DDSketch: A fast and fully-mergeable quantile sketch with relative-error guarantees
Summary statistics such as the mean and variance are easily maintained for large, distributed data streams, but order statistics (i.e., sample quantiles) can only be approximately summarized. There is extensive literature on maintaining quantile sketches where the emphasis has been on bounding the rank error of the sketch while using little memory. Unfortunately, rank error guarantees do not precl
)
Moment (mathematics) - Wikipedia
In mathematics, the moments of a function are certain quantitative measures related to the shape of the function's graph. If the function represents mass density, then the zeroth moment is the total mass, the first moment (normalized by total mass) is the center of mass, and the second moment is the moment of inertia. If the function is a probability distribution, then the first moment is the expe
A/Bテストã«ç”¨ã„られる統計的検定手法（ãƒã‚¸ãƒƒã‚¯ï¼‰ã®ã¾ã¨ã‚＆比較 | RCO Ad-Tech Lab Blog
リクルートデータ組織ã®ãƒ–ãƒã‚°ã‚’ã¯ã˜ã‚ã¾ã—ãŸã€‚â€»æœ€æ–°æƒ…å ±ã¯Recruit Data Blogã‚’ã”覧ãã ã•ã„。 Recruit Data Blogã¯ã“ã¡ã‚‰ 汎用人型雑用AIã® stakaya ã§ã™ã€‚ ãŸã¾ãŸã¾æ•°å¹´å‰ã«ç¤¾å†…ã®BLOGã«æ›¸ã„ãŸABテストã®ãƒã‚¸ãƒƒã‚¯ã®ã¾ã¨ã‚＆比較記事を発掘ã—ãŸã®ã§ã€ ã“ã®ã¾ã¾çœ らã›ã¦ã„ã‚‹ã®ã¯ã‚‚ã£ãŸã„ãªã„ãžã¨ã€ 圧倒的もã£ãŸã„ãªã„精神を発æ®ã—ã€ã‚·ã‚§ã‚¢ã•ã›ã¦ã„ãŸã ãã¾ã™ã€‚ ã‚ã®é ƒã¯ç§ã‚‚è‹¥ã‹ã£ãŸã€‚ 社内ã§ã¯â€å …物・真é¢ç›®ãƒ»ä¸€é€”â€ã§æœ‰åãªã‚‚ã®ã§ã—ã¦ã€ä¸‹è¨˜æ–‡ç« ãŒãŠç¡¬ã„ã®ã¯ã”勘å¼ãã ã•ã„。 ã¯ã˜ã‚㫠本記事ã¯ã€æ–½ç–ã®è©•ä¾¡æ‰‹æ³•ã¨ã—ã¦ã—ã°ã—ã°ç”¨ã„られるA/Bテスト（A/B testing)ã«ã¤ã„ã¦ã€ã§ãã‚‹é™ã‚ŠèƒŒå¾Œã«ã‚る仮定を明記ã—ãŸä¸Šã§ã€ã¾ã¨ã‚ãŸã‚‚ã®ã§ã‚る。 A/Bテストã¨ã¯ã€ä¸»ã«ã‚¤ãƒ³ã‚¿ãƒ¼ãƒãƒƒãƒˆãƒžãƒ¼ã‚±ãƒ†ã‚£ãƒ³ã‚°ã«ãŠã‘ã‚‹æ–½ç–ã®è‰¯å¦ã‚’判æ–ã™ã‚‹ãŸã‚ã«ã€2ã¤ã®æ–½ç–（通常ã€A・Bã¨è¨˜è¼‰ï¼‰ã‚’比較ã™
mots quotidiens.
