MachineLearning

A collection of foundational reinforcement learning (RL) algorithm implementations.
This repository serves as a learning playground with clear scaffolds and code for building, experimenting with, and understanding key RL methods.

---

Implemented Algorithms

• Q-Learning
  Tabular off-policy RL algorithm using value iteration.

• Deep Q-Networks (DQN)
  Neural network–based extension of Q-Learning for function approximation.
  Includes replay buffer and target network scaffolds.

• Actor-Critic
  Policy-gradient + value-based hybrid method.
  Separates actor (policy) and critic (value estimator) for stable training.

---

Repo Structure

MachineLearning/
├── q_learning/       # Q-Learning implementation and utilities
├── dqn/              # Deep Q-Network scaffold and training loop
├── actor_critic/     # Actor-Critic scaffold and modules
├── utils/            # Shared helpers (e.g., replay buffer, env wrappers)
└── README.md

---

Quickstart

1. Clone the repo

   git clone https://github.com/KhushalM/MachineLearning.git
   cd MachineLearning

2. Set up environment

   python -m venv .venv && source .venv/bin/activate
   pip install -r requirements.txt

3. Run a training script (example for Q-learning)

   python q_learning/train.py

4. Experiment
   Modify hyperparameters, environments, or extend scaffolds into more advanced RL methods.

---

Goals

• Provide minimal, readable implementations of core RL algorithms.
• Serve as a starting point for deeper experimentation (e.g., PPO, A2C, GRPO).
• Build intuition by connecting math → code → results.

---