スライド 1ノンパラメトリックベイズ モデルã®è¤‡é›‘ã•ãŒä¸æ˜Žãªå ´åˆ ã“ã‚Œã¾ã§ã«èª¬æ˜Žã—ã¦ããŸK-meansã€EMã€å¤‰åˆ†ãƒ™ã‚¤ã‚ºãªã©ã¯ã€ モデルã®è¤‡é›‘ã•ã€ãŸã¨ãˆã°ã‚¯ãƒ©ã‚¹ã‚¿ãƒªãƒ³ã‚°ã«ãŠã‘るクラスタ数 ã¯äºˆã‚分ã‹ã£ã¦ã„ã‚‹ã¨ã—ã¦ãƒ¢ãƒ‡ãƒ«æŽ¨å®šã—ãŸã€‚ ã—ã‹ã—ã€å®Ÿéš›ã¯ã‚¯ãƒ©ã‚¹ã‚¿æ•°ãŒä¸æ˜Žã®å ´åˆãŒå¤šã„。 ノンパラメトリックã¨ã¯ 観測データã«å¿œã˜ã¦ãƒ¢ãƒ‡ãƒ«è‡ªä½“ã®è¤‡é›‘ã•ã‚‚å¦ ç¿’ã™ã‚‹ クラスタリングã®å ´åˆã¯ã€ã‚¯ãƒ©ã‚¹ã‚¿æ•°ãŒäºˆã‚分ã‹ã£ ã¦ã„ãªã„å ´åˆã€‚観測データã«é©ã—ãŸã‚¯ãƒ©ã‚¹ã‚¿æ•°ã‚‚ 推定 データã‹ã‚‰å¦ç¿’ã™ã‚‹ã¨ã„ã†ã“ã¨ã®ç›´è¦³  有é™å€‹ã®æ£è¦åˆ†å¸ƒã‚’é…分比πkã§ã®æ··åˆãƒ¢ãƒ‡ãƒ«:å¼(NP1)  クラスタ数Kã®å€¤ã‚’観測データã‹ã‚‰æœ€é©åŒ–ã™ã‚‹ã“ã¨ã«ã‚ˆã£ ã¦æŽ¨å®šã™ã‚‹ã€‚  基本的アイデア：無é™æ¬¡å…ƒã®é€£ç¶šåˆ†å¸ƒã‹ã‚‰è¦³æ¸¬ãƒ‡ãƒ¼ã‚¿ ã«é©å¿œã—ãŸæœ‰é™æ¬¡å…ƒã®é›¢æ•£åˆ†å¸ƒã‚’å¦ç¿’ã™ã‚‹ã€‚ ( ) ( )∑= = K k kk NPxNxp 1 )1(|,| θππθ ç„¡é™æ¬¡
chineserestaurantprocess.dvi
Chinese Restaurant Processes
A Simple Proof of the Stick-Breaking Construction of the Dirichlet Process John Paisley Department of Computer Science Princeton University, Princeton, NJ 
http://www.stats.bris.ac.uk/~peter/papers/GreenCDP.pdf
Introduction to the Dirichlet Distribution and Related Processes Bela A. Frigyik, Amol Kapila, and Maya R. Gupta Department of Electrical Engineering University of Washington Seattle, WA 98195 {gupta}@ee.washington.edu ElectricalElectrical EngineeringEngi
ç†ç³»å¥³åマンガ「ç†ç³»ãªã‚«ãƒŽã‚¸ãƒ§ã€
xargs を使ã£ã¦ã‚«ã‚¸ãƒ¥ã‚¢ãƒ«ã«ä¸¦åˆ—å‡¦ç† - ãŸã”ã‚‚ã‚Šã™ãƒ¡ãƒ¢
https://people.csail.mit.edu/tommi/papers/SonGloJaa_optbook.pdf
2007-12-30 - ã¯ã‚‹ã‹ã•ã‚“ï¼ ã¯ã¦ãªãƒ€ã‚¤ã‚¢ãƒªãƒ¼
Google I/O 2011: Python@Google
{{#tags}}- {{label}}
{{/tags}}