Unsupervised learning: The curious pupil
Research Unsupervised learning: The curious pupil Published 25 June 2019 Authors Alexander Graves, Kelly Clancy One in a series of posts explaining the theories underpinning our research. Over the last decade, machine learning has made unprecedented progress in areas as diverse as image recognition, self-driving cars and playing complex games like Go. These successes have been largely realised by
囲ç¢AIブームã«ä¹—ã£ã¦ã€è‹¥æ‰‹æ£‹å£«ã®é–“ã§ã€ŒAWSã€ãŒå¤§æµè¡Œã€€ãã®ç†ç”±ã¨ã¯ï¼Ÿ
囲ç¢AIブームã«ä¹—ã£ã¦ã€è‹¥æ‰‹æ£‹å£«ã®é–“ã§ã€ŒAWSã€ãŒå¤§æµè¡Œã€€ãã®ç†ç”±ã¨ã¯ï¼Ÿï¼šé€±æœ«ã‚¨ãƒ³ãƒ—ラã“ã¼ã‚Œè©±ï¼ˆ1/4 ページ） 人間ã®èƒ½åŠ›ã‚’AIãŒå®Œå…¨ã«ä¸Šå›žã‚Šã¤ã¤ã‚る「囲ç¢ã€ã®ä¸–界。最近ã§ã¯ã€AIを活用ã—ãŸç ”究を行ã†æ£‹å£«ã‚‚増ãˆã¦ã„ã‚‹ãã†ã ãŒã€ãã®è£å´ã§AWSãŒè‹¥æ‰‹æ£‹å£«ã®ä¸ã§å¤§æµè¡Œã—ã¦ã„ã‚‹ã¨ã„ã†ã€‚一体何ãŒèµ·ã“ã£ã¦ã„ã‚‹ã®ã ã‚ã†ã‹ã€‚ ã“ã“æ•°å¹´ã€å°†æ£‹ã‚„囲ç¢ã¨ã„ã£ãŸãƒœãƒ¼ãƒ‰ã‚²ãƒ¼ãƒ ã®ä¸–ç•Œã§ã¯ã€AI（人工知能）ã®èƒ½åŠ›ãŒäººé–“を上回りã¤ã¤ã‚る。特ã«ã€Alphabet傘下ã®DeepMindãŒé–‹ç™ºã—ãŸå›²ç¢AI「AlphaGoã€ã¯ã€ä¸–ç•Œã®ãƒˆãƒƒãƒ—棋士を次々ã¨ç ´ã£ãŸã“ã¨ã§ã€æ˜¨ä»Šã®äººå·¥çŸ¥èƒ½ãƒ–ームã®â€œç«ä»˜ã‘å½¹â€ã¨ãªã£ãŸã®ã¯è¨˜æ†¶ã«æ–°ã—ã„。 最近ã§ã¯ã€ãƒ—ãƒæ£‹å£«ãŸã¡ã‚‚ç ”ç©¶ã«AIを使ã„始ã‚ã¦ã„ã‚‹ãŒã€ãã®å½±éŸ¿ã§ã€è‹¥ã„囲ç¢æ£‹å£«ãŸã¡ã®é–“ã§ä»Šã€ŒAWS（Amazon Web Services）ã€ã‚’利用ã™ã‚‹äººãŒæ€¥é€Ÿã«å¢—ãˆã¦ã„ã‚‹ã®ã ã¨ã„ã†ã€‚一体何ãŒèµ·ãã¦ã„ã‚‹
AlphaZero Shogiã®å¦ç¿’ã«ã‹ã‹ã‚‹ã‚³ã‚¹ãƒˆã‚’試算ã—ã¦ã¿ãŸ | ã‚„ãã†ã‚‰çŽ‹ å…¬å¼ã‚µã‚¤ãƒˆ
