25,161
edits
Changes
no edit summary
=Hex Programs=
After early trials with [https://www.game-ai-forum.org/icga-tournaments/program.php?id=276 Mongoose], the Hex programs [https://www.game-ai-forum.org/icga-tournaments/program.php?id=135 Wolve] ([[13th Computer Olympiad#Hex|2008]]) and [https://www.game-ai-forum.org/icga-tournaments/program.php?id=555 MoHex] ([[14th Computer Olympiad#Hex|2009]], [[15th Computer Olympiad#Hex|2010]], [[16th Computer Olympiad#Hex|2011]], [[17th Computer Olympiad#Hex|2013]], [[18th Computer Olympiad#Hex|2015]] and [[20th Computer Olympiad#Hex|2017]]) won Gold Medals in Hex at the [[Computer Olympiad]].
<span id="Wolve"></span>
==Wolve==
Wolve does a truncated [[Alpha-Beta]] search of two and up to four [[Ply|plies]], considering the huge [[Branching Factor]] of Hex.
<span id="MoHex"></span>
==MoHex==
Since 2009 [[Monte-Carlo Tree Search]] starts to dominate, and MoHex applies MCTS along with the [[UCT]] framework combined with the allmoves-as-first (AMAF) heuristic to select the best child during tree traversal <ref>[[Broderick Arneson]], [[Ryan Hayward]], [[Philip Henderson]] ('''2010'''). ''Monte Carlo Tree Search in Hex''. [[IEEE#TOCIAIGAMES|IEEE Transactions on Computational Intelligence and AI in Games]], Vol. 2, No. 4, [http://webdocs.cs.ualberta.ca/~hayward/papers/mcts-hex.pdf pdf]</ref>.
MoHex-CNN, which won the [[20th Computer Olympiad#Hex|13x13 competition of the 20th Computer Olympiad 2017]] is a [[Neural Networks#Convolutional|convolutional neural net]] (CNN) version of MoHex. At each new node of the Monte-Carlo search tree, a policy CNN biases child selection by initializing child visit and win counts with artificial values <ref>* [[Ryan Hayward]], [[Noah Weninger]] ('''2017'''). ''Hex 2017: MoHex wins the 11x11 and 13x13 tournaments''. [[ICGA Journal#39_34|ICGA Journal, Vol. 39, Nos. 3-4]]</ref>.
=Selected Publications=
