[B! statistics2] nfunatoã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯

nfunato id:nfunato

statistics2ã«é–¢ã™ã‚‹nfunatoã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯ (60)

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

${{author_name}}$
{{author_name}}{{created}}
{{ #comment }}{{ comment }}{{ /comment }}
- {{ label }}

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

è—¤äº•å››æ®µã§å¦ã¶æœ€å°¤æŽ¨å®šã€MAPæŽ¨å®šã€ãƒ™ã‚¤ã‚ºæŽ¨å®š - Qiita
Deleted articles cannot be recovered. Draft of this article would be also deleted. Are you sure you want to delete this article? è—¤äº•å››æ®µã®é€£å‹ãŒæ¢ã¾ã‚‰ãªã„ã§ã™ãã€‚ 21æ—¥ã®å¯¾å±€ã«å‹åˆ©ã—ã¦ã€é€£å‹è¨˜éŒ²ã‚’ï¼‘ä½ã‚¿ã‚¤ã®28é€£å‹ã¾ã§ä¼¸ã°ã—ã¦ãã¾ã—ãŸã€‚26æ—¥ã®å¯¾å±€ã§å‹åˆ©ã™ã‚Œã°å˜ç‹¬ãƒˆãƒƒãƒ—ã«ãªã‚Šã¾ã™ã€‚ ãã‚“ãªè—¤äº•å››æ®µã®å¯¾æˆ¦æˆç¸¾ã¯28å‹0è² ã€‚å‹çŽ‡ã§ã„ã†ã¨1.000ã§ã™ã€‚ã‚¯ãƒ©ã‚¯ãƒ©ã™ã‚‹ã‚ˆã†ãªæˆç¸¾ã§ã™ãŒã€ã“ã®ã€Œå‹çŽ‡ã€ã¨ã¯ä½•ã‹ã‚’å°‘ã—æ•°å¦çš„ã«ã¿ã¦ã¿ã¾ã—ã‚‡ã†ã€‚ å˜ç´”ã«è¨€è‘‰ã ã‘ã‚’ã¿ã‚‹ã¨ã€Œè—¤äº•å››æ®µãŒå‹åˆ©ã™ã‚‹ç¢ºçŽ‡ã€ã§ã¯ãªã„ã‹ã¨è€ƒãˆã‚‰ã‚Œã¾ã™ã€‚ã¤ã¾ã‚Š $$P(\text{å‹åˆ©}\ |\ \text{è—¤äº•å››æ®µ}) = 1.0$$ã‹ã®ã‚ˆã†ã«æ„Ÿã˜ã¾ã™ã€‚ ã§ã¯ã“ã“ã§ã€26æ—¥ã®å¯¾å±€ã§è—¤äº•å››æ®µãŒå‹åˆ©ã™ã‚‹ç¢ºçŽ‡ã¯ã©ã‚Œã
nfunato 2017/06/28
Qiita

statistics2

shogi
ãƒªãƒ³ã‚¯
Amazon.co.jp: ç¾ä»£æ•°ç†çµ±è¨ˆå¦ã®åŸºç¤Žï¼ˆå…±ç«‹è¬›åº§ æ•°å¦ã®é…åŠ› 11ï¼‰: ä¹…ä¿å·é”ä¹Ÿ (è‘—), æ–°äº•ä»ä¹‹ (ç·¨é›†), å°æž—ä¿Šè¡Œ (ç·¨é›†), æ–Žè—¤æ¯… (ç·¨é›†), å‰ç”°æœ‹åºƒ (ç·¨é›†): æœ¬
nfunato 2017/04/13
book

statistics2
ãƒªãƒ³ã‚¯
Help Me Decide
Statistics Textbook Options Each textbook offers its own flavor. All can be used as part of an intro-level statistics course. For further considerations or to ask questions, see thisÂ ForumsÂ post.
nfunato 2017/04/09
ebook