昨日ã®è¨˜äº‹ã§ã€AlphaZeroãŒã©ã‚Œãらã„ã®å¼·ã•ãªã®ã‹ã–ã£ãり試算ã—ãŸã®ã§ã™ãŒã€ã–ã£ãã‚Šã™ãŽã¦ã‹ãªã‚Šã„ã„åŠ æ¸›ã ã£ãŸã®ã§ã€LC0(LeelaChessZero)ã®benchmarkã‚’å…ƒã«ã€æ£ç¢ºã«è¨ˆç®—ã—ãªãŠã—ã¾ã™ã€‚ AlphaZeroãŒTPUv1×4ã§R4500相当。ニューラルãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã¯ResNet 20ブãƒãƒƒã‚¯ã§ãƒãƒ£ãƒãƒ«æ•°ã¯å…¬é–‹ã•ã‚Œã¦ã„ã¾ã›ã‚“ãŒã€npsã¯ã€80k(80,000)ã ãã†ã§ã™ã€‚(※) ※　https://arxiv.org/pdf/1712.01815.pdf 一方ã€LC0ã®Benchmarks(※)ã«ã‚ˆã‚‹ã¨ã€(CPUå´ã®)スレッド数ã«ã‚ˆã‚Šå¤šå°‘変動ã¯ã‚ã‚‹ã‚‚ã®ã®ã€ResNet 20ブãƒãƒƒã‚¯Ã—256ch構æˆã«ãŠã„ã¦ã€ä»¥ä¸‹ã®ã‚ˆã†ã«ãªã£ã¦ã„ã¾ã—ãŸã€‚ ・GTX 1080Ti : 9knps ・RTX 2080 : 26.1knps ・TITAN V : 31knps ・RTX 208
AlphaGoã®è¡æ’ƒå†ã³ — ã‚¿ãƒ³ãƒ‘ã‚¯è³ªæ§‹é€ äºˆæ¸¬ã§AlphaFoldãŒä»Šã¾ã§ã®ãƒ¢ãƒ‡ãƒ«ã«åœ§å‹ï½œZhubo.JP
原文記事: [阿尔法狗å†ä¸‹ä¸€åŸŽ | 蛋白结构预测AlphaFoldå¤§èƒœä¼ ç»Ÿäººç±»æ¨¡åž‹] (2018/12/03公開) ã€Œç ”ç©¶ã—ãŸã„タンパク質ãŒã‚ã‚‹ã®ã ãŒã€ãã®æ§‹é€ ã¨æ©Ÿèƒ½ãŒã‚ã‹ã‚‰ãªã„〠— ã“ã‚Œã¯åˆ†å細胞生物å¦ã®ç ”究者ãŒæ—¥ã€…ç›´é¢ã™ã‚‹æœ€å¤§ã®é›£é¡Œã®ä¸€ã¤ã§ã‚ã‚‹*a。アミノ酸é…列測定技術ãŒç™ºå±•ã™ã‚‹ä¸ã§ã€å¤šãã®ã‚¿ãƒ³ãƒ‘ク質ã®é…列ãŒãƒã‚¤ã‚¹ãƒ«ãƒ¼ãƒ—ット*1ã«è§£æžã•ã‚Œã¦ã„ã‚‹ãŒã€ã“ã®é…列決定ã®æ®µéšŽã‹ã‚‰å®Ÿéš›ã«3æ¬¡å…ƒæ§‹é€ ã‚’æ±ºå®šã™ã‚‹ã¾ã§ã®é–“ã«ã¯ã€æœªã ã«å¤§ããªè·é›¢ãŒã‚る。 生物ã®åŸºæœ¬å˜ä½ãŒç´°èƒžã ã¨ã—ã¦ã€ç´°èƒžã®åŸºæœ¬çš„ãªæ©Ÿèƒ½å˜ä½ã“ããŒã€è¤‡é›‘多å²ãªã‚¿ãƒ³ãƒ‘ク質ã®1ã¤1ã¤ã§ã‚る。ãã—ã¦ã¾ã•ã«ã‚¿ãƒ³ãƒ‘ク質ã®æ©Ÿèƒ½ã®æœ¬è³ªã‚’決定ã—ã¦ã„ã‚‹ã®ãŒã€ã‚¿ãƒ³ãƒ‘ク質ã®æ§‹é€ ã§ã‚る。タンパク質ã®æ©Ÿèƒ½ã‚’ç ”ç©¶ã—ãŸã‚Šã€ãれをターゲットã¨ã™ã‚‹è–¬å‰¤ã‚’開発ã—ãŸã„ã¨ã„ã†ã¨ãã€ã‚¿ãƒ³ãƒ‘ク質ã®æ§‹é€ ã¯ã¨ã¦ã‚‚é‡è¦ãªè¦ç´ ã®ä¸€ã¤ã«ãªã‚‹ã€‚ã ã‹ã‚‰ã“ãã€ç”Ÿç‰©å¦ã«ã¯ã€æ§‹é€ 生物å¦ã¨ã„ã†å¦å•é ˜åŸŸã¾ã§