9月ã®NLç ”(NL213)ã®åŽŸç¨¿ã‚’無事サブミットã—ã¾ã—ãŸã®ã§, ç½®ã„ã¦ãŠãã¾ã™ã€‚ 「ガウスéŽç¨‹ã«åŸºã¥ã連続空間トピックモデルã€. nl213cstm.pdf 発表ã¯é€šå¸¸ç™ºè¡¨ã®ä¸€ç•ªæœ€å¾Œã§, 9/13(金)ã®åˆå‰ä¸ã®ã‚»ãƒƒã‚·ãƒ§ãƒ³ã®ã‚ˆã†ã§ã™ã€‚ (NL213ã®ãƒ—ãƒã‚°ãƒ©ãƒ )。 ガウスéŽç¨‹ã«åŸºã¥ã„ã¦ã¯ã„ã‚‹ã‚‚ã®ã®, çµæžœçš„ã«RBMã§æ½œåœ¨å±¤ãŒ1/0ã§ã¯ãªã (ã¤ã¾ã‚Šã‚·ã‚°ãƒ¢ã‚¤ãƒ‰é–¢æ•°ã‚’使ã‚ãš), ガウス分布ã«å¾“ã†ã‚ˆã†ãªç”Ÿæˆãƒ¢ãƒ‡ãƒ«ã«ãªã‚‹ã®ã§, RBMã«ã¤ã„ã¦ä¸‹ã®ã‚ˆã†ãªç–‘å•ã‚’æŒã£ã¦ã„ã‚‹æ–¹(çµæ§‹å¤šã„ã¨æ€ã†)ã‚„, 最é©åŒ–ã®é›£ã—ã•ã« å›°ã£ã¦ã„る人ã«ã¯èˆˆå‘³ã‚’æŒã£ã¦ã‚‚らãˆã‚‹ã‚ˆã†ãªå†…容ã‹ã¨æ€ã£ã¦ã„ã¾ã™ã€‚ 甲府ã¯ä¸‰é·¹ã‹ã‚‰ã¯ç‰¹æ€¥ã‹ã„ã˜å·ã§1本ã§è¡Œã‘ã‚‹ã®ã§ã™ãŒ, 1日目ã«åº§é•·ã‚’ã™ã‚‹ã“ã¨ã« ãªã£ãŸã®ã§, 戻らãšã«åˆæ—¥ã¯æ™®é€šã«æ³Šã¾ã‚‹äºˆå®šã§ã™ã€‚ Rao-BlackwellizationãŒä½•ã®ã“ã¨ã‹, é•·ã„ã“ã¨å…¨ç„¶ã‚ã‹ã‚‰ãªã‹ã£ãŸã®ã ãŒ, "
統計å¦ã®ç¾åœ¨éŽåŽ»æœªæ¥ --方法論ã®è¦–点ã‹ã‚‰--
æ±Ÿå£ çœŸé€ ï¼ˆçµ±è¨ˆæ•°ç†ç ”究所，ç·åˆç ”究大å¦é™¢ï¼‰ 「ç¾åœ¨éŽåŽ»æœªæ¥ã€ã«ã¤ã„ã¦ã®åŽŸç¨¿ä¾é ¼ãŒããŸï¼Žã‚¿ã‚¤ãƒˆãƒ«ã¯èŒ«æ¼ ã¨ã—ã¦ãŠã‚Šï¼Œç§ã®ç‹ã„ç ”ç©¶ã®èˆˆå‘³ã®ç¯„囲，与ãˆã‚‰ã‚ŒãŸèƒ½åŠ›ã‚’ã¯ã‚‹ã‹ã«è¶…ãˆã¦ã„る．通常ã®æ„Ÿè¦šã§ã¯ï¼Œä¾é ¼ã‚’æ–り自分ã®ç ”究・教育ã«å°‚念ã™ã‚‹æ–¹ãŒè‰¯ã„ã«æ±ºã¾ã£ã¦ã„る．自分ã®ç„¡èƒ½ã•ã‚’冷é™ã«è¦‹ã‚Œã‚‹ã ã‘ã®è‹¦ã„経験ã¯ç©ã‚“ã§ã„ã‚‹ã—，ã“れを論ãšã‚‹ã¹ãã‚‚ã£ã¨é©å½“ãªäººã€…ãŒã„ã‚‹ã«ã¡ãŒã„ãªã„．ã“ã®ã‚ˆã†ã«è€ƒãˆãŸçµæžœï¼Œä¸é‡ã«ãŠæ–ã‚Šã—よã†ã¨æ€ã£ãŸï¼Žæ£ç›´è¨€ã£ã¦ã»ã‚“ã®ï¼‘・２年å‰ã¾ã§ã¯ï¼Œçµ±è¨ˆå¦ã®å°†æ¥ã«ä½•ã®ç–‘å•ç‚¹ã‚‚æŒã£ã¦ã„ãªã‹ã£ãŸï¼Žã€Œæ—¥æœ¬ã®çµ±è¨ˆå¦ã¯ï¼Œãã®é•·ã„活動ã®æ´å²ã®ä¸ã§è¼ã‹ã—ã„実績を挙ã’，今後もç€å®ŸãªæˆæžœãŒæœŸå¾…ã•ã‚Œã¦ã„る．多種多様ãªåˆ†é‡Žã«ãŠã„ã¦ï¼Œãƒ‡ãƒ¼ã‚¿ã‚’å¾—ã‚‹ãŸã‚ã®è¨ˆç”»ã‹ã‚‰å¾—られãŸãƒ‡ãƒ¼ã‚¿ã«åŸºã¥ã„ã¦ãªã•ã‚Œã‚‹çµ±è¨ˆæŽ¨æ¸¬ã¾ã§ï¼Œè‘—ã—ã„貢献ãŒå°†æ¥ã«æ¸¡ã£ã¦ç´„æŸã•ã‚Œï¼Œçµ±è¨ˆç§‘å¦ã¯è¼ã続ã‘ã‚‹ã ã‚ã†ï¼Žã€ã“ã†ä¿¡ã˜ã¦ã„ãŸï¼Žã—ã‹ã—ãªãŒã‚‰ç¾æ™‚点1999å¹´ã¯ï¼Œã¾ã•ã«
Empirical Bayes method - Wikipedia
