Unlike chess and go, poker players don't have full knowledge of the game state. [Read More]
Summary… * Machine-Made
At the end of January, Carnegie Mellon computer scientists achieved a major milestone: their algorithm, Libratus, beat a set of professional poker players in a 120,000-hand tournament.
Imperfect strategiesA perfect information game is relatively simple: all players can know the full state of the game, often just by looking at the board.
To handle imperfect information games, past work has focused on approaches derived from game theory.
The networks were trained on 10 million randomly drawn poker games.
Still, the DeepStack approach appears to be more general, since it doesn't rely on having a supercomputer at your disposal to update the system during breaks.
Opinion… * Man-Made