• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2016.10a
• /
• pp.405-408
• /
• 2016

Games like Go, Chess, Janggi have helped to brain development of the people. These games are developed by computer program. And many algorithms have been developed to allow myself to play. The person winning chess program was developed in the 1990s. But game of go is too large number of cases. So it was considered impossible to win professional go player. However, with the use of MCTS(Monte Carlo Tree Search) and CNN(Convolutional Neural Network), the performance of the go algorithm is greatly improved. In this paper, using CNN and MCTS were proceeding development of go algorithm. Using the manual of go learning CNN look for the best position, MCTS calculates the win probability in the game to proceed with simulation. In addition, extract pattern information of go using existing manual of go, plans to improve speed and performance by using it. This method is showed a better performance than general go algorithm. Also if it is receiving sufficient computing power, it seems to be even more improved performance.

• Smart Media Journal
• /
• v.9 no.4
• /
• pp.36-43
• /
• 2020

The Monte Carlo tree search (MCTS) is a popular method for implementing an intelligent game program. It has several hyper-parameters that require an optimization for showing the best performance. Due to the stochastic nature of the MCTS, the hyper-parameter optimization is difficult to solve. This paper uses the self-playing capability of the MCTS-based game program for optimizing the hyper-parameters. It seeks a winner path over the hyper-parameter space while performing the self-play. The top-q longest winners in the winner path compete for the final winner. The experiment using the 15-15-5 game (Omok in Korean name) showed a promising result.

• Journal of Korea Game Society
• /
• v.18 no.5
• /
• pp.77-82
• /
• 2018

Go is the most complex board game in which the computer can not search all possible moves using an exhaustive search to find the best one. Prior to AlphaGo, all powerful computer Go programs have used the Monte-Carlo Tree Search (MCTS) to overcome the difficulty in positional evaluation and the very large branching factor in a game tree. In this paper, we tried to find the best sequence of moves using an MCTS on a very small Go board. We found that a $2{\times}2$ Go game would be ended in a tie and the size of Komi should be 0 point; Meanwhile, in a $3{\times}3$ Go Black can always win the game and the size of Komi should be 9 points.

• Journal of Korea Game Society
• /
• v.18 no.6
• /
• pp.59-68
• /
• 2018

In spite of its simple rule, Go is one of the most complex strategic board games in the field of Artificial Intelligence (AI). Monte-Carlo Tree Search (MCTS) is an algorithm with best-first tree search, and has used to implement computer Go. We try to find the most promising first move using MCTS for playing a Go game on a board of size smaller than $9{\times}9$ Go board. The experimental result reveals that MCTS prefers to place the first move at the center in case of odd-sized Go boards, and at the central in case of even-sized Go boards.

• KIISE Transactions on Computing Practices
• /
• v.21 no.8
• /
• pp.572-577
• /
• 2015

Generally, the existing Artificial Intelligence (AI) systems were designed for specific purposes and their capabilities handle only specific problems. Alternatively, Artificial General Intelligence can solve new problems as well as those that are already known. Recently, General Video Game Playing the game AI version of General Artificial Intelligence, has garnered a large amount of interest among Game Artificial Intelligence communities. Although video games are the sole concern, the design of a single AI that is capable of playing various video games is not an easy process. In this paper, we propose a GreedyUCB1 algorithm and rollout method that were formulated using the knowledge from a game analysis for the Monte-Carlo Tree Search game AI. An AI that used our method was ranked fourth at the GVG-AI (General Video Game-Artificial Intelligence) competition of the IEEE international conference of CIG (Computational Intelligence in Games) 2014.

