Difference between revisions of "Yu Nasu"

From Chessprogramming wiki
Jump to: navigation, search
 
(10 intermediate revisions by the same user not shown)
Line 8: Line 8:
 
[https://en.wikipedia.org/wiki/Speech_recognition speech recognition] and [https://en.wikipedia.org/wiki/Speech_synthesis speech synthesis] as application of [https://en.wikipedia.org/wiki/Hidden_Markov_model hidden Markov model] and [[Neural Networks|neural networks]] <ref>[https://dblp.org/pers/hd/o/Ohtani:Yamato Yamato Ohtani], [[Yu Nasu]], [https://dblp.org/pers/hd/m/Morita:Masahiro Masahiro Morita], [https://dblp.org/pers/hd/a/Akamine:Masami Masami Akamine] ('''2015'''). ''[https://www.semanticscholar.org/paper/Emotional-transplant-in-statistical-speech-based-on-Ohtani-Nasu/5d6c1846bef6e8f796a84f6d7a381fdcffe1e1dd Emotional transplant in statistical speech synthesis based on emotion additive model]''. [https://dblp.org/db/conf/interspeech/interspeech2015.html#OhtaniNMA15 INTERSPEECH 2015]</ref>.  
 
[https://en.wikipedia.org/wiki/Speech_recognition speech recognition] and [https://en.wikipedia.org/wiki/Speech_synthesis speech synthesis] as application of [https://en.wikipedia.org/wiki/Hidden_Markov_model hidden Markov model] and [[Neural Networks|neural networks]] <ref>[https://dblp.org/pers/hd/o/Ohtani:Yamato Yamato Ohtani], [[Yu Nasu]], [https://dblp.org/pers/hd/m/Morita:Masahiro Masahiro Morita], [https://dblp.org/pers/hd/a/Akamine:Masami Masami Akamine] ('''2015'''). ''[https://www.semanticscholar.org/paper/Emotional-transplant-in-statistical-speech-based-on-Ohtani-Nasu/5d6c1846bef6e8f796a84f6d7a381fdcffe1e1dd Emotional transplant in statistical speech synthesis based on emotion additive model]''. [https://dblp.org/db/conf/interspeech/interspeech2015.html#OhtaniNMA15 INTERSPEECH 2015]</ref>.  
  
=Shogi=
+
=[[Shogi]]=
 
Along with [[Hisayori Noda]] (Nodchip), [[Jun Okabe]], [[Takahiro Suzuki]], [[Masatoshi Hidaka]], and at times [[Akio Kono]],
 
Along with [[Hisayori Noda]] (Nodchip), [[Jun Okabe]], [[Takahiro Suzuki]], [[Masatoshi Hidaka]], and at times [[Akio Kono]],
 
Yu Nasu was member of ''The Minstrel's Ballad: Tanuki's Reign''  
 
