Reinforcement Learning
Agents that learn to trade stocks using Q-learning, SARSA, and DQN on OpenAI Gymnasium.
What
Coursework from grad school. A set of notebooks that walk through classic reinforcement learning on small, legible environments before pointing the same methods at a toy stock-trading setup.
Approach
Started with tabular Q-learning and SARSA on gridworlds to build intuition for exploration versus exploitation. Moved to Deep Q-Networks with experience replay for continuous state spaces, and applied the same code to a tiny trading environment built on top of OpenAI Gymnasium.
Stack
Python, NumPy, PyTorch, Jupyter, Gymnasium. No GPU needed for any of it.