Microsoft Translator launching Neural Network based translations for all its speech languages - Microsoft Translator Blog
Microsoft Translator is now powering all speech translation through state-of-the-art neural networks. All speech translation apps that use this service, such as Skype Translator and the Microsoft Translator app for mobile devices, are now using neural network technology. Furthermore, the technology is available to all developers and end-users who want to use the Microsoft Translator speech API to
A Neural Network for Machine Translation, at Production Scale
Posted by Quoc V. Le & Mike Schuster, Research Scientists, Google Brain Team Ten years ago, we announced the launch of Google Translate, together with the use of Phrase-Based Machine Translation as the key algorithm behind this service. Since then, rapid advances in machine intelligence have improved our speech recognition and image recognition capabilities, but improving machine translation remai
ゼミã§åŠæ•™å¸«ã‚ã‚Šå¦ç¿’ã®ã“ã¨ã‚’発表ã—ãŸã®ã§ã‚¹ãƒ©ã‚¤ãƒ‰ã‚’公開ã—ã¾ã™ - yasuhisa's blog
ç†è§£ãŒä¸å®Œå…¨ã ã—ã€ã‚¼ãƒŸã§å†ç”Ÿæ ¸ãƒ’ルベクト空間ã®ã“ã¨ã‚„ã£ã¦ãªã‹ã£ãŸã‹ã‚‰é€”ä¸ã§æ··ã–ã£ã¦ãŸã‚Šã¨ã‹ã§å¤§åˆ†ã‚ã‚Œãªè³‡æ–™ã«ãªã£ã¦ã¾ã™ãŒã€å…¬é–‹ã—ã¦ã¿ã‚‹ãƒ†ã‚¹ãƒˆã€‚ åŠæ•™å¸«ã‚ã‚Šå¦ç¿’ from syou6162 次回ã¯ä½•ã‚’発表ã™ã‚‹ã‹ã€‚。。
特徴é¸æŠž - 機械å¦ç¿’ã®ã€Œæœ±é·ºã®æœWikiã€
特徴é¸æŠž (feature selection)†\(k\)個ã®ç‰¹å¾´é‡(属性)ã®ãƒ™ã‚¯ãƒˆãƒ«ã§è¨˜è¿°ã•ã‚Œã¦ã„る対象ã«æ©Ÿæ¢°å¦ç¿’ã®æ‰‹æ³•ã‚’é©ç”¨ã™ã‚‹ã¨ã™ã‚‹ï¼Ž\(k\)個ã®ç‰¹å¾´é‡ã‚’å…¨ã¦åˆ©ç”¨ã›ãšï¼Œãã®ä¸ã§æœ‰ç”¨ãªã‚‚ã®ã‚’é¸ã³å‡ºã™ã®ãŒç‰¹å¾´é¸æŠžï¼Žãã®ç›®çš„ã¯æ¬¡ã®äºŒã¤ï¼š 目的変数ã¨ç„¡é–¢ä¿‚ãªç‰¹å¾´é‡ã‚’使ã‚ãªã„ã“ã¨ã§äºˆæ¸¬ç²¾åº¦ã‚’å‘上ã•ã›ã‚‹ï¼Ž å¦ç¿’ã•ã‚ŒãŸé–¢æ•°ã‚’，定性的ã«è§£é‡ˆã—ã‚„ã™ãã™ã‚‹ï¼Ž ãã®å®Ÿç¾æ‰‹æ³•ã¯æ¬¡ã®äºŒã¤ã«åˆ†ã‘られる： フィルター法：目的変数ã¨å„特徴é‡ã¨ã®æƒ…å ±ã‚²ã‚¤ãƒ³ãªã©ã®ï¼Œç‰¹å¾´ã®è‰¯ã•ã®è¦æº–を使ã£ã¦é¸æŠžã™ã‚‹ï¼Ž ラッパー法：特徴é‡ã®éƒ¨åˆ†é›†åˆã‚’使ã£ã¦å®Ÿéš›ã«å¦ç¿’アルゴリズムをé©ç”¨ã—，交差確èªæ³•ãªã©ã§æ±‚ã‚ãŸæ±ŽåŒ–誤差を最å°ã«ã™ã‚‹ç‰¹å¾´é‡ã®éƒ¨åˆ†é›†åˆã‚’é¸æŠžã™ã‚‹ï¼Ž å¦ç¿’アルゴリズムをé©ç”¨ã™ã‚‹å¿…è¦ãŒã‚るラッパー法ã®æ–¹ãŒäºˆæ¸¬ç²¾åº¦ã¯å‘上ã™ã‚‹ãŒï¼Œãƒ•ã‚£ãƒ«ã‚¿ãƒ¼æ³•ã®æ–¹ãŒã¯ã‚‹ã‹ã«é«˜é€Ÿï¼Ž 特徴é‡ãŒ\(k\)個ã‚ã‚‹ã¨ã，全部㧠\(2^k-1\)個ã®ç‰¹å¾´é‡ã®éƒ¨