statistics2
ãƒªãƒ³ã‚¯
Rã§ç•°å¸¸æ¤œçŸ¥(1): ã“ã‚Œã¾ã§è‡ªåˆ†ãŒã‚„ã£ã¦ããŸã“ã¨ã®ãŠã•ã‚‰ã„ - æ¸‹è°·é§…å‰ã§åƒããƒ‡ãƒ¼ã‚¿ã‚µã‚¤ã‚¨ãƒ³ãƒ†ã‚£ã‚¹ãƒˆã®ãƒ–ãƒã‚°
ã¶ã£ã¡ã‚ƒã‘ä»Šæ›´æ„ŸãŒãªãã‚‚ãªã„ã‚“ã§ã™ãŒã€å®Ÿã¯ã“ã‚Œã¾ã§è‡ªåˆ†ã§ã¯ã»ã¨ã‚“ã©ç•°å¸¸æ¤œçŸ¥ãƒ»å¤‰åŒ–æ¤œçŸ¥ã‚’ã‚´ãƒªã‚´ãƒªã‚„ã£ãŸã“ã¨ãŒãªã‹ã£ãŸãªãã¨æ€ã£ãŸã®ã§ã—ãŸã€‚ãã£ã‹ã‘ã¯ã€æ™‚ã€…è‰²ã€…ãªæ‰‹æ³•ã®ãƒ†ã‚¹ãƒˆã«ä½¿ã£ã¦ã„ã‚‹ã“ã®UCIæ©Ÿæ¢°å¦ç¿’ãƒªãƒã‚¸ãƒˆãƒªã®ãƒ‡ãƒ¼ã‚¿ã‚»ãƒƒãƒˆã€‚ UCI Machine Learning Repository: Water Treatment Plant Data Set ã“ã‚Œã¯èª¬æ˜Žã«ã‚‚ã‚ã‚‹ã‚ˆã†ã«ã€ã¨ã‚ã‚‹éƒ½å¸‚éƒ¨ã®ä¸‹æ°´å‡¦ç†å ´ã®æ§˜ã€…ãªã‚»ãƒ³ã‚µã‹ã‚‰ã®ã‚¤ãƒ³ãƒ—ãƒƒãƒˆã‚’åˆã‚ã›ã¦æ—¥æ¬¡ã§ã¾ã¨ã‚ãŸãƒ‡ãƒ¼ã‚¿ã‚»ãƒƒãƒˆã§ã™ã€‚ã“ã®æ‰‹ã®ãƒ—ãƒ©ãƒ³ãƒˆãƒ‡ãƒ¼ã‚¿ã‚»ãƒƒãƒˆã«ã‚ã‚ŠãŒã¡ãªèª²é¡ŒãŒã¾ã•ã«ãã®ç•°å¸¸æ¤œçŸ¥ã§ã€è¦ã¯ä½•ã‹ä¸å…·åˆãŒã‚ã£ãŸæ—¥ä»˜ã‚’äº‹å¾Œã§è‰¯ã„ã®ã§æ¤œå‡ºã—ãŸã„ã¨ã„ã†ãŠè©±ã§ã™ã€‚ ç•°å¸¸æ¤œçŸ¥è‡ªä½“ã¯ã€ä»¥å‰ã“ã®ãƒ–ãƒã‚°ã§ã•ã‚‰ã£ã¨ã ã‘å–ã‚Šä¸Šã’ãŸã“ã¨ãŒã‚ã‚Šã¾ã™ã€‚ãã‚Œã¯{AnomalyDetection}ãƒ‘ãƒƒã‚±ãƒ¼ã‚¸ã®ç´¹ä»‹è¨˜äº‹ã€‚ ã“ã®æ™‚ã¯{AnomalyDetection}ãŒä¾
nfunato 2017/01/12
R

