Dynamic randomization and domain knowledge in Monte-Carlo Tree Search for Go knowledge-based systems

Chen, Keh-Hsun

doi:10.1016/j.knosys.2011.08.007

Cited by 11 publications

(13 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Dynamic randomization was used in [24]. It can increase the playing strength of a Go Intellect significantly while the simulations are beyond 128K per move.…”

Section: Related Workmentioning

confidence: 99%

A Middle Game Search Algorithm Applicable to Low-Cost Personal Computer for Go

Wang

et al. 2019

IEEE Access

View full text Add to dashboard Cite

Go Artificial Intellects(AIs) using deep reinforcement learning and neural networks have achieved superhuman performance, but they rely on powerful computing resources. They are not applicable to low-cost personal computer(PC). In our life, most entertainment programs of Go run on the general PC. A human Go master consider different strategies for different stages, especially for the middle stage that has a significant impact on winning or losing. To study arguably a more humanlike approach that is applicable to low-cost PC while not reducing chess power, this paper proposes a new search algorithm based on hypothesis testing and dynamic randomization for the middle stage of the game Go. Firstly, a new method to decide the intervals of different playing stages more reasonable based on hypothesis testing is proposed. Secondly, a new search algorithm including a layered pruning branch method, a comprehensive evaluation function and a new selecting node method is proposed. The pruning method based on domain knowledge and upper confidence bound formula(UCB) are all applied to subtract the branches from the lower evaluation score, which was ranked behind 20%. The comprehensive evaluation function with adjustable parameters is proposed to evaluate the tree nodes after pruning. The new selecting node method based on dynamic randomization is used to expand the tree by selecting a node randomly from the high-quality node interval. Finally, the experimental results show that the designed algorithm outperforms Gnugo3.6 and Gnugo3.8 in chess power while reducing average search time and average RAM cost for one move effectively on a 19×19 board.

show abstract

“…Dynamic randomization was used in [24]. It can increase the playing strength of a Go Intellect significantly while the simulations are beyond 128K per move.…”

Section: Related Workmentioning

confidence: 99%

A Middle Game Search Algorithm Applicable to Low-Cost Personal Computer for Go

Wang

et al. 2019

IEEE Access

View full text Add to dashboard Cite

show abstract

“…During the statistical estimation of the values which consists directly of many variable factors, the universal Monte Carlo algorithm is very useful. It is used in the modelling of knowledge and is related to performance evaluation in complex systems under uncertainty [25] for financial analysis [26] or estimates with dynamically changing knowledge [27].…”

Section: Related Workmentioning

confidence: 99%

An Approach to Tacit Knowledge Classification in a Manufacturing Company

Patalas‐Maliszewska

Śliwa

2019

Teh. vjesn.

View full text Add to dashboard Cite

This article attempts to classify the tacit knowledge acquired using the example of the research and development (R&D) department of a manufacturing company. Based on studies in the literature and direct interviews in the company analysed, the authors' own model for classifying the tacit knowledge in an R&D department was proposed. The description of this model has been divided into two parts. In the first part, viz., the classification of knowledge, through three planes: (1) selection of algorithm inputs and the grouping of knowledge accumulated in an enterprise; (2) algorithm activity, that is, the use of algorithms based on clustering them, for calculations; (3) interpretation of results. The Bayesian Network was used for this purpose, which was modelled on the defined relationships between representing of tacit knowledge. Then, on the basis of the case study, the classification of knowledge was prepared according to: (1) definition of knowledge in the R&D department and its modelling, (2) implementation of a suitable number of training sets, (3) verification of the knowledge base, that is, declaration of the value of the knowledge observed, followed by (4) assignment of the probability of returning to the node of network clusters containing interpretations of business benefits.

show abstract

“…For Connect-k games, the priority of candidate move depends on number of threats by the move. Monte-Carlo Tree Search (MCTS) is a best-first search, which uses stochastic simulations [1,[6][7][8]15,25]. It has advanced the development of computer Go substantially [6,16,40].…”

Section: Multistage Search and Pns Implementationmentioning

confidence: 99%

“…Since Wu [30] developed k-in-a-row or Connect-k games in 2005, Connect (19,19,6,2,1) or Connect6, which is derived from Gomoku, has been a popular research topic [30][31][32]38]. In Connect (m, n, k, p, q), two players alternately place p stones on an m Â n board for each move except for that the first player places q stones for the first move.…”

Section: Introductionmentioning

confidence: 99%