A

adversarial-search

Connect Four agent combining a Policy–Value CNN with Negamax search under a 2-second move budget. Experiments cover reinforcement learning, supervised learning from expert games, self-play, and search caching, with versioned data and configurable training pipelines.