statistics2
ãƒªãƒ³ã‚¯
å®Ÿè·µ çµ±è¨ˆãƒ¢ãƒ‡ãƒªãƒ³ã‚°å…¥é–€ ã€1. æ¦‚è¦ãƒ»ç›®æ¬¡ã€‘ - ã»ããç¬‘ã‚€
ã€å®£ä¼ã€‘2016/09/14 ã“ã®ãƒšãƒ¼ã‚¸ã«æ¥ãŸæ–¹ã¸ã€‚ã‚ãªãŸãŒæ±‚ã‚ã¦ã„ã‚‹æœ¬ã¯ã“ã‚Œã§ã™ã€‚ Stanã¨Rã§ãƒ™ã‚¤ã‚ºçµ±è¨ˆãƒ¢ãƒ‡ãƒªãƒ³ã‚° (Wonderful R) ä½œè€…: æ¾æµ¦å¥å¤ªéƒŽ,çŸ³ç”°åŸºåºƒå‡ºç‰ˆç¤¾/ãƒ¡ãƒ¼ã‚«ãƒ¼: å…±ç«‹å‡ºç‰ˆç™ºå£²æ—¥: 2016/10/25ãƒ¡ãƒ‡ã‚£ã‚¢: å˜è¡Œæœ¬ã“ã®å•†å“ã‚’å«ã‚€ãƒ–ãƒã‚° (10ä»¶) ã‚’è¦‹ã‚‹ã¾ãšã“ã‚Œã‚’äºˆç´„ã—ã¦ã‹ã‚‰ä¸‹è¨˜ã‚’èªã‚€ã¨ã„ã„ã§ã™ã€‚ ã€å®£ä¼çµ‚ã€‘ ã¯ã˜ã‚ã« çµ±è¨ˆãƒ¢ãƒ‡ãƒªãƒ³ã‚°ã¯ä»Šå¾Œã¾ã™ã¾ã™é‡è¦ã«ãªã£ã¦ãã‚‹æŠ€è¡“ã§ã™ã€‚ ç¾åœ¨ã€Web ä¸Šã«ã¯çµ±è¨ˆãƒ¢ãƒ‡ãƒªãƒ³ã‚°ã«é–¢ã™ã‚‹æ§˜ã€…ãªå„ªè‰¯è¨˜äº‹ãŒã‚ã‚Šã¾ã™ã€‚ ãã‚Œã‚‰ã®è¨˜äº‹ã¯ã€å®Œæˆã—ãŸãƒ¢ãƒ‡ãƒ«ã‚’ã‚¹ãƒžãƒ¼ãƒˆã«æç¤ºã—ã¦ã„ã‚‹ã‚‚ã®ãŒå¤šã„ã‚ˆã†ã§ã™ã€‚ ã—ã‹ã—ã€å®Ÿéš›ã®çµ±è¨ˆãƒ¢ãƒ‡ãƒªãƒ³ã‚°ã®ç¾å ´ã¯æ±ºã—ã¦ã‚¹ãƒžãƒ¼ãƒˆãªã‚‚ã®ã§ã¯ãªãã€æ§˜ã€…ãªæ³¥è‡ã„è©¦è¡ŒéŒ¯èª¤ã‚’è¡Œã„ãªãŒã‚‰åœ°é“ã«ãƒ¢ãƒ‡ãƒ«ã‚’æ§‹ç¯‰ã—ã¦ã„ãã¾ã™ã€‚ ã“ã®ä¸€é€£ã®è¨˜äº‹ã§ã¯ã€æœ€çµ‚çš„ãªãƒ¢ãƒ‡ãƒ«ã®å®Œæˆå½¢ã‚’ã„ããªã‚Šæç¤ºã™ã‚‹ã®ã§ã¯ãªãã€ãƒ¢ãƒ‡ãƒ«æ§‹ç¯‰ã®æ‰‹é †ã‚’ã‚¹ãƒƒãƒ†ãƒ—ãƒã‚¤ã‚¹
nfunato 2017/01/05
statistics2
ãƒªãƒ³ã‚¯
Kerasã«ã‚ˆã‚‹2ã‚¯ãƒ©ã‚¹ãƒã‚¸ã‚¹ãƒ†ã‚£ãƒƒã‚¯å›žå¸° - äººå·¥çŸ¥èƒ½ã«é–¢ã™ã‚‹æ–å‰µéŒ²
ã¾ãšã¯ã‚‚ã£ã¨ã‚‚ç°¡å˜ãª2ã‚¯ãƒ©ã‚¹ãƒã‚¸ã‚¹ãƒ†ã‚£ãƒƒã‚¯å›žå¸°ãƒ¢ãƒ‡ãƒ«ã‚’Kerasã§æ›¸ã„ã¦ã¿ã‚‹ã€‚å‰ã«Theanoã§ã‚„ã£ãŸï¼ˆ2015/5/19ï¼‰ã®ã‚’Kerasã§æ›¸ãæ›ãˆãŸã ã‘ã€‚ãƒã‚¸ã‚¹ãƒ†ã‚£ãƒƒã‚¯å›žå¸°ã¯ã€å›žå¸°ã¨ã¤ãã‘ã©åˆ†é¡žã®ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã§ã€éš ã‚Œå±¤ãŒãªãã€æ´»æ€§åŒ–é–¢æ•°ã«ã‚·ã‚°ãƒ¢ã‚¤ãƒ‰é–¢æ•°ã‚’ä½¿ã£ãŸãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆã¨ã—ã¦ãƒ¢ãƒ‡ãƒ«åŒ–ã§ãã‚‹ã€‚ ãƒ‡ãƒ¼ã‚¿ã¯ã€PRMLã®4ç« ã®ex2data1 ã‚’ä½¿ã†ã€‚1åˆ—ç›®ã¨2åˆ—ç›®ãŒãƒ‡ãƒ¼ã‚¿ã§3åˆ—ç›®ãŒã‚¯ãƒ©ã‚¹ãƒ©ãƒ™ãƒ«ï¼ˆ0ã¾ãŸã¯1ã®2ã‚¯ãƒ©ã‚¹ï¼‰ã€‚ ã‚½ãƒ¼ã‚¹ã‚³ãƒ¼ãƒ‰ï¼šex2data1.py ãƒ‡ãƒ¼ã‚¿ã®ãƒãƒ¼ãƒ‰ã¨æ£è¦åŒ– ãƒ‡ãƒ¼ã‚¿ã‚’èªã¿è¾¼ã‚€ãƒ©ã‚¤ãƒ–ãƒ©ãƒªã«ã¯pandasãªã©ã‚‚ã‚ã‚‹ãŒã€ã“ã“ã§ã¯numpy.genfromtxt()ã‚’ä½¿ã†ã€‚XãŒãƒ‡ãƒ¼ã‚¿ã§äºŒæ¬¡å…ƒãƒ‡ãƒ¼ã‚¿ã®ãƒªã‚¹ãƒˆã€tãŒãƒ©ãƒ™ãƒ«ã§0ã¾ãŸã¯1ã®ãƒªã‚¹ãƒˆã€‚ # load training data data = np.genfromtxt(os.path.join('data', 'ex2d
nfunato 2016/10/31
statistics2
ãƒªãƒ³ã‚¯
éšŽå±¤ãƒ™ã‚¤ã‚ºãƒ¢ãƒ‡ãƒ«ã¨WAIC - StatModeling Memorandum
ã“ã®è¨˜äº‹ã§ã¯éšŽå±¤ãƒ™ã‚¤ã‚ºãƒ¢ãƒ‡ãƒ«ã®å ´åˆã®WAICã¨ã¯ä½•ã‹ã€ã¾ãŸãã®å ´åˆã®WAICã®é«˜é€Ÿãªç®—å‡ºæ–¹æ³•ã«ã¤ã„ã¦æ›¸ãã¾ã™ã€‚ èƒŒæ™¯ ä»¥ä¸‹ã®2ã¤ã®è³‡æ–™ã‚’å‚ç…§ã—ã¦ãã ã•ã„ã€‚[1]ã«äºŒç¨®é¡žã®å®Ÿè£…ãŒè¼‰ã£ã¦ã„ã¾ã™ã€‚[2]ã«æ˜Žå¿«ãªç†è«–çš„è£œè¶³ãŒè¼‰ã£ã¦ã„ã¾ã™ã€‚ [1] éšŽå±¤ãƒ™ã‚¤ã‚ºã¨WAIC (æ¸…æ°´å…ˆç”Ÿã®è³‡æ–™ã§ã™ã€slideshare) [2] éšŽå±¤ãƒ™ã‚¤ã‚ºæ³•ã¨WAIC (æ¸¡è¾ºå…ˆç”Ÿã®è³‡æ–™ã§ã™ã€pdf, html) ãƒ¢ãƒ‡ãƒ«1 è³‡æ–™[1]ã«ã‚ã‚‹ãƒ¢ãƒ‡ãƒ«ã‚’æ‰±ã„ã¾ã™ã€‚ã™ãªã‚ã¡ã€ ã“ã“ã§ã¯äººæ•°ã€ã¯äººã®ã‚¤ãƒ³ãƒ‡ãƒƒã‚¯ã‚¹ã§ã™ã€‚ã¯å€‹äººå·®ã‚’è¡¨ã™å€¤ã«ãªã‚Šã¾ã™ã€‚ã“ã®ãƒ¢ãƒ‡ãƒ«ã«ãŠã„ã¦ã¯ã‚’è§£æžçš„ã«ç©åˆ†æ¶ˆåŽ»ã™ã‚‹ã“ã¨ãŒã§ãã¦ã€è² ã®äºŒé …åˆ†å¸ƒã‚’ä½¿ã†ä»¥ä¸‹ã®ãƒ¢ãƒ‡ãƒ«å¼ã¨ç‰ä¾¡ã«ãªã‚Šã¾ã™ã€‚ ã“ã“ã§ã¯äºˆæ¸¬ã¨ã—ã¦ï¼ˆWAICã¨ã—ã¦ï¼‰2é€šã‚Šè€ƒãˆã¦ã¿ã¾ã—ã‚‡ã†ã€‚ ä»¥é™ã§ã¯äº‹å¾Œåˆ†å¸ƒã«ã‚ˆã‚‹å¹³å‡ã‚’ã€åˆ†æ•£ã‚’ã¨æ›¸ãã“ã¨ã«ã—ã¾ã™ã€‚ (1) ã‚’æŒã¤ãŒã€è¿½åŠ ã§æ–°ã—ã1ã¤ã®ã‚µãƒ³ãƒ—ãƒ«ã‚’å¾—ã‚‹å ´åˆ ã“ã®å ´åˆã«
nfunato 2016/09/06
statistics2
ãƒªãƒ³ã‚¯
Amazon.co.jp: Machine Learning: A Probabilistic Perspective (Adaptive Computation and Machine Learning series): Murphy, Kevin P.: æœ¬
nfunato 2016/08/15
book