Empirical Bayes methods are procedures for statistical inference in which the prior probability distribution is estimated from the data. This approach stands in contrast to standard Bayesian methods, for which the prior distribution is fixed before any data are observed. Despite this difference in perspective, empirical Bayes may be viewed as an approximation to a fully Bayesian treatment of a hie
生æˆãƒ¢ãƒ‡ãƒ«ã¨åˆ¤åˆ¥ãƒ¢ãƒ‡ãƒ« - アルゴリズムマニア2.0
Discriminative models, not discriminative traininghttp://research.microsoft.com/research/pubs/view.aspx?type=Technical%20Report&id=998ãšãƒ¼ãƒ¼ãƒ¼ãƒ¼ãƒ¼ãƒ¼ãƒ¼ãƒ¼ãƒ¼ãƒ¼ãƒ¼ãƒ¼ãƒ¼ã£ã¨å‰ã«ã“れを見ã¦ã€èªã¿ãŸã„ãªã¨ã¯æ€ã£ãŸã‚‚ã®ã®ã€ãŸã‹ã ã‹A4ã§1æžšã®æ–‡ç« ã‚’èªã‚€ä½™åŠ›ã‚‚ãªãã€å¿˜ã‚Œã¦ã„ãŸã®ã§ã™ãŒã€ãµã¨æ€ã„出ã—ã¦èªã‚“ã§ã¿ã¾ã—ãŸã€‚ Due to the model assumptions, the estimations of and 0 decouple, so the best is the same as in (3).ã“ã“ã¾ã§ã¯OKãªã‚“ã§ã™ãŒã€ãã£ã‹ã‚‰å…ˆãŒç†è§£ã§ãã¾ã›ã‚“ã§ã—ãŸã€‚具体的ã«ã¯It must be because a model of the form (5) fits
Latent Process Decompositionã®ãŸã‚ã®collapsed変分ベイズ推定ã®è§£èª¬ | photogenic blue note
下記論文ã«ã‚る，latent process decompositionã®ãŸã‚ã®collapsed変分ベイズ推定ã«ã¤ã„ã¦ï¼Œã‚¬ã‚¦ã‚¹åˆ†å¸ƒã«ã‚ˆã‚‹è¿‘ä¼¼ã®ã¨ã“ã‚ã‚’ã®ãžã„ã¦ï¼Œå¼ã®å°Žå‡ºã‚’解説ã—ãŸæ–‡æ›¸ã‚’書ãã¾ã—ãŸï¼Žã“ã“をクリックã—ã¦ãã ã•ã„． Yiming Ying, Peng Li and Colin Campbell. A marginalized variational bayesian approach to the analysis of array data. BMC Proceedings 2008, 2(Suppl 4):S7. [2009/03/27 追記] 大å¦ã®Webサーãƒï¼Œãƒ¡ãƒ¼ãƒ«ã‚µãƒ¼ãƒãŒç§»è»¢ã®ãŸã‚休æ¢ä¸ã§ã™ï¼Žä¸Šè¨˜ã®ç§ãŒä½œæˆã—ãŸè§£èª¬æ–‡æ›¸ã¯ï¼Œé–“é•ã„ãŒã‚ã£ãŸã®ã§ä¿®æ£ã—ã¾ã—ãŸï¼Žæ–°ã—ã„ãƒãƒ¼ã‚¸ãƒ§ãƒ³ã¯ï¼ŒWebサーãƒãŒå¾©æ´»ã—次第，アップã—ã¾ã™ï¼Ž ã¾ãŸï¼Œå®Ÿè£…も終ãˆã¦ï¼Œwineデータセットã«ã¤ã„ã¦ï¼Œã»
Statistics vs. Machine Learning, fight! | AI and Social Science – Brendan O'Connor
