BridgeHand2Vec Bridge Hand Representation
This addresses incomplete information challenges in contract bridge for AI researchers, though it is incremental as it builds on existing neural network methods for game representation.
The paper tackles the problem of representing bridge hands for AI applications by proposing BridgeHand2Vec, a neural network that embeds hands into a vector space to reflect strength and enable interpretable distances, achieving state-of-the-art results on trick estimation with concrete numbers.
Contract bridge is a game characterized by incomplete information, posing an exciting challenge for artificial intelligence methods. This paper proposes the BridgeHand2Vec approach, which leverages a neural network to embed a bridge player's hand (consisting of 13 cards) into a vector space. The resulting representation reflects the strength of the hand in the game and enables interpretable distances to be determined between different hands. This representation is derived by training a neural network to estimate the number of tricks that a pair of players can take. In the remainder of this paper, we analyze the properties of the resulting vector space and provide examples of its application in reinforcement learning, and opening bid classification. Although this was not our main goal, the neural network used for the vectorization achieves SOTA results on the DDBP2 problem (estimating the number of tricks for two given hands).