This group developed a reinforcement learning agent for BattleSnake, a synchronous multiplayer strategy game. Impressively, the group designed their game-playing agent, AlphaSnake Zero, by drawing from ideas from DeepMind's AlphaGo Zero and making many clever, custom design decisions. Their agent is quite competitive in the BattleSnake worldwide leaderboard.
Report
This project investigated methods for automatically classifying YouTube videos into one of two categories, using features like video thumbnails, transcripts, and video statistics. The group constructed a custom dataset using the YouTube API, and a considerable amount of attention was given to using state-of-the-art methods for extracting sophisticated features, methods that went beyond what was covered in the course.
Report