Yu Nasu was member of ''The Minstrel's Ballad: Tanuki's Reign''  
<ref>[https://groups.google.com/d/msg/shogi-l/-fO7GP6Zzww/8pG4UYeNAAAJ WCSC26 participation list] by [[Reijer Grimbergen]], [https://groups.google.com/forum/#!forum/shogi-l SHOGI-L], April 08, 2016</ref>  
+
<ref>[https://groups.google.com/d/msg/shogi-l/-fO7GP6Zzww/8pG4UYeNAAAJ WCSC26 participation list] by [[Reijer Grimbergen]], [[Computer Chess Forums|SHOGI-L]], April 08, 2016</ref>  
<ref>[https://groups.google.com/d/msg/shogi-l/v-tznD0IQe0/37KYlSYtAgAJ The 27th World Computer Shogi Championship: participant list] by [[Takenobu Takizawa]], [https://groups.google.com/forum/#!forum/shogi-l SHOGI-L], April 17, 2017</ref>
+
<ref>[https://groups.google.com/d/msg/shogi-l/v-tznD0IQe0/37KYlSYtAgAJ The 27th World Computer Shogi Championship: participant list] by [[Takenobu Takizawa]], [[Computer Chess Forums|SHOGI-L]], April 17, 2017</ref>
and subsequently is member of the ''Ziosoft Computer Shogi Club'', developing and representing the Shogi playing entities ''Shouten Gensou Knights of Tanuki'' <ref>[http://www2.computer-shogi.org/wcsc27/team.html 第27回世界コンピュータ将棋選手権 参加チーム]</ref>, ''the end of genesis T.N.K.evolution turbo type D''  
+
and subsequently is member of the ''Ziosoft Computer Shogi Club'', developing and representing the Shogi playing entities ''Shouten Gensou Knights of Tanuki'' ([[WCSC27|2017]], derived from ''Yaneura King'' by [[Motohiro Isozaki]] aka Yaneurao)
 +
<ref>[[Hisayori Noda]], [[Jun Okabe]], [[Takahiro Suzuki]], [[Masatoshi Hidaka]], [[Yu Nasu]], [[Akio Kono]] ('''2017'''). ''Shouten Gensou Knights of Tanuki Appeal Document''. The Minstrel's Ballad: Tanuki's Reign, [http://www2.computer-shogi.org/wcsc27/appeal/tanuki-/appeal.pdf pdf] (Japanese)</ref>
 +
<ref>[http://www2.computer-shogi.org/wcsc27/team.html 第27回世界コンピュータ将棋選手権 参加チーム wcsc27]</ref>,  
 +
''the end of genesis T.N.K.evolution turbo type D''  
 
<ref>[http://www2.computer-shogi.org/wcsc28/ 第28回世界コンピュータ将棋選手権 wcsc28]</ref>
 
<ref>[http://www2.computer-shogi.org/wcsc28/ 第28回世界コンピュータ将棋選手権 wcsc28]</ref>
<ref>[https://www.uuunuuun.com/single-post/2019/05/28/Installation-instruction-of-shogi-engines-v2019-May Installation instruction of shogi engines], May 28, 2019</ref>
+
<ref>[https://www.uuunuuun.com/single-post/2019/05/28/Installation-instruction-of-shogi-engines-v2019-May Installation instruction of shogi engines], May 28, 2019</ref> and [[Tanu-King]] at the [[WCSC29]] <ref>[http://www2.computer-shogi.org/wcsc29/ 第29回世界コンピュータ将棋選手権 wcsc29]</ref>.  
and the [[WCSC29|29th World Computer Shogi Champion]] <ref>[http://www2.computer-shogi.org/wcsc29/ 第29回世界コンピュータ将棋選手権 wcsc29]</ref> [[YaneuraOu]] <ref>[http://yaneuraou.yaneu.com/ やねうら王 公式サイト | コンピューター将棋 やねうら王 公式サイト]</ref> <ref>[https://github.com/yaneurao/YaneuraOu GitHub - yaneurao/YaneuraOu: YaneuraOu is the World's Strongest Shogi engine(AI player), WCSC29 1st winner, educational and USI compliant engine]</ref>.
 
  
=NNUE=
+
=[[NNUE]]=
 
In 2018, Yu Nasu introduced a new [[Evaluation Function|evaluation function]] based on a shallow [[Neural Networks|neural network]]
 
In 2018, Yu Nasu introduced a new [[Evaluation Function|evaluation function]] based on a shallow [[Neural Networks|neural network]]
 
whose heavily overparametrized input layer outputs were [[Incremental Updates|incremental updated]] in [[Make Move|make]] and [[Unmake Move|unmake move]].
 
whose heavily overparametrized input layer outputs were [[Incremental Updates|incremental updated]] in [[Make Move|make]] and [[Unmake Move|unmake move]].
 
This approach, dubbed [[NNUE]] (&#398;U&#1048;&#1048; Efficiently Updatable Neural Networks), turned out to become extremely powerful.  
 
