Fast and Knowledge-Free Deep Learning for General Game Playing (Student Abstract)
This work addresses the problem of efficient and knowledge-free deep learning for general game playing, but it is incremental as it builds on AlphaZero with modifications for speed and reduced knowledge requirements.
The authors tackled the challenge of adapting AlphaZero to General Game Playing by developing a method that uses MCTS playing instead of self-play, a value network only, and attention layers, enabling faster model generation without assumptions about action space or board topology. They implemented this in the Regular Boardgames GGP system and showed it outperforms the UCT baseline for most games efficiently.
We develop a method of adapting the AlphaZero model to General Game Playing (GGP) that focuses on faster model generation and requires less knowledge to be extracted from the game rules. The dataset generation uses MCTS playing instead of self-play; only the value network is used, and attention layers replace the convolutional ones. This allows us to abandon any assumptions about the action space and board topology. We implement the method within the Regular Boardgames GGP system and show that we can build models outperforming the UCT baseline for most games efficiently.