Neuro-Evolution in Multi-Player Pente
نویسنده
چکیده
Pente is a derivative of the Japanese game Go-moku, both of which are normally played with only two players. We extend the game of Pente to three players and study the ability of neuro-evolution via the Enforced Sub-Populations (ESP) algorithm to evolve Pente players for 7 by 7 boards capable of beating pairs of opponents taken from a set of five simple handcoded opponents. We also compare the performance of feed forward networks to that of simple recurrent networks and simple recurrent networks that pay attention to the board by reading inputs from it on every player’s turn, not just their own. Evolving networks that beat all pairs of opponents in threeplayer games proves difficult, and we also find that against the given opponents feed forward networks are superior to either type of simple recurrent network.
منابع مشابه
Neuro-Evolution Through Augmenting Topologies Applied To Evolving Neural Networks To Play Othello
Many different approaches to game playing have been suggested including alpha-beta search, temporal difference learning, genetic algorithms, and coevolution. Here, a powerful new algorithm for neuroevolution, Neuro-Evolution for Augmenting Topologies (NEAT), is adapted to the game playing domain. Evolution and coevolution were used to try and develop neural networks capable of defeating an alph...
متن کاملTexture Evolution in Low Carbon Steel Fabricated by Multi-directional Forging of the Martensite Starting Structuree
It has been clarified that deformation and annealing of martensite starting structure can produce ultrafine-grained structure in low carbon steel. This study aims to investigate the texture evolution and mechanical properties of samples with martensite structure deformed by two different forging processes. The martensitic steel samples were forged by plane strain compression and multi-directio...
متن کاملEvolving Opponents for Interesting Interactive Computer Games
In this paper we introduce experiments on neuro-evolution mechanisms applied to predator/prey multi-character computer games. Our test-bed is a modified version of the well-known Pac-Man game. By viewing the game from the predators’ (i.e. opponents’) perspective, we attempt off-line to evolve neural-controlled opponents capable of playing effectively against computer-guided fixed strategy playe...
متن کاملModified Pareto archived evolution strategy for the multi-skill project scheduling problem with generalized precedence relations
In this research, we study the multi-skill resource-constrained project scheduling problem, where there are generalized precedence relations between project activities. Workforces are able to perform one or several skills, and their efficiency improves by repeating their skills. For this problem, a mathematical formulation has been proposed that aims to optimize project completion time, reworki...
متن کاملNeuro-Evolution for Multi-Agent Policy Transfer in RoboCup Keep-Away
An objective of transfer learning is to improve and speedup learning on target tasks after training on a different, but related source tasks. This research is a study of comparative Neuro-Evolution (NE) methods for transferring evolved multi-agent policies (behaviors) between multi-agent tasks of varying complexity. The efficacy of five variants of two NE methods are compared for multi-agent po...
متن کامل