statistics2

machinelearning
ãƒªãƒ³ã‚¯
Markov Logic Network (1) - Qiita
Markov Logic Network(MLN) ã¨ã„ã†ãƒ¢ãƒ‡ãƒ«ã«ã¤ã„ã¦å‹‰å¼·ã—ãŸéš›ã®ãƒŽãƒ¼ãƒˆã§ã™ã€‚èª¤ã‚ŠãŒã‚ã‚‹ã‹ã‚‚ã—ã‚Œã¾ã›ã‚“ã®ã§ã”æ³¨æ„ä¸‹ã•ã„ã€‚ã¨ã‚Šã‚ãˆãšä½¿ã†ç‚ºã«ã¯Alchemyã®ã‚ˆã†ãªã‚½ãƒ•ãƒˆã‚¦ã‚§ã‚¢ã‚‚ã‚ã‚Šã¾ã™ã€‚Matthew Richardsonã¨Pedro Domingosã®è«–æ–‡ã‚’èªã¿ãŸã„ã¨æ€ã„ã¾ã™ã€‚ æ¦‚è¦ Markov Logic Network $L$ ã¨ã¯ ä¸€éšŽè¿°èªžè«–ç†ã®è«–ç†å¼ $F_i$ ã¨å®Ÿæ•°å€¤ã®é‡ã¿ $w_i$ ã®çµ„ $(F_i, w_i)$ ã®é›†åˆ ã®äº‹ã€‚ä¾‹ãˆã°ä¸‹ã®ã‚ˆã†ãªã‚‚ã®ã€‚ $\forall x(\rm{Smokes}(x)\Rightarrow \rm{Cancer}(x)): 1.5$ $\forall x\forall y(\rm{Friends}(x,y)\Rightarrow (\rm{Smokes}(x)\Leftrightarrow\rm{Smokes}(y))
nfunato 2016/07/16
statistics2