This approach, dubbed [[NNUE]] (&#398;U&#1048;&#1048; Efficiently Updatable Neural Networks), turned out to become extremely powerful.  
 
NNUE was used along with a [[Stockfish]] based Shogi adaptation in [[YaneuraOu]], as well in other Shogi playing entities such as  
 
NNUE was used along with a [[Stockfish]] based Shogi adaptation in [[YaneuraOu]], as well in other Shogi playing entities such as  
[[Kristallweizen-kai]] <ref>[https://github.com/Tama4649/Kristallweizen/ GitHub - Tama4649/Kristallweizen: 第29回世界コンピュータ将棋選手権 準優勝のKristallweizenです。]</ref>,
+
[[Kristallweizen]] <ref>[https://github.com/Tama4649/Kristallweizen/ GitHub - Tama4649/Kristallweizen: 第29回世界コンピュータ将棋選手権 準優勝のKristallweizenです。]</ref>,
 
apparently with superhuman [[AlphaZero]] strength <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=2&t=72754 The Stockfish of shogi] by [[Larry Kaufman]], [[CCC]], January 07, 2020</ref>.
 
apparently with superhuman [[AlphaZero]] strength <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=2&t=72754 The Stockfish of shogi] by [[Larry Kaufman]], [[CCC]], January 07, 2020</ref>.
YaneuraOu team member [[Hisayori Noda]] aka Nodchip proved that Nasu's idea also works well in [[Chess|western chess]], yielding to [[Stockfish NNUE]] <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74059 Stockfish NN release (NNUE)] by [[Henk Drost]], [[CCC]], May 31, 2020</ref>.
+
YaneuraOu team member [[Hisayori Noda]] aka Nodchip proved that Nasu's idea also works well in [[Chess|western chess]], yielding to [[Stockfish NNUE]] <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=2&t=74059 Stockfish NN release (NNUE)] by [[Henk Drost]], [[CCC]], May 31, 2020</ref> and in September 2020 to '''Stockfish 12''' <ref>[https://blog.stockfishchess.org/post/628172810852925440/stockfish-12 Stockfish 12], The Stockfish Team, [https://blog.stockfishchess.org/ Stockfish Blog], September 02, 2020</ref>.
  
 
=Selected Publications=
 
=Selected Publications=
Line 32: Line 34:
 
* [https://dblp.org/pers/hd/y/Yokoyama:Ryo Ryo Yokoyama], [[Yu Nasu]], [https://dblp.org/pers/hd/s/Shinoda:Koichi Koichi Shinoda], [https://dblp.org/pers/hd/i/Iwano:Koji Koji Iwano] ('''2012'''). ''[https://www.semanticscholar.org/paper/Overlapped-Speech-Detection-in-Meeting-Using-and-Yokoyama-Nasu/862c1571c47ca77ce5f227b47eef77a35c264a53 Overlapped Speech Detection in Meeting Using Cross-Channel Spectral Subtraction and Spectrum Similarity]''. [https://dblp.org/db/conf/interspeech/interspeech2012.html INTERSPEECH 2012]
 
* [https://dblp.org/pers/hd/y/Yokoyama:Ryo Ryo Yokoyama], [[Yu Nasu]], [https://dblp.org/pers/hd/s/Shinoda:Koichi Koichi Shinoda], [https://dblp.org/pers/hd/i/Iwano:Koji Koji Iwano] ('''2012'''). ''[https://www.semanticscholar.org/paper/Overlapped-Speech-Detection-in-Meeting-Using-and-Yokoyama-Nasu/862c1571c47ca77ce5f227b47eef77a35c264a53 Overlapped Speech Detection in Meeting Using Cross-Channel Spectral Subtraction and Spectrum Similarity]''. [https://dblp.org/db/conf/interspeech/interspeech2012.html INTERSPEECH 2012]
 