AlphaZeroã«æŠ•äº†å®£è¨€ã—ãªã„ã¨ã„ã‘ãªã„ã‹ã‚‚知れãªã„ - コンピュータ将棋 Qhapaq
AlphaZeroã®è«–æ–‡ãŒå‡ºã¦ãã¾ã—ãŸã€‚arxivã«å‡ºã—ã¦ãŸè«–æ–‡ã¯ã‚¤ã‚±ã¦ãªã‹ã£ãŸã§ã™ãŒä»Šå›žã®ã¯ã‚¤ã‚±ã¦ã¾ã™ã€‚ å‰å›žã®è«–æ–‡ã«å¯¾ã—ã¦ã¯ä»Šã®å¾¡æ™‚世ã§WCSC27ã®elmoã«å‹ã£ãŸãらã„㧠convincingly defeated a world champion program ã¨ã‹è¨€ã‚ãªã„ã§ã‚ˆã¨ã‹ã€æŠ•äº†ã®åŸºæº–ãŒ900点ãªã®æ„味ãŒã‚ã‹ã‚‰ã‚“ã¨ã‹ã€å®šè·¡ã¯ã©ã†ã—ãŸã®ã‹ã¨ã‹ã€æ£‹èœãŒãªã„ã®ã¯ã©ã†ã„ã†ã“ã¨ã‹ã¨ã‹æ•£ã€…ãªè©•ä¾¡ã‚’ã—ãŸè¨˜æ†¶ãŒã‚ã‚Šã¾ã™ã€‚ ãŒã€ãªã‚“ã¨ä»Šå›žã¯æ£‹èœãŒã‚ã‚Šã¾ã™ã€‚ã—ã‹ã‚‚羽生竜王ã«æ£‹èœã‚³ãƒ¡ã‚’è²°ã£ãŸã‚ˆã†ã§ã™ã€‚棋èœã®å…¬é–‹ã«ä¼´ã„ã€æˆ¦åž‹ã®åˆ†å¸ƒã‚‚明らã‹ã«ãªã‚Šã¾ã—ãŸã€‚Zero言ã„ãªãŒã‚‰ã¡ã‚ƒã‚“ã¨å±…飛車を指ã—ã¾ã™ã€‚WCSC28ã§CrazyShogiも矢倉組んã§ãŸã®ã§é©šãã»ã©ã§ã¯ãªã„ã‹ã‚‚知れã¾ã›ã‚“。 ãã—ã¦ã€æ£‹èœã®å…¬é–‹ã¨åŒã˜ãらã„イケã¦ã‚‹ã®ãŒaperypaq(Apery-Qhapaq関数ã€ç•¥ã—ã¦aperypaq。2
Alpha Zeroã®è¡æ’ƒã¨æŠ€è¡“的失æ¥ï½œå±±æœ¬ä¸€æˆðŸš—TURING
2016å¹´ã€Google DeepMind社ã‹ã‚‰æã‚ã—ã„è«–æ–‡ãŒå‡ºã•ã‚ŒãŸã€AlphaGoãã®åã‚’å† ã—ãŸå›²ç¢ãƒ—ãƒã‚°ãƒ©ãƒ ãŒæ—¢å˜ã®å›²ç¢ã‚½ãƒ•ãƒˆã«å‹çŽ‡99ï¼…ã‚’å©ã出ã—ãŸã®ã 。AlphaGoã¯å¼·åŒ–å¦ç¿’ã¨Deep Learningを組ã¿åˆã‚ã›ãŸå›²ç¢ãƒ—ãƒã‚°ãƒ©ãƒ ã§ã€ãã®å¹´ã«æœ€å¼·ã®å›²ç¢æ£‹å£«ã®ä¸€äººã§ã‚ã‚‹æŽä¸–ドルã•ã‚“ã«4å‹1è² ã§å‹åˆ©ã—ãŸã€‚ãã®å¾Œã‚‚進æ©ã‚’続ã‘ã¦ä»Šã®AlphaGoã®å¼·ã•ã¯äººé¡žãŒä½“æ„Ÿã§ãるレベルを超ãˆã‚‹ã»ã©å¼·ããªã£ãŸã¨äºˆæƒ³ã•ã‚Œã‚‹ã€‚ 