logic
ãƒªãƒ³ã‚¯
ã€Œpå€¤ã‚„æœ‰æ„æ€§ã«æ‹˜ã‚ŠéŽãŽã‚‹ãªã€p < 0.05ã‹ã©ã†ã‹ãŒå…¨ã¦ã‚’æ±ºã‚ã‚‹æ™‚ä»£ã¯ã‚‚ã†çµ‚ã‚ã‚‰ã›ã‚ˆã†ã€ã¨ã„ã†ã‚¢ãƒ¡ãƒªã‚«çµ±è¨ˆå¦ä¼šã®å£°æ˜Žã«å¯¾ã™ã‚‹æŸãƒ‡ãƒ¼ã‚¿ã‚µã‚¤ã‚¨ãƒ³ãƒ†ã‚£ã‚¹ãƒˆã®ãƒ–ãƒã‚°ã®å£°æ˜Žã«å¯¾ã™ã‚‹ç´ äºº()ã®å£°æ˜Ž - é©šç•°ã®ã‚¢ãƒ‹ãƒ²ã‚¿ç¤¾ä¼šå¾©å¸°ã®äºˆå‚™
ASAã®å£°æ˜Žã¨ãã®ãƒ—ãƒ¬ã‚¹ãƒªãƒªãƒ¼ã‚¹(100%äºˆæ¸¬ã§ã¯ãªã„)ãŒè©±é¡Œã«ãªã£ã¦ã„ã‚‹ã€‚ è‹±èªžè‡ªä½“ã¯å¹³æ˜“ã ãŒé¢å€’ãªã‚‰ã°æŸãƒ‡ãƒ¼ã‚¿ã‚µã‚¤ãƒ¤äººãŒæ—¥æœ¬èªžã§è¨˜äº‹ã‚’æ›¸ã„ã¦ã„ã‚‹ã€‚ å°‘ãªãã¨ã‚‚ã€ã“ã®è¨˜äº‹ã¨ã„ã†ã‹på€¤ã«ã¤ã„ã¦å€‹äººçš„è¦‹è§£ã‚’è¨€ã†ãªã‚‰ã°ã€ Scientific conclusions and business or policy decisions should not be based only on whether a p-value passes a specific threshold. ã¨ã„ã†ã®ã¯éžå¸¸ã«ã‚ˆãã‚ã‹ã‚‹ã®ã ãŒã€ã ã‹ã‚‰ã¨è¨€ã£ã¦på€¤ã¯ã‚¯ã‚½!!ã„ã¾ã®æ™‚ä»£ã¯ãƒ¢ãƒ‡ãƒªãƒ³ã‚°!!(è¨€ã„éŽãŽ)ã¨ã„ã†ã®ã¯ã€ãŸã¶ã‚“ã€è‡ªåˆ†ã®åˆ†é‡Žã§ã¯ã¾ã på€¤ã¯ãƒ¢ãƒ‡ãƒªãƒ³ã‚°ã¨ã‹ä»–ã®ãªã‚“ã‹ã™ã”ã„ã‚„ã¤ã«ã¯ã¨ã£ã¦å¤‰ã‚ã‚‰ãšã€ã“ã®å…ˆ15å¹´ã¯på€¤ãŒä¸»æµã§ã¯ãªã„ã‹ã¨æ€ã†ã€‚ ã¨ã„ã†ã®ã‚‚ã€ã‚„ã¯ã‚Špå€¤ãŒã“ã“ã¾ã§ä¸»æµãªã®ã¯ã€på€¤ã¨ã„ãˆã°ã¾ã‚ãã‚ŒãŒä½•ã‚’æ„å‘³ã—ã¦
nfunato 2016/06/11
statistics2
ãƒªãƒ³ã‚¯
CS 281B / Stat 241B: Home
People Professor: Michael Jordan ([email protected]) Office: 401 Evans, 2-8660; 731 Soda, 2-3806 Office hours: Tues, 3-4 (401 Evans); Thurs 1-2 (731 Soda) TA: Chao Chen ([email protected]) Office: 385 Evans Office hours: Mon, 1-2; Weds, 1-2 Course Description: This course will provide an introduction to advanced statistical and computational methods for the modeling of complex, multi
nfunato 2016/06/06
statistics2
ãƒªãƒ³ã‚¯
Google
ä¸–ç•Œä¸ã®ã‚ã‚‰ã‚†ã‚‹æƒ…å ±ã‚’æ¤œç´¢ã™ã‚‹ãŸã‚ã®ãƒ„ãƒ¼ãƒ«ã‚’æä¾›ã—ã¦ã„ã¾ã™ã€‚ã•ã¾ã–ã¾ãªæ¤œç´¢æ©Ÿèƒ½ã‚’æ´»ç”¨ã—ã¦ã€ãŠæŽ¢ã—ã®æƒ…å ±ã‚’è¦‹ã¤ã‘ã¦ãã ã•ã„ã€‚
nfunato 2016/02/20
"KL divergence"

statistics2
ãƒªãƒ³ã‚¯
A Tutorial on Python - Statistics Views
nfunato 2016/02/19
Jupyter

