Creating Action Heuristics for General Game Playing Agents

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26230%2F16%3APU136052" target="_blank" >RIV/00216305:26230/16:PU136052 - isvavai.cz</a>

Výsledek na webu

<a href="https://link.springer.com/chapter/10.1007/978-3-319-39402-2_11" target="_blank" >https://link.springer.com/chapter/10.1007/978-3-319-39402-2_11</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/978-3-319-39402-2_11" target="_blank" >10.1007/978-3-319-39402-2_11</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Creating Action Heuristics for General Game Playing Agents

Popis výsledku v původním jazyce

Monte-Carlo Tree Search (MCTS) is the most popular search algorithm used in General Game Playing (GGP) nowadays mainly because of its ability to perform well in the absence of domain knowledge. Several approaches have been proposed to add heuristics to MCTS in order to guide the simulations. In GGP those approaches typically learn heuristics at runtime from the results of the simulations. Because of peculiarities of GGP, it is preferable that these heuristics evaluate actions rather than game positions. We propose an approach that generates heuristics that estimate the usefulness of actions by analyzing the game rules as opposed to the simulation results. We present results of experiments that show the potential of our approach.

Název v anglickém jazyce

Creating Action Heuristics for General Game Playing Agents

Popis výsledku anglicky

Monte-Carlo Tree Search (MCTS) is the most popular search algorithm used in General Game Playing (GGP) nowadays mainly because of its ability to perform well in the absence of domain knowledge. Several approaches have been proposed to add heuristics to MCTS in order to guide the simulations. In GGP those approaches typically learn heuristics at runtime from the results of the simulations. Because of peculiarities of GGP, it is preferable that these heuristics evaluate actions rather than game positions. We propose an approach that generates heuristics that estimate the usefulness of actions by analyzing the game rules as opposed to the simulation results. We present results of experiments that show the potential of our approach.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2016

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Computer Games, CGW 2015

ISBN

—

ISSN

1865-0929

e-ISSN

—

Počet stran výsledku

16

Strana od-do

149-164

Název nakladatele

Springer Verlag

Místo vydání

Berlín

Místo konání akce

Buenos Aires, Argentina

Datum konání akce

26. 6. 2015

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

000385788000011