* [https://dblp.org/pers/hd/o/Ohtani:Yamato Yamato Ohtani], [[Yu Nasu]], [https://dblp.org/pers/hd/m/Morita:Masahiro Masahiro Morita], [https://dblp.org/pers/hd/a/Akamine:Masami Masami Akamine] ('''2015'''). ''[https://www.semanticscholar.org/paper/Emotional-transplant-in-statistical-speech-based-on-Ohtani-Nasu/5d6c1846bef6e8f796a84f6d7a381fdcffe1e1dd Emotional transplant in statistical speech synthesis based on emotion additive model]''. [https://dblp.org/db/conf/interspeech/interspeech2015.html#OhtaniNMA15 INTERSPEECH 2015] <ref>[https://en.wikipedia.org/wiki/Hidden_semi-Markov_model hidden semi-Markov model from Wikipedia]</ref> <ref>[https://en.wikipedia.org/wiki/Kernel_eigenvoice Kernel eigenvoice from Wikipedia]</ref>
 
* [https://dblp.org/pers/hd/o/Ohtani:Yamato Yamato Ohtani], [[Yu Nasu]], [https://dblp.org/pers/hd/m/Morita:Masahiro Masahiro Morita], [https://dblp.org/pers/hd/a/Akamine:Masami Masami Akamine] ('''2015'''). ''[https://www.semanticscholar.org/paper/Emotional-transplant-in-statistical-speech-based-on-Ohtani-Nasu/5d6c1846bef6e8f796a84f6d7a381fdcffe1e1dd Emotional transplant in statistical speech synthesis based on emotion additive model]''. [https://dblp.org/db/conf/interspeech/interspeech2015.html#OhtaniNMA15 INTERSPEECH 2015] <ref>[https://en.wikipedia.org/wiki/Hidden_semi-Markov_model hidden semi-Markov model from Wikipedia]</ref> <ref>[https://en.wikipedia.org/wiki/Kernel_eigenvoice Kernel eigenvoice from Wikipedia]</ref>
* [[Yu Nasu]] ('''2018'''). ''&#398;U&#1048;&#1048; Efficiently Updatable Neural-Network based Evaluation Functions for Computer Shogi''.  Ziosoft Computer Shogi Club, [https://github.com/ynasu87/nnue/blob/master/docs/nnue.pdf pdf], [https://www.apply.computer-shogi.org/wcsc28/appeal/the_end_of_genesis_T.N.K.evolution_turbo_type_D/nnue.pdf pdf] (Japanese with English abstract)
+
* [[Hisayori Noda]], [[Jun Okabe]], [[Takahiro Suzuki]], [[Masatoshi Hidaka]], [[Yu Nasu]], [[Akio Kono]] ('''2017'''). ''Shouten Gensou Knights of Tanuki Appeal Document''. The Minstrel's Ballad: Tanuki's Reign, [http://www2.computer-shogi.org/wcsc27/appeal/tanuki-/appeal.pdf pdf] (Japanese)
 +
* [[Yu Nasu]] ('''2018'''). ''&#398;U&#1048;&#1048; Efficiently Updatable Neural-Network based Evaluation Functions for Computer Shogi''.  Ziosoft Computer Shogi Club, [https://github.com/ynasu87/nnue/blob/master/docs/nnue.pdf pdf], [https://www.apply.computer-shogi.org/wcsc28/appeal/the_end_of_genesis_T.N.K.evolution_turbo_type_D/nnue.pdf pdf] (Japanese with English abstract) [https://github.com/asdfjkl/nnue GitHub - asdfjkl/nnue translation] <ref>[http://www.talkchess.com/forum3/viewtopic.php?f=2&t=76250 Translation of Yu Nasu's NNUE paper] by [[Dominik Klein]], [[CCC]], January 07, 2021</ref>
  
 
=External Links=
 
=External Links=
 
* [https://www.researchgate.net/scientific-contributions/69800429_Yu_Nasu Yu Nasu's research works | Tokyo Institute of Technology, Tokyo (TITech) and other places], [https://en.wikipedia.org/wiki/ResearchGate ResearchGate]
 
