
   2. Maximum a Posteriori Policy Optimization
   3. NeuPL: Neural Population Learning
   4. Learning more skills through optimistic exploration
   5. When should agents explore?
3. Google brain research
   1. QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation
   2. FollowNet: Robot Navigation by Following Natural Language Directions with Deep Reinforcement Learning
   3. Interpolated Policy Gradient: Merging On-Policy and Off-Policy Gradient Estimation for Deep Reinforcement Learning
   4. Scalable Deep Reinforcement Learning Algorithms for Mean Field
   5. Value-Based Deep Reinforcement Learning Requires Explicit Regularisation
   6. Air Learning: a deep reinforcement learning gym for autonomous aerial robot visual navigation
   7. Deep Reinforcement Learning at the Edge of the Statistical Precipice
   8. Exploration in Reinforcement Learning with Deep Covering Options
4. Microsoft research
   1. Deep Reinforcement-learning for Dialogue Generation

Resources

• OpenAI papers
• DeepMind papers
• Google papers
• Microsoft papers