[B! histogram] yassã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯

yass id:yass

histogramã«é–¢ã™ã‚‹yassã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯ (7)

${{author_name}}$

{{{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}}

Streaming histograms... faster
yass 2013/11/24
" The algorithm is online/streaming, so it only needs on pass over the data and it can always give you an estimate of the distribution so far. "

quantile

histogram

java

streaming
ãƒªãƒ³ã‚¯
Home - Metamarkets
All is well and good if we content ourselves with computations that can be distributed efficiently such as summing hourly revenue to produce daily revenue, or calculating click-through rates. In the language of Gray et al., the former calculation is distributive: we can sum the raw event prices to produce hourly revenue over each combination of dimensions and in turn sum this intermediary for furt
yass 2013/09/15
algorithms

histogram

quantile

toread

druid
ãƒªãƒ³ã‚¯
Something went wrong...
yass 2013/02/12
histogram

EMD

similarity
ãƒªãƒ³ã‚¯
ãƒ’ã‚¹ãƒˆã‚°ãƒ©ãƒ ã®åˆ†é¡žå¹…ã®ç›®å®‰ã€ç ”ç©¶è€…å‘ã‘ã€‘ | æ²¿å²¸ç’°å¢ƒã®ç ”ç©¶æ—¥è¨˜
ãƒ’ã‚¹ãƒˆã‚°ãƒ©ãƒ ã¯ãƒ‡ãƒ¼ã‚¿ã®åˆ†å¸ƒã‚’ã¿ã‚‹ã¨ãã«ä½¿ã„ã¾ã™ãŒã€ åˆ†é¡žå¹…ã‚’æ±ºã‚ã‚‹ç›®å®‰ã‚’çŸ¥ã£ã¦ãŠãã¨ä¾¿åˆ©ã§ã™ã€‚ è‡ªåˆ†è‡ªèº«ã«å¯¾ã™ã‚‹ãƒ¡ãƒ¢ã®æ„å‘³ã‚‚è¾¼ã‚ã¦ã€ã“ã“ã«æ›¸ã„ã¦ãŠãã¾ã™ã€‚ Sturges(1926) k = 1 + log2(n) = 1 + 3.32 log10(n) ã“ã“ã§ã€kã¯åˆ†é¡žã‚¯ãƒ©ã‚¹ã®æ•°ã€nã¯ã‚µãƒ³ãƒ—ãƒ«æ•°ã€log2ã¯2ã‚’åº•ã¨ã™ã‚‹logã®æ„å‘³ã€‚ ä¾‹ãˆã°ã€n=1000ã®å ´åˆã€k=10.96â‰’11ã¨ãªã‚‹ã®ã§ã€ ãƒ‡ãƒ¼ã‚¿ã®æœ€å¤§å€¤ã€æœ€å°å€¤ã‚’è¦‹ã¦åˆ†é¡žå¹…ã‚’æ±ºã‚ã‚‹ã€‚ ãŸã ã—ã€Sturgesã®æ–¹æ³•ã¯ãƒ’ã‚¹ãƒˆã‚°ãƒ©ãƒ ãŒæ»‘ã‚‰ã‹ã«ãªã‚Šã™ãŽã‚‹ã®ã§ä½¿ã‚ãªã„æ–¹ãŒè‰¯ã„ã¨ã„ã‚ã‚Œã¦ã„ã¾ã™ã€‚ Scott(1979) h = 3.5 s n^{-1/3} ã“ã“ã§ã€hã¯åˆ†é¡žå¹…ã€sã¯æ¨™æº–åå·®ã€n^{-1/3}ã¯ã‚µãƒ³ãƒ—ãƒ«æ•°ã®-1/3ä¹—ã‚’è¡¨ã‚ã™ã€‚ Freedman and Diaconis(1981) h = 2 IQR n^{-1/3} ã“ã“ã§ã€I
yass 2013/02/11
histogram