2017年も終ã‚ã‚Šã®ã“ã‚ã€Google DeepMind社ã‹ã‚‰ã¾ãŸé€”方もãªã„è«–æ–‡ãŒç™ºè¡¨ã•ã‚ŒãŸã€‚囲ç¢ã¨ã»ã¼åŒã˜æ‰‹æ³•ã§æœ€å¼·ãƒ¬ãƒ™ãƒ«ã®ãƒã‚§ã‚¹ã‚„将棋プãƒã‚°ãƒ©ãƒ を超ãˆãŸã¨ã„ã†ã“ã¨ã ã£ãŸã€‚実際ã®ã¨ã“ã‚æ£ç¢ºã«è¶…ãˆãŸã®ã‹ã©ã†ã‹ã¡ã‚‡ã£ã¨ã ã‘ç–‘å•ã‚‚ã‚ã‚‹ã®ã ãŒã€ã¾ãšå‰æã¨ã—ã¦å½¼ã‚‰ã®æ–°æ‰‹æ³•ãŒé€”方もãªã„æˆæžœã‚’ã‚ã’ãŸã“ã¨ç´ ç›´ã«ç¥ç¦ã—ãŸã„。彼らã¯è‡ªåˆ†ãŸã¡ã®ãƒ—ãƒã‚°ãƒ©ãƒ ã‚’Alpha Zeroã¨å付ã‘ãŸã€‚ コンピュ
AlphaZero: Shedding new light on the grand games of chess, shogi and Go | DeepMind
I can’t disguise my satisfaction that it plays with a very dynamic style, much like my own! Garry Kasparov, Former World Chess Champion This ability to learn each game afresh, unconstrained by the norms of human play, results in a distinctive, unorthodox, yet creative and dynamic playing style. Chess Grandmaster Matthew Sadler and Women’s International Master Natasha Regan, who have analysed thous
ã€Shogi】AlphaZeroã®æ£‹èœãŒå…¬é–‹ã•ã‚ŒãŸ - ã‚³ãƒ³ãƒ”ãƒ¥ãƒ¼ã‚¿å°†æ£‹ç ”ç©¶Blog
ã¯ã˜ã‚ã« 2017å¹´12æœˆé ƒã«è©±é¡Œã«ãªã£ãŸã€DeepMind社ã«ã‚ˆã‚‹AlphaZeroã®å°†æ£‹ç‰ˆã§ã‚ã‚‹AlphaShogiãŒç¬¬27回世界コンピュータ将棋é¸æ‰‹æ¨©ã§å„ªå‹ã—ãŸãƒãƒ¼ã‚¸ãƒ§ãƒ³ã®elmoã¨å¯¾æˆ¦ã—ãŸ100å±€ã®æ£‹èœãŒå…¬é–‹ã•ã‚ŒãŸã€‚ There are also 100 shogi games for players to enjoy with a top 10 selected by Shogi legend Yoshiharu Habu https://t.co/ZJDoaon5z0— DeepMind (@DeepMindAI) 2018å¹´12月6æ—¥ ãã®æ£‹èœã‚’早速ダウンãƒãƒ¼ãƒ‰ã—ã¦å†…容を確èªã—ã¦ã¿ãŸã®ã§ãã®æ„Ÿæƒ³ã‚’書ã„ã¦ã¿ãŸã„ã¨æ€ã†ã€‚ スãƒãƒ³ã‚µãƒ¼ãƒªãƒ³ã‚¯ 棋èœã‚’見ã¦ã®æ„Ÿæƒ³ AlphaShogiã¯å±…飛車党ã§ç›¸æŽ›ã‹ã‚Šã¨è§’æ›ã‚ã‚Šã€æ¨ªæ©å–ã‚Šé’野æµã‚’好むよã†ã 。 以å‰ã«ç™ºè¡¨ã•ã‚ŒãŸè«–æ–‡https://arx
