T. Bossomaier, A. Knittel, and A. Snyder (Australia)
games, reinforcement learning, graphs, artificial intelli gence
This paper studies scalability of collective intelligence ap plied to game play of combinatorially difficult games, in particular the game of Dots and Boxes. Using a graph iso morphism of game states, we represent moves as graph rewriting operations. We investigate scaling with game size, number of agents and the use of move graphs of low Kolmogorov complexity.
Important Links:
Go Back
