PROJECT TITLE :

N-Grams and the Last-Good-Reply Policy Applied in General Game Playing

ABSTRACT:

The aim of general game taking part in (GGP) is to create programs capable of taking part in a big selection of various games at an knowledgeable level, given solely the rules of the sport. The foremost successful GGP programs currently use simulation-based Monte Carlo tree search (MCTS). The performance of MCTS depends heavily on the simulation strategy used. In this paper, we have a tendency to introduce improved simulation ways for GGP that we implement and take a look at in the GGP agent CadiaPlayer, which won the International GGP competition in both 2007 and 2008. There are two aspects to the improvements: initial, we show that a simple $epsilon$-greedy exploration strategy works better within the simulation play-outs than the softmax-based mostly Gibbs measure currently used in CadiaPlayer and, second, we tend to introduce a general framework based on N-grams for learning promising move sequences. Collectively, these enhancements lead to a a lot of improved performance of CadiaPlayer. For instance, in our take a look at suite consisting of 5 totally different two-player turn-primarily based games, they led to an impressive average win rate of roughly 70%. The enhancements are also shown to be effective in multiplayer and simultaneous-move games. We have a tendency to additionally perform experiments with the last-good-reply policy (LGRP). The LGRP combined with N-grams is additionally tested. The LGRP has already been shown to achieve success in Go programs and we tend to demonstrate that it also has promise in GGP.