ã‚„ã™ã„ on Twitter: "タンパク質ãŒä¸‰æ¬¡å…ƒç©ºé–“ã§ã©ã†æŠ˜ã‚Šç•³ã¾ã‚Œã¦ã‚‹ã‹ï¼ˆç’°å¢ƒã«é‚„ã£ã¦ã„ãç´ æやアルツãƒã‚¤ãƒžãƒ¼ãªã©ã®ç—…ã®æ²»ç™‚法発見ã«å½¹ç«‹ã¤å•é¡Œï¼‰ã‚’予測ã™ã‚‹2å¹´ã«ä¸€åº¦ã®å›½éš›ã‚³ãƒ³ãƒšCASPã§ã€ã‚°ãƒ¼ã‚°ãƒ«å会社DeepMindã«ã‚ˆã‚‹AlphaFoldãŒæ·±å±¤å¦ç¿’を活用… https://t.co/gtRjtYE3KV"
タンパク質ãŒä¸‰æ¬¡å…ƒç©ºé–“ã§ã©ã†æŠ˜ã‚Šç•³ã¾ã‚Œã¦ã‚‹ã‹ï¼ˆç’°å¢ƒã«é‚„ã£ã¦ã„ãç´ æやアルツãƒã‚¤ãƒžãƒ¼ãªã©ã®ç—…ã®æ²»ç™‚法発見ã«å½¹ç«‹ã¤å•é¡Œï¼‰ã‚’予測ã™ã‚‹2å¹´ã«ä¸€åº¦ã®å›½éš›ã‚³ãƒ³ãƒšCASPã§ã€ã‚°ãƒ¼ã‚°ãƒ«å会社DeepMindã«ã‚ˆã‚‹AlphaFoldãŒæ·±å±¤å¦ç¿’を活用… https://t.co/gtRjtYE3KV
IBM Watsonã®åŒ»ç™‚分野ã§ã®è‹¦æˆ¦ãŒæ˜Žã‚‰ã‹ã«ï½œã¨ãã•ã‚“｜MYCOPING
ãƒã‚¤ãƒ†ã‚¯æ¥ç•Œã®ãƒ‹ãƒ¥ãƒ¼ã‚¹ã‚„仕事ã®ãŠå½¹ç«‹ã¡æƒ…å ±ãªã©ã€çš†ã•ã‚“ã®å½¹ã«ç«‹ã¤ãƒã‚¿ã‚’毎日紹介ã™ã‚‹è¨˜äº‹ã‚’ã¯ã˜ã‚ã¦ã¿ã‚ˆã†ã‹ã¨æ€ã„ã¾ã™ã€‚ãã‚Œã«åŠ ãˆã¦ã€Œä»Šæ—¥ã®ã¤ã¶ã‚„ãã€ã¨ã„ã†ã“ã¨ã§ã€å€‹äººã§é€²ã‚ã¦ã„るサイドプãƒã‚¸ã‚§ã‚¯ãƒˆã®ã“ã¨ã€æ—¥ã€…ã®çµŒå–¶ã§ã®æ°—ã¥ãや考察ãªã©ã‚’綴ã£ã¦ã„ãã¾ã™ã€‚特ã«ã‚µã‚¤ãƒ‰ãƒ—ãƒã‚¸ã‚§ã‚¯ãƒˆã«ã¤ã„ã¦ã¯ã€ä»•äº‹ã¨è‚²å…ã«æ¯Žæ—¥è¿½ã‚ã‚ŒãªãŒã‚‰ã€ã„ã¾ã¾ã§ãƒ—ãƒãƒ€ã‚¯ãƒˆé–‹ç™ºã‚‚事æ¥é–‹ç™ºã‚‚ã—ãŸã“ã¨ãŒãªã„ã€Œç´ äººã€ãŒæ‰‹æŽ¢ã‚Šã§é€²ã‚ã¦ã„る様åを綴ã£ã¦ã„ãã®ã§é–¢å¿ƒãŒã‚ã‚‹æ–¹ã¯ã€åˆæœˆç„¡æ–™ã§ã™ã®ã§ãœã²ãƒžã‚¬ã‚¸ãƒ³è³¼èªã—ã¦ã¿ã¦ãã ã•ã„ï¼ IBMã®äººå·¥çŸ¥èƒ½ã€ŒWatson(ワトソン)ã€ã«ã‚ˆã‚‹åŒ»ç™‚診æ–システムã¯ã€Œå®Ÿç”¨ã«è€ãˆã†ã‚‹ã‚‚ã®ã§ã¯ãªã„ã€ã¨ã„ã†ä¸»å¼µ (GIGAZINE) 最近IBMã®WatsonãŒåŒ»ç™‚分野ã§ã€Œå½¹ç«‹ãŸãšã€ã§ã‚ã‚‹ã€ã¨ã„ã†ãƒ‹ãƒ¥ãƒ¼ã‚¹ãŒå¢—ãˆã¦ã„ã¾ã™ã€‚ã“ã®è¨˜äº‹ã‚‚ãã®ä¸€ã¤ã€‚胸ã®ç—›ã¿ã§è¨ªã‚ŒãŸæ‚£è€…ã«ã€Œçã—ã„感染症ã€ã‚’ã¾ãšç–‘ã£ãŸã‚Šã€ãŒã‚“診æ–ã§æ•™ç§‘書ã«è¼‰ã£ã¦ã„るよ
DeepMind, meet Android
Impact DeepMind, meet Android Published 8 May 2018 Authors James Smith, Simon Rosen, Chris Gamble We’re delighted to announce a new collaboration between DeepMind for Google and Android, the world’s most popular mobile operating system. Together, we’ve created two new features that will be available to people with devices running Android P later this year: Adaptive Battery: A smart battery managem
Google DeepMindã€æ–‡å—ã‚„ç”»åƒãªã©ã‚’æã時ã®ãƒ–ラシストãƒãƒ¼ã‚¯ã‚’強化å¦ç¿’を用ã„ã¦æ•µå¯¾çš„ã«è¨“ç·´ã™ã‚‹ã€ŒSPIRALã€ç™ºè¡¨
Google DeepMindã€æ–‡å—ã‚„ç”»åƒãªã©ã‚’æã時ã®ãƒ–ラシストãƒãƒ¼ã‚¯ã‚’強化å¦ç¿’を用ã„ã¦æ•µå¯¾çš„ã«è¨“ç·´ã™ã‚‹ã€ŒSPIRALã€ç™ºè¡¨ 2018-04-08 Google DeepMindã¯ã€æ•°å—ã€æ–‡å—ã€è‚–åƒç”»ãªã©ã‚’æãã¨ãã®ãƒ–ラシストãƒãƒ¼ã‚¯ã‚’機械å¦ç¿’を用ã„ã¦æŽ¨è«–ã™ã‚‹ã€ŒSPIRAL（Synthesizing Programs for Images using Reinforced Adversarial Learning）ã€ã‚’発表ã—ã¾ã—ãŸã€‚ 論文：Synthesizing Programs for Images using Reinforced Adversarial Learning 著者：Yaroslav Ganin,Tejas Kulkarni,Igor Babuschkin,S. M. Ali Eslami,Oriol Vinyals 本稿ã¯ã€æ©Ÿæ¢°å¦ç¿’を用ã„ã¦ç†ã®ã‚¹ãƒˆãƒãƒ¼ã‚¯ã‚’考慮ã—ãªãŒã‚‰ç”Ÿ