statistics
ãƒªãƒ³ã‚¯
ãƒ’ã‚¹ãƒˆã‚°ãƒ©ãƒ ã®æ¯”è¼ƒ - ã«ã®ã›ãæ—¥è¨˜
OpenCVã‚’ä½¿ã£ã¦ãƒ’ã‚¹ãƒˆã‚°ãƒ©ãƒ ã®æ¯”è¼ƒã‚’è¡Œã£ã¦ã¿ã‚‹ã€‚ ã©ã†ã‚„ã‚‹ã‹ã¨ã„ã†ã¨ã€äºŒã¤ã®ç”»åƒã®ãƒ’ã‚¹ãƒˆã‚°ãƒ©ãƒ ã‚’è¨ˆç®—ã—ã¦ã‹ã‚‰ã€cvCompareHistã‚’ç”¨ã„ã¦ãã‚Œã‚‰ã‚’æ¯”è¼ƒã™ã‚Œã°ã„ã„ã€‚ ä»Šå›žã¯ä»¥ä¸‹ã®äºŒã¤ã®ç”»åƒã‚’ä½¿ã£ã¦ã¿ãŸã€‚è‡ªåˆ†ã®æ‰‹ã®ç”»åƒã§ã™ã€‚ myhand1.jpg myhand2.jpg cvCompareHistã«ã¯å››ã¤ã®ãƒ¡ã‚½ãƒƒãƒ‰(Correlationã€Chi-squareã€Intersectionã€Bhattacharyya distance)ãŒã‚µãƒãƒ¼ãƒˆã•ã‚Œã¦ã„ã‚‹ã€‚ Correlationã‚’ç”¨ã„ã‚‹ã¨ çµæžœãŒé«˜ã„ã»ã†ãŒã‚ˆã‚Šãƒžãƒƒãƒã—ã¦ã„ã‚‹ å®Œç’§ãªãƒžãƒƒãƒã ã£ãŸå ´åˆã¯1 å®Œç’§ãªãƒŸã‚¹ãƒžãƒƒãƒã®å ´åˆã¯-1 Chi-squareã‚’ç”¨ã„ã‚‹ã¨ çµæžœãŒä½Žã„ã»ã†ãŒã‚ˆã‚Šãƒžãƒƒãƒã—ã¦ã„ã‚‹ å®Œç’§ãªãƒžãƒƒãƒã ã£ãŸå ´åˆã¯0 å®Œç’§ãªãƒŸã‚¹ãƒžãƒƒãƒã ã£ãŸå ´åˆã¯éš›é™ã®ãªã„å€¤ãŒå¾—ã‚‰ã‚Œã‚‹(ãƒ’ã‚¹ãƒˆã‚°ãƒ©ãƒ ã®ã‚µã‚¤ã‚ºã«ä¾å˜ã™ã‚‹) Intersectio
yass 2013/02/11
image