statistics2
ãƒªãƒ³ã‚¯
ã€Minibatch SGDã€‘Goã§ç¢ºçŽ‡çš„å‹¾é…é™ä¸‹æ³•ã‚’æ›¸ã„ã¦ã¿ãŸ |
ç¢ºçŽ‡çš„å‹¾é…é™ä¸‹æ³•ã‚’æ›¸ã“ã†ã¨æ€ã„, å¦ç¿’ã®å‚è€ƒã¨ãªã‚‹å®Ÿè£…ã‚’ Webã§æŽ¢ã—ã¦ã¿ã‚‹ã¨ Pythonå®Ÿè£…ãŒå¤šã, ä¸ã§ã‚‚ NumPy, Matplotlibã‚’ä½¿ã£ã¦ã„ã‚‹ä¾‹ãŒå¤šã‹ã£ãŸã€‚[1] ã›ã£ã‹ããªã®ã§ Go (1.4.1)ã§æ›¸ã„ã¦ã¿ã‚ˆã†ã¨æ€ã„ Goã§ NumPyã«ã‚ãŸã‚‹è¡Œåˆ—è¨ˆç®—ãƒ‘ãƒƒã‚±ãƒ¼ã‚¸ã¨ Matplotlibã«ã‚ãŸã‚‹ãƒ—ãƒãƒƒãƒˆç³»ã®ãƒ‘ãƒƒã‚±ãƒ¼ã‚¸ã‚’æŽ¢ã—ã¦ã„ãŸã¨ã“ã‚, gonum/matrix ã¨ gonum/plotãŒ, è‰¯ã•ãã†ã§è§¦ã£ã¦ã¿ãŸã€‚ gonum/matrix gonum/matrix ã¯Goã®è¡Œåˆ—è¨ˆç®—ãƒ‘ãƒƒã‚±ãƒ¼ã‚¸ã€‚å†…éƒ¨çš„ã«ã¯ blas (Basic Linear Algebra Subprograms)ã‚’ä½¿ã£ã¦ã„ã‚‹ã€‚ numpy ã¨ã®å¯¾å¿œé–¢ä¿‚ã‚’æŽ¢ã™ã®ã«æ™‚é–“ãŒã‹ã‹ã£ã¦ã—ã¾ã£ãŸã€‚Dense, Matrix, Vectorã‚ãŸã‚Šã®åž‹ã‚’è¦šãˆãªã„ã¨ä¸Šæ‰‹ãä½¿ã„ã“ãªã›ãªã„ã‹ã‚‚ã€‚ è¡Œåˆ—ç©, è¦ç´ ç©, å†…ç©
nfunato 2016/01/14
golang

machinelearning

statistics2
ãƒªãƒ³ã‚¯
ã€Minibatch SGDã€‘ç¢ºçŽ‡çš„å‹¾é…é™ä¸‹æ³•ã‚’ goroutine ã§ä¸¦è¡ŒåŒ–ã—ã¦ã¿ã‚‹ |
å‰ã«æ›¸ã„ãŸã€MSGDã€‘Goã§ç¢ºçŽ‡çš„å‹¾é…é™ä¸‹æ³•ã‚’æ›¸ã„ã¦ã¿ãŸã€gonum/matrixã€‘ ã®ç¶šãã§ goroutine ã§ MSGD (Mini-batch Stochastic Gradient Descent)ã‚’ä¸¦è¡Œå‡¦ç†åŒ–ã—ã¦ã¿ã¾ã—ãŸã€‚ å‚è€ƒã«ã•ã›ã¦é ‚ã„ãŸæ›¸ç±ã¯, ã‚ªãƒ³ãƒ©ã‚¤ãƒ³æ©Ÿæ¢°å¦ç¿’ (æ©Ÿæ¢°å¦ç¿’ãƒ—ãƒãƒ•ã‚§ãƒƒã‚·ãƒ§ãƒŠãƒ«ã‚·ãƒªãƒ¼ã‚º)ã§ã™ã€‚ ç¢ºçŽ‡çš„å‹¾é…é™ä¸‹æ³•ã®ä¸¦åˆ—åŒ– å‰å›žã¯ Minibatch ã«åˆ†å‰²ã—ã¦ Minibatchå˜ä½ã§ ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ã‚’æ›´æ–°ã™ã‚‹ MSGD ã‚’æ›¸ã„ãŸã€‚ ç¢ºçŽ‡çš„å‹¾é…é™ä¸‹æ³• (ã«é™ã‚‰ãšã‚ªãƒ³ãƒ©ã‚¤ãƒ³å¦ç¿’ã«é–¢ã™ã‚‹æœ€é©åŒ–æ‰‹æ³•?) ã®ä¸¦åˆ—åŒ–ã«ã¤ã„ã¦, å¤§ã¾ã‹ã«ä¸‹è¨˜ã®3ã¤ã®ã‚¢ãƒ—ãƒãƒ¼ãƒãŒã‚ã‚‹ã€‚ Bulk Synchronous Parallelï¼ˆBSP) : å„ãƒ—ãƒã‚»ã‚¹ãŒ1å›žåŒæœŸã™ã‚‹ãŸã³ã«å…¨ä½“ã‚’åŒæœŸã™ã‚‹ Parallel SGD : å„ãƒ—ãƒã‚»ã‚¹ã§å¦ç¿’ã—, æœ€å¾Œã«1åº¦ã ã‘ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ãƒ™ã‚¯ãƒˆãƒ«ã®å¹³å‡ã‚’è¨ˆ
nfunato 2016/01/14
golang