AlphaGoを開発ã—ãŸGoogleã®äººå·¥çŸ¥èƒ½ç ”究部門ãŒã€Œæ©Ÿæ¢°å¦ç¿’ã«ã‚ˆã£ã¦æ›¸ãé †ã‚’è€ƒãˆãªãŒã‚‰æ–‡å—ã‚’æãAIã€ã‚’開発
書é“や似顔絵ã«è¦‹ã‚‰ã‚Œã‚‹ã‚ˆã†ã«ã€ç§ãŸã¡äººé–“ã¯çµµç”»ãƒ»æ›¸ãƒ»ç«‹ä½“ã®ä½œå“を見ãŸæ™‚ã€ãƒ‡ã‚¶ã‚¤ãƒ³ã«æ„Ÿå‹•ã™ã‚‹ã ã‘ã§ã¯ãªãã€ç†é‹ã³ã‚„æ§‹é€ ãªã©ã‚’考ãˆã¦ç†è§£ã—ãŸä¸Šã§ä½œå“ã‚’å†ç¾ã™ã‚‹ã“ã¨ãŒã§ãã¾ã™ã€‚与ãˆã‚‰ã‚ŒãŸç”»åƒã‚’å†ç¾ã™ã‚‹äººå·¥çŸ¥èƒ½(AI)ã¯ä»Šã¾ã§ã‚‚発表ã•ã‚Œã¦ãã¾ã—ãŸãŒã€æœ€å¼·ã®å›²ç¢AI・AlphaGoを開発ã—ãŸGoogle DeepMindãŒç™ºè¡¨ã—ãŸã€ŒSynthesizing Programs for Images using Reinforced Adversarial Learningã€(敵対的強化å¦ç¿’を用ã„ãŸæ画プãƒã‚°ãƒ©ãƒ åˆæˆ)ã€ç•¥ã—ã¦ã€ŒSPIRALã€ã¯ã€å¯¾è±¡ã¨ãªã‚‹æ–‡å—ã‚„ç”»åƒã®æ›¸ãé †ã‚„æ§‹é€ ã‚’è€ƒãˆã¦å¦ç¿’ã—ãªãŒã‚‰ã€è‡ªã‚‰æ画プãƒã‚°ãƒ©ãƒ をアップデートã—ã¦ã„ãフレームワークã ã¨ã®ã“ã¨ã§ã™ã€‚ Learning to write programs that generate images | DeepMind https://
ã‚¢ãƒ«ãƒ•ã‚¡ç¢ | Netflix (ãƒãƒƒãƒˆãƒ•ãƒªãƒƒã‚¯ã‚¹)
シンプルã§ã„ã¦å¥¥ãŒæ·±ã„ボードゲームã€å›²ç¢ã€‚コンピュータプãƒã‚°ãƒ©ãƒ ã¨å›²ç¢ã®å人ãŒç¹°ã‚Šåºƒã’ãŸç†±æˆ¦ã¯ã€äººå·¥çŸ¥èƒ½ãŒäººé–“を超ãˆã‚ˆã†ã¨ã™ã‚‹æ´å²çš„ãªæˆ¦ã„ã§ã‚‚ã‚ã£ãŸã€‚
AI and Deep Learning in 2017 – A Year in Review
The year is coming to an end. I did not write nearly as much as I had planned to. But I’m hoping to change that next year, with more tutorials around Reinforcement Learning, Evolution, and Bayesian Methods coming to WildML! And what better way to start than with a summary of all the amazing things that happened in 2017? Looking back through my Twitter history and the WildML newsletter, the followi
