Taiyou / DeepRL-Nanodegree

This is repository for Deep Reinforcement learning Nanodegree

This is repository for Deep RL nanodegree in Udacity.

Contents of the sections Textbook by Sutton and Bart

1. Introduction to Deep Reinforcement Learning

Lesson: Temporal-Difference Methods.

• Expected Sarsa
• Q-Learning
• Optimistic Q-Learning

Lesson: Solve OpenAI Gym's Taxi-v2 Task

• Hierarchical Reinforcement Learning with the MAXQ Value Function Decomposition

Lesson: RL in Continuous Spaces

• Discretization
• Tile Coding
• Coarse Coding
• Function Approximation

2. Value-Based Methods

Deep RL

• Learning to see and act
• Video: Deep Q-Learning
• Human-level control through deep reinforcement learning

Deep RL for Robotics

• the Japanese robot company Fanuc
• Robot learns via trial and error like a human
• DEEP LEARNING IN PRODUCTION & WAREHOUSING WITH AMAZON ROBOTICS

Opportunities

• UC Berkeley Course

Deep Q-networks (DQN)

• Experience Relay: A Deeper Look at Experience Replay
• Fixed Q Targets:
• Doueble DQN: Deep Reinforcement Learning with Double Q-learning
• Prioritized Experience Replay: Prioritized Experience Replay
  • Some transition information should be sampled.
• Dueling Network Dueling Network Architectures for Deep Reinforcement Learning
  • Pros:

Some other proposed methods

• Rainbow: Rainbow: Combining Improvements in Deep Reinforcement Learning
• Asynchronous Methods for Deep Reinforcement Learning
• Distribution DQN: A Distributional Perspective on Reinforcement Learning
• Noisy DQN: Noisy Networks for Exploration

3. Policy-Based Methods

4. Multi-Agent Reinforcement Learning

Extracurricular

1. Special Topics in Deep Reinforcement Learning

2. Neural Networks in PyTorch

3. Computing Resources

4. C++ Programming