ML Training Dashboard

This page trains a neural network to play Commander decisions. Follow these steps in order on your first run:

1. Run batch simulations with ML logging enabled (from the Simulator page) to generate decision log files.
2. Start Training below — leave “Rebuild dataset” checked so it parses your logs into a dataset automatically.
3. Check Evaluation Results after training. Aim for >60% test accuracy before moving on.
4. (Optional) Run PPO reinforcement learning to fine-tune the policy via self-play. Only recommended after a solid supervised baseline.
5. Run a Tournament to benchmark your model against heuristic and random agents and get a win-rate score.