This is a repository containing all codes/reports/images of the three-phase FinalProject in Artifical Intelligence(DATA130008.01). All the agents in this project are submitted and evaluated via Kaggle.
- Adapted from the Connect X competition in Kaggle.
- You need to implement a MCTS(Monte Carlo Tree Search) agent for the game.
- You need to choose one game from Connect X, Santa, Kore and Lux AI to implement a RL(Reinforcement Learning) agent.
- The RL should be based on value/policy iteration in this phase.
- You need to make a poster and give a presentation in no more than 2 minutes to clarify your agent implemented in FinalProject2.
- The method is not confined to value/policy iteration any more. Any method is allowed in this phase.
- Requirements can be found here.
We choose Santa in phase 2. The best agents of our team in Connect X and Santa all rank second in the leaderboard. Below are the tables display the Kaggle scores of our submissions.
Connext X (rank 2/26) |
Santa (rank 2/15) |
Submissions | Scores |
---|
agent | 550.4 | agent2 | 887.8 | agent4 | 1894.9 | agent5 | 2327.7 | |
Submissions | Scores |
---|
agent-test | 585.8 | agent-test-2 | 1066.1 | agent-1 | 2661.4 | agent-5 | 1453.6 | |