• Journal of Korea Game Society
• /
• v.14 no.3
• /
• pp.47-54
• /
• 2014

The game of Go is one of the oldest games and originated at least more than 2,500 years ago. In game programming the most successful approach is to use game tree searches using evaluation functions. However it is really difficult to construct feasible evaluation function in computer Go. Monte-Carlo Tree Search(MCTS) has created strong computer Go programs such as MoGo and CrazyStone which defeated human Go professionals played on the $9{\times}9$ board. MCTS is based on the winning rate estimated by Monte-Carlo simulation. Prior to implementing MCTS into computer Go, we tried to measure each winning rate of three positions, center, corner and side, in Tic-Tac-Toe playing as the best first move. The experimental result revealed that the center is the best, a corner the next and a side the last as the best first move.

• Journal of Korea Game Society
• /
• v.16 no.6
• /
• pp.79-90
• /
• 2016

Recently, many researchers have paid attention to the improved generation of AI agent in the area of game industry. Monte-Carlo Tree Search(MCTS) is one of the algorithms to search an optimal solution through random search with perfect information, and it is suitable for the purpose of calculating an approximate value to the solution of an equation which cannot be expressed explicitly. Games in Trading Card Game(TCG) genre such as the heartstone has imperfect information because the cards and play of an opponent are not predictable. In this study, MCTS is suggested in imperfect information card games so as to generate AI agents. In addition, the practicality of MCTS algorithm is verified by applying to heartstone game which is currently used.

• Journal of Korea Game Society
• /
• v.16 no.4
• /
• pp.79-86
• /
• 2016

Monte-Carlo Tree Search(MCTS) is a best-first tree search algorithm and has been successfully applied to various games, especially to the game of Go. We evaluate the performance of MCTS playing against each other in the game of Tic-Tac-Toe. It reveals that the first player always has an overwhelming advantage to the second player; and we try to find out the reason why the first player is superior to the second player in spite of the fact that the best game result should be a draw. Since MCTS is a statistical algorithm based on the repeated random sampling, it cannot adequately tackle an urgent problem that needs a strategy, especially for the second player. For this, we propose a strategic MCTS(S-MCTS) and show that the S-MCTS player never loses a Tic-Tac-Toe game.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.17 no.5
• /
• pp.119-124
• /
• 2017

In 1997, IBM's DeepBlue won the world chess championship, Garry Kasparov, and recently, Google's AlphaGo won all three games against Ke Jie, who was ranked 1st among all human Baduk players worldwide, interest in deep running has increased rapidly. DeepPurple, proposed in this paper, is a AI chess engine based on deep learning. DeepPurple Chess Engine consists largely of Monte Carlo Tree Search and policy network and value network, which are implemented by convolution neural networks. Through the policy network, the next move is predicted and the given situation is calculated through the value network. To select the most beneficial next move Monte Carlo Tree Search is used. The results show that the accuracy and the loss function cost of the policy network is 43% and 1.9. In the case of the value network, the accuracy is 50% and the loss function cost is 1, respectively.

• Journal of Information Technology Services
• /
• v.16 no.4
• /
• pp.137-148
• /
• 2017

Researches on the Artificial Intelligence has been explosively activated in various fields since the advent of AlphaGo. Particularly, researchers on the application of multi-layer neural network such as deep learning, and various machine learning algorithms are being focused actively. In this paper, we described a development of an artificial intelligence Janggi game based on reinforcement learning algorithm and MCTS (Monte Carlo Tree Search) algorithm with accumulated game data. The previous artificial intelligence games are mostly developed based on mini-max algorithm, which depends only on the results of the tree search algorithms. They cannot use of the real data from the games experts, nor cannot enhance the performance by learning. In this paper, we suggest our approach to overcome those limitations as follows. First, we collects Janggi expert's game data, which can reflect abundant real game results. Second, we create a graph structure by using the game data, which can remove redundant movement. And third, we apply the reinforcement learning algorithm and MCTS algorithm to select the best next move. In addition, the learned graph is stored by object serialization method to provide continuity of the game. The experiment of this study is done with two different types as follows. First, our system is confronted with other AI based system that is currently being served on the internet. Second, our system confronted with some Janggi experts who have winning records of more than 50%. Experimental results show that the rate of our system is significantly higher.