machinelearning

statistics2
ãƒªãƒ³ã‚¯
çµ±è¨ˆå¦ã®åŸºç¤Žã®åŸºç¤Ž
2. è‡ªå·±ç´¹ä»‹ ãƒ»Facebookãƒšãƒ¼ã‚¸ https://www.facebook.com/matsukenbook ãƒ»Twitterã‚¢ã‚«ã‚¦ãƒ³ãƒˆ @kenmatsu4 ãƒ»Qiitaã§ãƒ–ãƒã‚°ã‚’æ›¸ã„ã¦ã„ã¾ã™ï¼ˆçµ±è¨ˆã€æ©Ÿæ¢°å¦ç¿’ã€Pythonç‰ï¼‰ http://qiita.com/kenmatsu4 ï¼ˆ3500 contributionã‚’è¶…ãˆã¾ã—ãŸï¼ï¼‰ ãƒ»è¶£å‘³ - ãƒãƒ³ãƒ‰ã§ãƒ™ãƒ¼ã‚¹ã‚’å¼¾ã„ãŸã‚Šã—ã¦ã„ã¾ã™ã€‚ - ä¸»ã«æ±å—ã‚¢ã‚¸ã‚¢ã¸ãƒãƒƒã‚¯ãƒ‘ãƒƒã‚¯æ—…è¡Œã«è¡Œã£ãŸã‚Šã—ã¾ã™ ï¼ˆã‚«ãƒ³ãƒœã‚¸ã‚¢ã€ãƒŸãƒ£ãƒ³ãƒžãƒ¼ã€ãƒãƒ³ã‚°ãƒ©ãƒ‡ã‚·ãƒ¥ã€æ–°ç–†ã‚¦ã‚¤ã‚°ãƒ«è‡ªæ²»åŒº etcï¼‰ æ—…è¡Œã®å†™çœŸ : http://matsu-ken.jimdo.com Twitterã‚¢ã‚¤ã‚³ãƒ³
nfunato 2015/12/08
slideshare

statistics2
ãƒªãƒ³ã‚¯
ã€Œå†…ç©ãŒè¦‹ãˆã‚‹ã¨çµ±è¨ˆå¦ã‚‚è¦‹ãˆã‚‹ã€ç¬¬5å›ž ãƒ—ãƒã‚°ãƒ©ãƒžã®ãŸã‚ã®æ•°å¦å‹‰å¼·ä¼š ç™ºè¡¨è³‡æ–™
ã€Œç¬¬5å›ž ãƒ—ãƒã‚°ãƒ©ãƒžã®ãŸã‚ã®æ•°å¦å‹‰å¼·ä¼š ç™ºè¡¨è³‡æ–™ (2015/11/21[sat])ã€ å†…å®¹ã¯çµ±è¨ˆå¦ã®ç´ é¤ŠãŒã‚ã‚‹æ–¹ã«ã¯åŸºæœ¬çš„ãªäº‹é …ã§ã™ãŒã€ãƒ™ã‚¯ãƒˆãƒ«ã¨å†…ç©ã§è¦‹æ–¹ã‚’å¤‰ãˆã¦ã¿ãŸã¨ã„ã†ç‚¹ã¨ã€ã‚ã¾ã‚Šçµ±è¨ˆå¦ã«è¦ªã—ã¿ãŒãªã„æ–¹ã«ã‚‚ç†è§£ã—ã¦ã‚‚ã‚‰ãˆã‚‹ã‚ˆã†ãªã¾ã¨ã‚ã«ãªã£ã¦ã„ã‚‹ã€ã¨ã„ã†ã¨ã“ã‚ã«æœ¬ã‚¹ãƒ©ã‚¤ãƒ‰ã®ç‹¬è‡ªæ€§ãŒã‚ã‚‹ã¨è€ƒãˆã¦ã„ã¾ã™ã®ã§ã€ãã®è¾ºã‚Šè‰¯ã‘ã‚Œã°ã”è¦§ãã ã•ã„ï¼¾ï¼¾Read less
nfunato 2015/12/08
slideshare

statistics2
ãƒªãƒ³ã‚¯
å±€æ‰€çš„ãªè¿‘ä¼¼ã«ã‚ˆã‚‹MCMCã®é«˜é€ŸåŒ–è«–æ–‡ã‚’èªã‚“ã - xiangze's sparse blog
å±€æ‰€çš„ãªè¿‘ä¼¼ã«ã‚ˆã‚‹MCMCã®é«˜é€ŸåŒ–è«–æ–‡ãŒè©±é¡Œã«ãªã‚Šã¾ã—ãŸã€‚ http://japan.zdnet.com/article/35073667/ è«–æ–‡Accelerating Asymptotically Exact MCMC for Computationally Intensive Models via Local Approximations ã‚’èªã‚“ã§ç†è§£ã—ãŸç¯„å›²ã®å†…å®¹ã‚’æ›¸ãã¾ã™ã€‚ ã‚¢ã‚¤ãƒ‡ã‚¢ ç‰©ç†ç¾è±¡ã®ãƒ¢ãƒ‡ãƒ«ã¯(ç¢ºçŽ‡)å¾®åˆ†æ–¹ç¨‹å¼ã‚’ä½¿ã£ã¦æ›¸ã‹ã‚Œã¾ã™ãŒã€ãã®ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ãŒæœªçŸ¥ã®ã“ã¨ãŒå¤šãã€MCMCã‚’ä½¿ã£ã¦ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ã‚’æŽ¨å®šã™ã‚‹ã“ã¨ãŒè¡Œã‚ã‚Œã¦ã„ã¾ã™ã€‚ ã—ã‹ã—ç¾å®Ÿçš„ãªå•é¡Œã§ã¯ãƒ¢ãƒ‡ãƒ«ãŒå·¨å¤§ã«ãªã£ã¦ã€MCMCã®å„ã‚¹ãƒ†ãƒƒãƒ—ã§ãã‚Œã‚’è©•ä¾¡ã™ã‚‹ã®ã¯è¨ˆç®—é‡ãŒå¤šããªã‚Šã™ãŽã¦ã—ã¾ã„ã¾ã™ã€‚ ãã“ã§ãƒ¢ãƒ‡ãƒ«ã‚’è¿‘ä¼¼ã™ã‚‹ã¨ã„ã†ã“ã¨ãªã®ã§ã™ãŒã€ã‚µãƒ³ãƒ—ãƒªãƒ³ã‚°ã«è¿‘ä¼¼ã—ãŸäº‹å¾Œåˆ†å¸ƒã‚’ä½¿ã†ã¨ã„ã†ã‚¢ã‚¤ãƒ‡ã‚¢ã¯Approximate Bayes
nfunato 2015/11/21
statistics2