ãŠçŸ¥ã‚‰ã›
障害
ランã‚ング
リリースã€éšœå®³æƒ…å ±ãªã©ã®ã‚µãƒ¼ãƒ“スã®ãŠçŸ¥ã‚‰ã›
最新ã®äººæ°—エントリーã®é…ä¿¡
j次ã®ãƒ–ックマーク
kå‰ã®ãƒ–ックマーク
lã‚ã¨ã§èªã‚€
eコメント一覧を開ã
oページを開ã
DeepMind on Twitter: "Curious about AI and want to learn more? Introducing the DeepMind podcast! @FryRsquared will be taking you behind t… https://t.co/uAwg7fRyJx"
将棋ソフト「mEssiahã€å…¬å¼ã•ã‚“ã®ãƒ„イート: "ãŸã¶ã‚“æŒã¡æ™‚間や探索速度ã«å¿œã˜ã¦æœ€é©ãªãƒã‚¤ãƒ‘ーパラメータをè¨å®šã—ã¦ã‚„らãªã„ã¨ã„ã‘ãªã„é¡žã ã¨æ€ã„ã¾ã™ã€‚ãƒãƒªã‚·ãƒ¼ãƒãƒƒãƒˆã¨ãƒãƒªãƒ¥ãƒ¼ãƒãƒƒãƒˆã®ç²¾åº¦ã«å¼·ãä¾å˜ã™ã‚‹æŽ¢ç´¢ç²¾åº¦ã‚’å‘上ã™ã‚‹æ–¹æ³•ã¨ã—ã¦ã¯ã¨ã¦ã‚‚ç°¡å˜ã§ã„ã„方法ã ã¨æ€ã†ã®ã§ã€å–り入れãªã„手ã¯ãªã„ã§ã™ã。… "
DeepMindã•ã‚“ã®ãƒ„イート: "The full peer-reviewed @sciencemagazine evaluation of #AlphaZero is here - a single algorithm that creatively masters chess, shogi and Go through self-play https://t.co/E3OPUOIFw4… https://t.co/Vn24pfbcGy"
å°é³¥éŠã‚Šã‚‡ã†ï¼ˆã‚„ãŒã¦ãƒ¬ãƒ™ãƒ«3ã«ãªã‚‹ï¼‰ã•ã‚“ã®ãƒ„イート: "ディープマインドã¯ã‚³ãƒ³ã‚¹ã‚¿ãƒ³ãƒˆã«æˆæžœã‚’出ã—続ã‘ã¦ã„ãŸã€‚全然é™ã‹ã˜ã‚ƒãªã‹ã£ãŸã€‚ãƒã‚§ã‚¹ã€å°†æ£‹ã€å›²ç¢ã«ä½¿ãˆã‚‹æ±Žç”¨æ€§ã‚’æŒã¤æ‰‹æ³•ã€Œã‚¢ãƒ«ãƒ•ã‚¡ã‚¼ãƒã€ã€å˜ä¸€ã‚¨ãƒ¼ã‚¸ã‚§ãƒ³ãƒˆã§è¤‡æ•°ã®ã‚²ãƒ¼ãƒ ã‚’è¡Œãˆã‚‹IMPALAã€ãれを発展ã•ã›ã¦ãã®ã‚²ãƒ¼ãƒ 群ã§äººé–“ã®ä¸å¤®å€¤ã¾ã§å¼•ã上ã’ãŸãƒžãƒ«ãƒã‚¿ã‚¹ã‚¯æ·±å±¤å¼·åŒ–å¦ç¿’PopArt-IMPARA"
GitHub - deepmind/graph_nets: Build Graph Nets in Tensorflow
{{#tags}}- {{label}}
{{/tags}}