T. Cazenave, Yen-Chi Chen, Guanting Chen, Shi-Yu Chen, Xian-Dong Chiu, J. Dehos, Maria Elsa, Qucheng Gong, Hengyuan Hu, Vasil Khalidov, Cheng-Ling Li, Hsin-I Lin, Yu-Jin Lin, Xavier Martinet, Vegard Mella, J. Rapin, Baptiste Rozière, Gabriel Synnaeve, F. Teytaud, O. Teytaud, Shi-Cheng Ye, Yi-Jun Ye, Shi-Jim Yen, Sergey Zagoruyko
Since DeepMind’s AlphaZero, Zero learning quickly became the state-of-the-art method for many board games. It can be improved using a fully convolutional structure (no fully connected layer). Using such an architecture plus global pooling, we can create bots independent of the board size. The training can be made more robust by keeping track of the best checkpoints during the training and by training against them. Using these features, we release Polygames, our framework for Zero learning, with its library of games and its checkpoints. We won against strong humans at the game of Hex in 19 × 19, including the human player with the best ELO rank on LittleGolem; we incidentally also won against another Zero implementation, which was weaker than humans: in a discussion on LittleGolem, Hex19 was said to be intractable for zero learning. We also won in Havannah with size 8: win against the strongest player, namely Eobllor, with excellent opening moves. We also won several first places at the TAAI 2019 competitions and had positive results against strong bots in various games.
{"title":"Polygames: Improved Zero Learning","authors":"T. Cazenave, Yen-Chi Chen, Guanting Chen, Shi-Yu Chen, Xian-Dong Chiu, J. Dehos, Maria Elsa, Qucheng Gong, Hengyuan Hu, Vasil Khalidov, Cheng-Ling Li, Hsin-I Lin, Yu-Jin Lin, Xavier Martinet, Vegard Mella, J. Rapin, Baptiste Rozière, Gabriel Synnaeve, F. Teytaud, O. Teytaud, Shi-Cheng Ye, Yi-Jun Ye, Shi-Jim Yen, Sergey Zagoruyko","doi":"10.3233/ICG-200157","DOIUrl":"https://doi.org/10.3233/ICG-200157","url":null,"abstract":"Since DeepMind’s AlphaZero, Zero learning quickly became the state-of-the-art method for many board games. It can be improved using a fully convolutional structure (no fully connected layer). Using such an architecture plus global pooling, we can create bots independent of the board size. The training can be made more robust by keeping track of the best checkpoints during the training and by training against them. Using these features, we release Polygames, our framework for Zero learning, with its library of games and its checkpoints. We won against strong humans at the game of Hex in 19 × 19, including the human player with the best ELO rank on LittleGolem; we incidentally also won against another Zero implementation, which was weaker than humans: in a discussion on LittleGolem, Hex19 was said to be intractable for zero learning. We also won in Havannah with size 8: win against the strongest player, namely Eobllor, with excellent opening moves. We also won several first places at the TAAI 2019 competitions and had positive results against strong bots in various games.","PeriodicalId":14829,"journal":{"name":"J. Int. Comput. Games Assoc.","volume":"111 3S 1","pages":"244-256"},"PeriodicalIF":0.0,"publicationDate":"2020-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80728922","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"TCEC Cup 5","authors":"G. Haworth, Nelson Hernandez","doi":"10.3233/icg-200150","DOIUrl":"https://doi.org/10.3233/icg-200150","url":null,"abstract":"","PeriodicalId":14829,"journal":{"name":"J. Int. Comput. Games Assoc.","volume":"26 1","pages":"207-210"},"PeriodicalIF":0.0,"publicationDate":"2020-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74761762","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Recent trends in mathematics education emphasize discovery learning over drill. This has proven to be a bad idea in some cases, for the simple reason that practice is required to learn basic arithmetic skills. Drills in arithmetic skills can be made interesting through gamification. This study proposes a family of puzzles that gamify arithmetic practice. The puzzles are designed with an evolutionary algorithm forming an instance of automatic content generation. Two methods of evolutionary puzzle design are presented and discussed. The first method used transformed the problem into an almost trivial optimization. The second algorithm was designed to avoid the flaws of the first and produced a huge variety of puzzles. A hardness measure, based on the difficulty experienced by the evolutionary puzzle generator, is employed. The hardness measure is tested on a large collection of puzzles produced with the evolutionary automatic content generation system. An initial assumption, that all the pieces in the puzzle must be used to achieve a maximum score, was shown to be incorrect in puzzles located via automatic search. Two classes of puzzle are defined: those where the optimal solution uses all pieces and those where the optimal solution fails to use at least one piece. The latter sort of puzzle were found to be far more common in the search space.