python
ãƒªãƒ³ã‚¯
ç·‘æœ¬ in Python (1): ç¬¬2ç« - tnoda-journal
ä»–ã«ã‚‚å¤šãã®åŒæ§˜ã®ãƒ–ãƒã‚°ãƒã‚¹ãƒˆãŒã‚ã‚‹ã®ã§ä½•ç•ªç…Žã˜ã‹åˆ†ã‹ã‚Šã¾ã›ã‚“ãŒï¼Œä¹…ä¿å…ˆç”Ÿã®ç·‘æœ¬ ã‚’ Python ã§æ›¸ãè‡ªä¸»ç·´ã‚’å§‹ã‚ã‚‹ã“ã¨ã«ã—ã¾ã—ãŸã€‚ç¬¬2ç« ã‹ã‚‰å§‹ã‚ã¾ã™ã€‚ GitHub ãƒªãƒã‚¸ãƒˆãƒªã® URL ã¯ https://github.com/tnoda/midorihon-in-python ã§ã™ã€‚ ç¬¬2ç« ç¢ºçŽ‡åˆ†å¸ƒã¨çµ±è¨ˆãƒ¢ãƒ‡ãƒ«ã®æœ€å°¤æŽ¨å®šPandas ä½¿ã„ã¾ã™ã€‚ %matplotlib inline import matplotlib.pyplot as plt plt.style.use('ggplot') from numpy.random import randn import numpy as np from pandas import Series, DataFrame import pandas as pd pd.__version__ u'0.17.0' .RData ãƒ‡ãƒ¼ã‚¿ã®èªã¿
nfunato 2015/11/13
book

midoribon

statistics2
ãƒªãƒ³ã‚¯
ç¢ºçŽ‡çš„å‹¾é…é™ä¸‹æ³•ã‚’æœ€é©åŒ–ã«ä½¿ãˆã‚‹{sgd}ãƒ‘ãƒƒã‚±ãƒ¼ã‚¸ã‚’ä¸€é€šã‚Šè©¦ã—ã¦ã¿ã‚‹ - æ¸‹è°·é§…å‰ã§åƒããƒ‡ãƒ¼ã‚¿ã‚µã‚¤ã‚¨ãƒ³ãƒ†ã‚£ã‚¹ãƒˆã®ãƒ–ãƒã‚°
ã¡ã‚‡ã£ã¨å‰ã«è©±é¡Œã«ãªã£ã¦ãŸã‚“ã§ã™ãŒã€ä½•ã§ã‚‚CRANã«ç¢ºçŽ‡çš„å‹¾é…é™ä¸‹æ³•(Stochastic Gradient Descent)ã‚’å®Ÿè£…ã—ãŸ{sgd}ã¨ã„ã†ãƒ‘ãƒƒã‚±ãƒ¼ã‚¸ãŒå…¬é–‹ã•ã‚Œã¦ã„ã‚‹ãã†ã§ã€‚JSSæŽ²è¼‰äºˆå®šã®Vignetteã‚‚ã‚ã‚‹ã¿ãŸã„ã§ã™ã€‚ CRAN - Package sgd Stochastic gradient decent methods for estimation with large data sets ã¨ã„ã†ã“ã¨ã§ã€ãƒŸãƒ¼ãƒãƒ¼ã‹ã‚‚ã§ã™ãŒè©¦ã—ã«ã¡ã‚‡ã£ã¨è§¦ã£ã¦ã¿ã‚ˆã†ã‹ã¨æ€ã„ã¾ã™ã€‚ã¡ãªã¿ã«ä»Šå›žæ¤œè¨¼ã«ç”¨ã„ãŸãƒžã‚·ãƒ³ã¯Windows 7 (64bit)ã€8ã‚³ã‚¢ (3.60GHz)ã€ãƒ¡ãƒ¢ãƒª32GBã¨ã„ã†ã‚¹ãƒšãƒƒã‚¯ã§ã™ã€‚ ç¢ºçŽ‡çš„å‹¾é…é™ä¸‹æ³•ã«ã¤ã„ã¦ æœ€è¿‘ã¯MLPã‚·ãƒªãƒ¼ã‚ºã¨ã‹è‰¯ã„ãƒ†ã‚ã‚¹ãƒˆãŒæ²¢å±±å‡ºã¦ã„ã‚‹ã®ã§ã€ãã¡ã‚‰ã§ã‚‚èªã‚“ã§ã¿ã¦ä¸‹ã•ã„*1ã€‚ ã‚ªãƒ³ãƒ©ã‚¤ãƒ³æ©Ÿæ¢°å¦ç¿’ (æ©Ÿæ¢°å¦ç¿’ãƒ—ãƒãƒ•ã‚§ãƒƒã‚·ãƒ§ãƒŠãƒ«ã‚·ãƒªãƒ¼ã‚º) ä½œ
nfunato 2015/11/09
machinelearning

statistics2
ãƒªãƒ³ã‚¯
1 2 3 æ¬¡ã®ãƒšãƒ¼ã‚¸