2020; RL Algorithms Implementation

Introduction

So around end of 2018 I started learning about Reinforcement Learning from Coursera and from David Silver's UCL course. Over time I focused on implementing the algorithms I learned from scratch, mainly in Tensorflow 1.x and then later in 2.x. I also ended up implementing a lot of them during the Covid lockdowns in India.

Check it out at: https://github.com/otoshuki/RL_Projects

This repository contains all of my custom implementations for the following:

1. Crossentropy

2. Deep Crossentropy

3. Value Iteration

4. Q-Learning with e - greedy exploration

5. SARSA

6. Expected Value SARSA with e-greedy exploration

7. Q-Learning with Experience Replay

8. Deep Q-Learning on TF1.x and TF2

9. Deep Q-Learning on TF2 with Experience replay and Target network

10. REINFORCE - Policy Gradient algorithm on TF1.x and TF2

11. Double Q-Learning on TF2 with Experience replay

12. Advantage Actor-Critic on TF2

13. Proximal Policy Optimization

14. Twin Delayed DDPG

There are a lot of good sources for the explanation behind most of these algorithms, so I will refrain from the explanations.