2025-12-15

Reinforcement Learning Mario Project

machine-learning deep-learning reinforcement-learning python

Overview

A deep reinforcement learning project that trains an AI agent to master Super Mario Bros levels autonomously. The agent uses convolutional neural networks to process game frames and learns optimal strategies through policy gradient methods.

Key Features

  • Implements Deep Q-Network (DQN) algorithm for decision-making
  • Uses OpenAI Gym environment for Mario Bros simulation
  • Processes raw pixel data through CNN layers for state representation
  • Applies experience replay for stable training
  • Achieves progressive level completion through iterative learning

Technologies

Python PyTorch OpenAI Gym Deep Learning Reinforcement Learning

Source Code

View on GitHub