Browsing by Author "Wu, Ti-Rong"

Now showing 1 - 3 of 3

Multilabeled Value Networks for Computer Go
(2018-12-01)
Wu, Ti-Rong
;
Wu, I-Chen
;
Chen, Guan-Wun
;
Wei, Ting-Han
;
Wu, Hung-Chun
This paper proposes a new approach to a novel value network architecture for the game Go, called a multilabeled (ML) value network. In the ML value network, different values (win rates) are trained simultaneously for different settings of komi, a compensation given to balance the initiative of playing first. The ML value network has three advantages: 1) it outputs values for different komi; (2) it supports dynamic komi; and (3) it lowers the mean squared error (MSE). This paper also proposes a new dynamic komi method to improve game-playing strength. This paper also performs experiments to demonstrate the merits of the architecture. First, the MSE of the ML value network is generally lower than the value network alone. Second, the program based on the ML value network wins by a rate of 67.6% against the program based on the value network alone. Third, the program with the proposed dynamic komi method significantly improves the playing strength over the baseline that does not use dynamic komi, especially for handicap games. To our knowledge, up to date, no handicap games have been played openly by programs using value networks. This paper provides these programs with a useful approach to playing handicap games.
On Strength Adjustment for MCTS-Based Programs
(2019-01-01)
Wu, I-Chen
;
Wu, Ti-Rong
;
Liu, An-Jen
;
Guei, Hung
;
Wei, Tinghan
This paper proposes an approach to strength adjustment for MCTS-based game-playing programs. In this approach, we use a softmax policy with a strength index to choose moves. Most importantly, we filter low quality moves by excluding those that have a lower simulation count than a pre-defined threshold ratio of the maximum simulation count. We perform a theoretical analysis, reaching the result that the adjusted policy is guaranteed to choose moves exceeding a lower bound in strength by using a threshold ratio. The approach is applied to the Go program ELF OpenGo. The experiment results show that is highly correlated to the empirical strength; namely, given a threshold ratio 0.1, z is linearly related to the Elo rating with regression error 47.95 Elo where -2 <= z <= 2. Meanwhile, the covered strength range is about 800 Elo ratings in the interval of in. With the ease of strength adjustment using, we present two methods to adjust strength and predict opponents' strengths dynamically. To our knowledge, this result is state-of-the-art in terms of the range of strengths in Elo rating while maintaining a controllable relationship between the strength and a strength index.
Strength Adjustment and Assessment for MCTS-Based Programs [Research Frontier]
(2020-08-01)
Liu, An-Jen
;
Wu, Ti-Rong
;
Wu, I-Chen
;
Guei, Hung
;
Wei, Ting-Han
This paper proposes an approach to strength adjustment and assessment for Monte-Carlo tree search based game-playing programs. We modify an existing softmax policy with a strength index to choose moves. The most important modification is a mechanism which filters low-quality moves by excluding those that have a lower simulation count than a pre-defined threshold ratio of the maximum simulation count. Through theoretical analysis, we show that the adjusted policy is guaranteed to choose moves exceeding a lower bound in strength by using a threshold ratio. Experimental results show that the strength index is highly correlated to the empirical strength. With an index value between ?2, we can cover a strength range of about 800 Elo ratings. The strength adjustment and assessment methods were also tested in real-world scenarios with human players, ranging from professionals (strongest) to kyu rank amateurs (weakest). For amateur levels, we tested our mechanism on two popular Go online platforms - Fox Weiqi and Tygem. The result shows that our method can adjust program strength to different ranks stably. In terms of strength assessment, we proposed a new dynamic strength adjustment method, then used it to evaluate human professionals, predicting reliably their playing strengths within 15 games. Lastly, we collected survey responses asking players about strength perception, entertainment, and general comments for different aspects of analysis. To our best knowledge, this result is state-ofthe- art in terms of the range of strengths in Elo rating while maintaining a controllable relationship between the strength and a strength index.