COP-KMEANS(Constrained K-means)を試ã—ã¦ã¿ãŸ - ã®ã‚“ã³ã‚Šèªæ›¸æ—¥è¨˜
制約付ãクラスタリング・åŠæ•™å¸«ã‚りクラスタリングã¯ã€ã‚¯ãƒ©ã‚¹ã‚¿ãƒªãƒ³ã‚°ã‚’ã™ã‚‹éš›ã«åˆ¶ç´„を与ãˆã‚‹ã“ã¨ã§ç²¾åº¦ã‚’å‘上ã•ã›ã‚‹æ‰‹æ³•ã§ã™ã€‚制約ã¯2ã¤ã®ãƒ‡ãƒ¼ã‚¿é–“ã®é–¢ä¿‚を定義ã—ãŸã€ä»¥ä¸‹ãŒã‚ˆã使ã‚れるよã†ã§ã™ã€‚ must-link åŒã˜ã‚¯ãƒ©ã‚¹ã‚¿ã«æ‰€å±žã—ãªã‘ã‚Œã°ãªã‚‰ãªã„ cannot-link åŒã˜ã‚¯ãƒ©ã‚¹ã‚¿ã«æ‰€å±žã—ã¦ã¯ãªã‚‰ãªã„ 今回ã¯ã¨ã‚Šã‚ãˆãšåˆ¶ç´„付ãクラスタリングã®è«–æ–‡ã§å¤šã引用ã•ã‚Œã¦ã„ã¦ã€ä»¥ä¸‹ã®è«–文をå‚考ã«å®Ÿè£…ã—ã¦ã¿ã¾ã—ãŸã€‚手法ãŒK-meansã‚’å°‘ã—改良ã—ãŸã ã‘ã§ä¸€ç•ªç°¡å˜ãã†ã ã£ãŸã®ã¨ã€æœ€æ–°ã®å‹•å‘ã¾ã§èª¿ã¹ãã‚Œãªã‹ã£ãŸã®ã§ã€ã¾ãšã¯ã¨ã£ã‹ã‹ã‚Šã¨ã—ã¦ã€‚ "Constrained K-means Clustering with Background Knowledge", by Kiri Wagstaff, Claire Cardie, Seth Rogers, and Stefan Schroedl. ICML 20
トーナメントã¨å¤šå€¤åˆ†é¡ž - DO++
今やã£ã¦ã‚‹ç ”究ã§ã€ãƒˆãƒ¼ãƒŠãƒ¡ãƒ³ãƒˆå•é¡Œã‚’調ã¹ã‚‹æ©Ÿä¼šãŒã‚ã‚Šã¾ã—ãŸã€‚ トーナメントã¯ç§ã‚‚知らãªã‹ã£ãŸã®ã§ã™ãŒã€å‹è€…ã‚„é †ä½ã‚’決ã‚ã‚‹æ–¹å¼ã®ã“ã¨ã‚’指ã—ã€ã„ã‚ゆる二人ãšã¤æˆ¦ã£ã¦ç”Ÿã残ã£ã¦ã„ãæ–¹å¼ã¯ãƒŽãƒƒã‚¯ã‚¢ã‚¦ãƒˆãƒˆãƒ¼ãƒŠãƒ¡ãƒ³ãƒˆã¨ã„ã‚れるãã†ã§ã™(wikipedia)。 ＃10000人戦ã†æ™‚ã«ãƒŽãƒƒã‚¯ã‚¢ã‚¦ãƒˆãƒˆãƒ¼ãƒŠãƒ¡ãƒ³ãƒˆã§ã¯ä½•å›žè©¦åˆãŒè¡Œã‚れるã‹ã¨ã„ã†ã®ã¯ã‚ˆãã‚る質å•ã§ã™ã。 ã§ã€ã“ã®ãƒˆãƒ¼ãƒŠãƒ¡ãƒ³ãƒˆæ–¹å¼ã¨ã„ã†ã®ã¯èª¿ã¹ã¦ã¿ã‚‹ã¨éžå¸¸ã«æ§˜ã€…ãªã‚‚ã®ãŒã‚ã‚Šã¾ã™ 例ãˆã°ã‚¹ã‚¤ã‚¹å¼ãƒˆãƒ¼ãƒŠãƒ¡ãƒ³ãƒˆã¯ã€æœ€åˆã¯ãƒ©ãƒ³ãƒ€ãƒ ãªçµ„ã¿åˆã‚ã›ã§å¯¾æˆ¦ã€æ¬¡ã¯å‹è€…åŒå£«ã¨æ•—者åŒå£«ã€ãã®æ¬¡ã¯å…¨å‹ãƒ»ï¼‘å‹ï¼‘敗・２戦全敗ã®ãã‚Œãžã‚ŒãŒãƒ»ãƒ»ã¨ã„ã†ãµã†ã«åŒã˜æˆç¸¾ã®äººåŒå£«ã§æˆ¦ã†æ–¹å¼ã§ã™ã€‚レーティングを計算ã—ã¦ã€ãƒ¬ãƒ¼ãƒ†ã‚£ãƒ³ã‚°ãŒè¿‘ã„ã‚‚ã®åŒå£«ã‚’戦ã‚ã›ã‚‹ã‚ˆã†ãªæ‹¡å¼µã‚‚ã‚ã‚Šã¾ã™ã€‚è¿‘ã„ã®ã¯å°†æ£‹ã§ã‚„ã£ã¦ã‚‹ã‚ˆã†ãªã‚‚ã®ã§ã™ã。 利点ã¯å…¨ã¦ã®äººãŒåŒã˜è©¦åˆæ•°ã§æˆ¦ã„ã€ã¾ãŸåŽ³å¯†ãªé †ä½ãŒæ±ºã‚ã‚„ã™ã„ã“ã¨ãŒã‚ã‚Šã¾
1
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
{{#tags}}- {{label}}
{{/tags}}