Reinforcement Learning Journey 🚀

Welcome to the Reinforcement Learning Journey!
This repository is a personal and practical exploration of classic and modern reinforcement learning (RL) algorithms, with hands-on code and experiments.

---

📚 What You'll Find

• Q-Learning:
  Classic tabular Q-learning for discrete environments (e.g., FrozenLake).
  Simple, interpretable, and a great RL starting point!

• Deep Q-Network (DQN):
  Neural network-based Q-learning for environments with larger or continuous state spaces.
  Includes experience replay, target networks, and more.

• REINFORCE:
  The foundational policy gradient algorithm.
  Learn how to directly optimize policies using episode returns.

---

🤖 Agents on the go

Implementing RL Algorithms on multiple games.

---

🛠️ Getting Started

1. Clone the repository

   git clone https://github.com/cannibalcheeseburger/reinforcement-learning.git
   cd reinforcement-learning
   

2. Set up the environment

   conda env create -f environment.yml
   conda activate rl
   

3. Run the Notebooks

   • Open JupyterLab or Jupyter Notebook and explore the projects in each subfolder.

---

💡 Highlights

• Modular code: Each algorithm in its own folder for clarity.
• Reproducible experiments: Environment files and scripts included.
• HPC-ready: Scripts for running jobs on high-performance clusters.
• Learning by doing: Code is commented and designed for experimentation.

---

👾 Algorithms Covered

• Q-Learning (Tabular)
• Deep Q-Network (DQN)
• REINFORCE (Policy Gradient)
• More coming soon!

---

📈 Example Results

• Q-Learning: Solves FrozenLake-v1 (4x4 and 8x8) with tabular methods.
• DQN: Learns to balance CartPole and tackle more complex environments.
• REINFORCE: Demonstrates policy gradient learning on discrete tasks.