AI-Pong documentation

AI-Pong evolves a small neural network population to play Pong. The Python simulation and genetic algorithm use PyTorch, while Arcade renders one representative play zone and its network.

Get started

Install the project requirements and launch training from the repository root:

python -m pip install -r requirements.txt
python -m src.main

python -m src.main opens the rendered training runner. For headless runs with periodic progress reports, use python -m src.train. Both run continuously until interrupted unless --generations sets a limit. To view the most recently saved elite model, run python -m src.tester.

Publishing this site

The deployment workflow publishes this documentation under /AI-Pong/docs/ and the project homepage under /AI-Pong/ on pushes to main or manual dispatch. Before its first run, a repository maintainer must set Settings → Pages → Build and deployment → Source to GitHub Actions.

Guides

• Architecture: population evaluation, neural network, simulation, and rendering.
• Homepage model demo: browser inference and exporting future checkpoints.
• Fitness function: what the genetic algorithm rewards.
• Training tips: timeouts, population size, and mutation.
• Release 1.4.0: CLI reference, saved-run formats, lifecycle, migration notes, and known limitations.
• Coordinate system: simulation coordinates and Arcade rendering.
• Roadmap: current project direction.