Content

Chronological log of training experiments, each documenting the configuration, results, and lessons learned.

---

Table of contents

• 1. Baseline
• 2. Enhance Boss Fight Training
• 3. Action Test: Monte Carlo Search
• 4. Action Test: MC search to beat boss
• 5. Better Action Space
• 6. The First Setback
• 7. Behavior Cloning — A New Baseline
• 8. Behavior Data Collection
• 9. Fixed Anchor for Save and Load
• 10. Search to Win the Game