{"title":"A polyomino puzzle for arithmetic practice","authors":"Jérémy Foxcroft, D. Ashlock","doi":"10.3233/icg-200170","DOIUrl":"https://doi.org/10.3233/icg-200170","url":null,"abstract":"Recent trends in mathematics education emphasize discovery learning over drill. This has proven to be a bad idea in some cases, for the simple reason that practice is required to learn basic arithmetic skills. Drills in arithmetic skills can be made interesting through gamification. This study proposes a family of puzzles that gamify arithmetic practice. The puzzles are designed with an evolutionary algorithm forming an instance of automatic content generation. Two methods of evolutionary puzzle design are presented and discussed. The first method used transformed the problem into an almost trivial optimization. The second algorithm was designed to avoid the flaws of the first and produced a huge variety of puzzles. A hardness measure, based on the difficulty experienced by the evolutionary puzzle generator, is employed. The hardness measure is tested on a large collection of puzzles produced with the evolutionary automatic content generation system. An initial assumption, that all the pieces in the puzzle must be used to achieve a maximum score, was shown to be incorrect in puzzles located via automatic search. Two classes of puzzle are defined: those where the optimal solution uses all pieces and those where the optimal solution fails to use at least one piece. The latter sort of puzzle were found to be far more common in the search space.","PeriodicalId":14829,"journal":{"name":"J. Int. Comput. Games Assoc.","volume":"81 1","pages":"272-286"},"PeriodicalIF":0.0,"publicationDate":"2020-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73381348","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This contribution reports about an almost forgotten attempt to find themes in chess composition automatically by machine. Based on investigations of the late International Master of Chess Composition Dieter Müller in the 1970s, the writers describe some of his results. They hope to fill a gap in the great history of chess programming.
{"title":"A program for analyzing two move chess problems","authors":"D. Müller","doi":"10.3233/icg-200162","DOIUrl":"https://doi.org/10.3233/icg-200162","url":null,"abstract":"This contribution reports about an almost forgotten attempt to find themes in chess composition automatically by machine. Based on investigations of the late International Master of Chess Composition Dieter Müller in the 1970s, the writers describe some of his results. They hope to fill a gap in the great history of chess programming.","PeriodicalId":14829,"journal":{"name":"J. Int. Comput. Games Assoc.","volume":"6 1","pages":"171-177"},"PeriodicalIF":0.0,"publicationDate":"2020-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86083755","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
It is been 46 years since the first computer chess world championship was held in Stockholm in 1974, and there have now been 24 more with over 700 games played. This paper investigates a number of questions that a curious soul might ask regarding the results of these games, especially how they ended. The data provided herein attempts to answer several related questions. In addition, it presents data showing how ratings of the top engines have increased over the years.
{"title":"An analysis of games played in world computer chess championships over the years","authors":"M. Newborn","doi":"10.3233/icg-200148","DOIUrl":"https://doi.org/10.3233/icg-200148","url":null,"abstract":"It is been 46 years since the first computer chess world championship was held in Stockholm in 1974, and there have now been 24 more with over 700 games played. This paper investigates a number of questions that a curious soul might ask regarding the results of these games, especially how they ended. The data provided herein attempts to answer several related questions. In addition, it presents data showing how ratings of the top engines have increased over the years.","PeriodicalId":14829,"journal":{"name":"J. Int. Comput. Games Assoc.","volume":"13 1","pages":"99-105"},"PeriodicalIF":0.0,"publicationDate":"2020-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89386033","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
On August 31, 1970, an experiment began that continues to this day. The first chess tournament for computers was held as part of the Association for Computing Machinery’s (ACM’s) National Conference. The interest generated was tremendous, leading to ACM sponsoring an annual event until 1994. Chess competitions continue to this day, allowing for 50 years of data on the growth of artificial intelligence capabilities in this domain. During this period, program ratings have soared from roughly 1400 in 1970 to over 3500 today. The 1970 event was the first continuous competition in computer science history, and it represents the longest ongoing experiment in computer science history.1
{"title":"The 1970 United States computer chess championship: The start of the longest-running experiment in computer science history","authors":"J. Schaeffer","doi":"10.3233/icg-200149","DOIUrl":"https://doi.org/10.3233/icg-200149","url":null,"abstract":"On August 31, 1970, an experiment began that continues to this day. The first chess tournament for computers was held as part of the Association for Computing Machinery’s (ACM’s) National Conference. The interest generated was tremendous, leading to ACM sponsoring an annual event until 1994. Chess competitions continue to this day, allowing for 50 years of data on the growth of artificial intelligence capabilities in this domain. During this period, program ratings have soared from roughly 1400 in 1970 to over 3500 today. The 1970 event was the first continuous competition in computer science history, and it represents the longest ongoing experiment in computer science history.1","PeriodicalId":14829,"journal":{"name":"J. Int. Comput. Games Assoc.","volume":"51 1","pages":"72-85"},"PeriodicalIF":0.0,"publicationDate":"2020-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73162782","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"The SSDF Rating List, 2020-07, in memoriam Tony Hedlund","authors":"Lars Sandin, G. Haworth","doi":"10.3233/icg-200155","DOIUrl":"https://doi.org/10.3233/icg-200155","url":null,"abstract":"","PeriodicalId":14829,"journal":{"name":"J. Int. Comput. Games Assoc.","volume":"4 1","pages":"242"},"PeriodicalIF":0.0,"publicationDate":"2020-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87207814","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：