nice place to have a meeting: Las Vegas in August Hah. Or rather, ouch! I had two thoughts reading this. (1) Poor statisticians. Machine learners invent annoying new terms, sound cooler, and have all the fun. (2) What’s wrong with statistics? They have way less funding and influence than it seems they might deserve. There are several issues going on here, both substantive and cultural: There might
1-text-A5-size.dvi
1-text-A5-size : 2008/9/29(15:45) 第 4 ç« æ¨™æœ¬åˆ†å¸ƒè«–ã¨æ¼¸è¿‘分布論 1 1.1 ランダム標本 標本分布論ã®æž 組㿠定義 4.1  X1 , X2 , . . . , Xn ãŒæ¯é›†å›£åˆ†å¸ƒ f (x) ã‹ã‚‰ã®æ¨™æœ¬ã®å¤§ãã•ãŒ n ã®ãƒ©ãƒ³ãƒ€ãƒ 標本ã§ã‚ ã‚‹ã¨ã¯ï¼ŒX1 , X2 , . . . , Xn ã¯äº’ã„ã«ç‹¬ç«‹ãªç¢ºçŽ‡å¤‰æ•°åˆ—ã§ã‚ã£ã¦ï¼Œå„ Xi , i = 1, 2, . . . , n, ã¯ç¢ºçŽ‡ 関数ã¾ãŸã¯ç¢ºçŽ‡å¯†åº¦é–¢æ•° f (x) ã«å¾“ã†ã¨ãã‚’ã„ã†ï¼Žã¾ãŸï¼ŒX1 , X2 , . . . , Xn ã¯ç‹¬ç«‹åŒä¸€ã«ç¢ºçŽ‡é–¢æ•° ã¾ãŸã¯ç¢ºçŽ‡å¯†åº¦é–¢æ•° f (x) ã«å¾“ã†ã¨ã‚‚ã„ã†ï¼Ž 多ãã®å ´åˆã¯ n > 1 ã§ã‚る．ã¾ãŸï¼Œ æ³¨æ„ 4.1   åŒæ™‚確率密度関数㯠= (X1 , X2 , . . . , Xn ) ã®åŒæ™‚確率関数ã¾ãŸã¯ n f (x1 , x2 ,
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
WSDM2019tutorial
Variational Bayesian Dirichlet-Multinomial Allocation for Exponential Family Mixtures Shipeng Yu1,2 , Kai Yu2 , Volker Tresp2 , and Hans-Peter Kriegel1 1 Institute for Computer Science, University of Munich, Germany 2 Siemens Corporate Technology, Munich
データã¯ã©ã‚Œãらã„ã‚ã£ãŸã‚‰å分㋠- ã ã«ãˆã‚‹blog
Statistics 210B: Lecture Notes
Elements of Statistical Learning: data mining, inference, and prediction. 2nd Edition.
UCLA Statistics | Website
1-text-A5-size.dvi
1-text-A5-size.dvi
testproblem-2009-01-20.dvi
今野   (2007 å¹´ 1 月 9 æ—¥) æƒ…å ±çµ±è¨ˆå¦ã®è©¦é¨“å•é¡Œï¼ˆ 試験時間㯠75 分） 解ç”上ã®æ³¨æ„ 特別ãªæŒ‡ç¤ºãŒã‚ã‚‹å ´åˆã‚’除ã，ç”ãˆãŒåˆã£ã¦ã„ã‚‹ã‹ã© ã†ã‹ã‚ˆã‚Šã‚‚解ç”ã®é€”ä¸éŽç¨‹ã®è«–ç†çš„展開 ã‚’é‡è¦–ã—
{{#tags}}- {{label}}
{{/tags}}