RL Gymnasium

This repo is an exploratory codebase for learning and interacting with reinforcement learning concepts, libraries, and notebooks.

Intro

The goal of this project is to learn through practice by implementing a variety of reinforcement learning algorithms. The project can be broken up into two sections: algorithms, and integrations. Where integrations will be interfaces with other packages that provide test environments for RL.

Notes

• Action-Value
• Stationary k-armed Bandit Problems
• Non-stationary k-armed Bandit Problems

Algorithms

1. greedy
2. ε-greedy
3. Optimistic Initial Values
4. Upper-Confidence-Bound (UCB)
5. Bandit Gradient Algorithm
6. Associative Search (Contextual Bandits)

Component Algorithms

1. argmax

Tests

1. k-armed Bandit

Integrations

Outro