Technique analysis and designing of program with UCT algorithm for NoGo

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Citation (Scopus)

Abstract

As a typical example of dynamic search algorithm, the UCT algorithm was initially used on the computerized game of GO. This paper briefly introduces the Markov Decision process, the Multi-armed Bandit model, and the Upper-Confidence Bandit formula. It analyzes the source and structure of the UCT algorithm in theory, and proves that the UCT algorithm is suitable for the design of the program of NoGo. According to the characteristics of NoGo, in the paper we improved the algorithm in terms of move generation and data reuse. We also tried to establish an off-line knowledge database for research. With experimental data we have tested and evaluated the above methods. The above algorithm and technology have been successfully used in WTShadows the NoGo game program, which enabled us to have won the champion in national competition.

Original languageEnglish
Title of host publication2013 25th Chinese Control and Decision Conference, CCDC 2013
Pages923-928
Number of pages6
DOIs
Publication statusPublished - 2013
Event2013 25th Chinese Control and Decision Conference, CCDC 2013 - Guiyang, China
Duration: 25 May 201327 May 2013

Publication series

Name2013 25th Chinese Control and Decision Conference, CCDC 2013

Conference

Conference2013 25th Chinese Control and Decision Conference, CCDC 2013
Country/TerritoryChina
CityGuiyang
Period25/05/1327/05/13

Keywords

  • Dynamic Move Queue
  • Knowledge Base
  • MAB Model
  • Markov Decision Process
  • NoGo
  • UCT Algorithm

Fingerprint

Dive into the research topics of 'Technique analysis and designing of program with UCT algorithm for NoGo'. Together they form a unique fingerprint.

Cite this