* [https://www.researchgate.net/scientific-contributions/69800429_Yu_Nasu Yu Nasu's research works | Tokyo Institute of Technology, Tokyo (TITech) and other places], [https://en.wikipedia.org/wiki/ResearchGate ResearchGate]
* [http://yaneuraou.yaneu.com/ やねうら王 公式サイト | コンピューター将棋 やねうら王 公式サイト] (YaneuraOu)
+
* [http://yaneuraou.yaneu.com/ やねうら王 公式サイト | コンピューター将棋 やねうら王 公式サイト YaneuraOu Official Website] by [[Motohiro Isozaki]] aka Yaneurao
* [https://github.com/yaneurao/YaneuraOu GitHub - yaneurao/YaneuraOu: YaneuraOu is the World's Strongest Shogi engine(AI player), WCSC29 1st winner, educational and USI compliant engine]
+
*  [http://yaneuraou.yaneu.com/2019/06/24/%E5%B0%86%E6%A3%8B%E3%82%BD%E3%83%95%E3%83%88%E9%96%8B%E7%99%BA%E8%80%85%E3%81%8Cstockfish%E3%81%AB%E8%B2%A2%E7%8C%AE%E3%81%99%E3%82%8B%E6%97%A5/ 将棋ソフト開発者がStockfishに貢献する日 The day when shogi software developers contribute to Stockfish] by [[Yu Nasu]], June 2019 » [[Stockfish NNUE]]
 +
* [https://github.com/yaneurao/YaneuraOu GitHub - yaneurao/YaneuraOu: YaneuraOu is the World's Strongest Shogi engine(AI player), WCSC29 1st winner, educational and USI compliant engine], hosted by [[Motohiro Isozaki]] aka Yaneurao
  
 
=References=  
 
=References=  
Line 44: Line 48:
 
[[Category:Shogi Programmer|Nasu]]
 
[[Category:Shogi Programmer|Nasu]]
 
[[Category:Researcher|Nasu]]
 
[[Category:Researcher|Nasu]]
 +
[[Category:Stockfish Contributor|Nasu]]

Latest revision as of 22:59, 22 May 2021

Home * People * Yu Nasu

Yu Nasu, (那須悠)
a Japanese computer scientist and Shogi programmer. While affiliated with the Tokyo Institute of Technology [1], and the Knowledge Media Laboratory, Toshiba, his research focused on natural language processing, speech recognition and speech synthesis as application of hidden Markov model and neural networks [2].

Shogi

Along with Hisayori Noda (Nodchip), Jun Okabe, Takahiro Suzuki, Masatoshi Hidaka, and at times Akio Kono, Yu Nasu was member of The Minstrel's Ballad: Tanuki's Reign [3] [4] and subsequently is member of the Ziosoft Computer Shogi Club, developing and representing the Shogi playing entities Shouten Gensou Knights of Tanuki (2017, derived from Yaneura King by Motohiro Isozaki aka Yaneurao) [5] [6], the end of genesis T.N.K.evolution turbo type D [7] [8] and Tanu-King at the WCSC29 [9].

NNUE

In 2018, Yu Nasu introduced a new evaluation function based on a shallow neural network whose heavily overparametrized input layer outputs were incremental updated in make and unmake move. This approach, dubbed NNUE (ƎUИИ Efficiently Updatable Neural Networks), turned out to become extremely powerful. NNUE was used along with a Stockfish based Shogi adaptation in YaneuraOu, as well in other Shogi playing entities such as Kristallweizen [10], apparently with superhuman AlphaZero strength [11]. YaneuraOu team member Hisayori Noda aka Nodchip proved that Nasu's idea also works well in western chess, yielding to Stockfish NNUE [12] and in September 2020 to Stockfish 12 [13].

Selected Publications

[14]

External Links

References

Up one level