histogram

similarity

opencv
ãƒªãƒ³ã‚¯
Earth Mover's Distance (EMD) - äººå·¥çŸ¥èƒ½ã«é–¢ã™ã‚‹æ–å‰µéŒ²
Earth Mover's Distance (EMD) ã«ã¤ã„ã¦èª¿ã¹ãŸã“ã¨ã‚’æ•´ç†ã—ã¦ãŠãã¾ã™ã€‚EMDã¯ã€ãƒ¦ãƒ¼ã‚¯ãƒªãƒƒãƒ‰è·é›¢ã®ã‚ˆã†ãªè·é›¢å°ºåº¦ã®ä¸€ã¤ã§ã€äºŒã¤ã®åˆ†å¸ƒã®é–“ã®è·é›¢ã‚’æ¸¬ã‚‹ã“ã¨ãŒã§ãã¾ã™ã€‚è¨€èªžå‡¦ç†ã§ã¯ã‚ã¾ã‚Šèžã„ãŸã“ã¨ãªã‹ã£ãŸã®ã§ã™ãŒã€ç”»åƒå‡¦ç†ã‚„éŸ³å£°å‡¦ç†ã§ã¯æ¯”è¼ƒçš„æœ‰åãªè·é›¢å°ºåº¦ã®ã‚ˆã†ã§ã™ã€‚ EMDãŒä½¿ãˆã‚‹å•é¡Œè¨å®šã¯ä¸‹å›³ã®ã‚ˆã†ã«ãªã‚Šã¾ã™ã€‚ EMDã¯ç‰¹å¾´é‡ã¨é‡ã¿ã®é›†åˆï¼ˆã‚·ã‚°ãƒãƒãƒ£ã¨å‘¼ã¶ï¼‰ã§ä¸Žãˆã‚‰ã‚Œã‚‹åˆ†å¸ƒPã¨åˆ†å¸ƒQã®é–“ã®è·é›¢ã§ã™ã€‚ã“ã“ã§ã€ç‰¹å¾´é‡é–“ã§ã¯è·é›¢ ãŒå®šç¾©ã•ã‚Œã¦ã„ã‚‹ã®ãŒå‰æã§ã™ã€‚ç‰¹å¾´é‡ãŒãƒ™ã‚¯ãƒˆãƒ«ã®ã¨ãã¯ãƒ¦ãƒ¼ã‚¯ãƒªãƒƒãƒ‰è·é›¢ã€ç‰¹å¾´é‡ãŒç¢ºçŽ‡åˆ†å¸ƒã®ã¨ãã¯ã‚«ãƒ«ãƒãƒƒã‚¯ãƒ»ãƒ©ã‚¤ãƒ–ãƒ©ãƒ¼è·é›¢ï¼ˆæƒ…å ±é‡ï¼‰ãªã©ã§ã™ã€‚EMDã¯ã€ç‰¹å¾´é‡ã®é›†åˆãŒ2ã¤ä¸Žãˆã‚‰ã‚ŒãŸã¨ãã«ã€1å€‹1å€‹ã®ç‰¹å¾´é‡é–“ã®è·é›¢ã‚’ã‚‚ã¨ã«ã€ç‰¹å¾´é‡é›†åˆé–“ã®è·é›¢ã‚’æ±‚ã‚ã‚‰ã‚Œã‚‹ã‚“ã§ã™ãã€‚ã“ã‚Œã¯ã™ã”ã„ã€‚ é‡ã¿ã¯å…·ä½“çš„ãªå¿œç”¨ã«ã‚ˆã£ã¦ä½¿ã„æ–¹ãŒå¤‰ã‚ã‚Šã¾ã™ãŒã€ãã®ç‰¹å¾´é‡ã®é‡è¦åº¦ã‚’
yass 2013/02/11
histogram

similarity

image

EMD
ãƒªãƒ³ã‚¯
é¡žä¼¼ç”»åƒæ¤œç´¢ã‚·ã‚¹ãƒ†ãƒ ã‚’ä½œã‚ã† - äººå·¥çŸ¥èƒ½ã«é–¢ã™ã‚‹æ–å‰µéŒ²
C++ç‰ˆã®OpenCVã‚’ä½¿ã£ã¦ã‚«ãƒ©ãƒ¼ãƒ’ã‚¹ãƒˆã‚°ãƒ©ãƒ ã‚’ç”¨ã„ãŸé¡žä¼¼ç”»åƒæ¤œç´¢ã‚’å®Ÿé¨“ã—ã¦ã¿ã¾ã—ãŸã€‚ãƒãƒƒãƒå‡¦ç†ãªã©ã®ã‚¹ã‚¯ãƒªãƒ—ãƒˆã¯Pythonã‚’ä½¿ã£ã¦ã¾ã™ãŒã€Perlã§ã‚‚Rubyã§ã‚‚ä¼¼ãŸã‚ˆã†ãªæ„Ÿã˜ã§ã§ãã¾ã™ã€‚ æŒ‡å®šã—ãŸç”»åƒã¨é¡žä¼¼ã—ãŸç”»åƒã‚’æ¤œç´¢ã™ã‚‹ã‚·ã‚¹ãƒ†ãƒ ã¯é¡žä¼¼ç”»åƒæ¤œç´¢ã‚·ã‚¹ãƒ†ãƒ ã¨è¨€ã„ã¾ã™ã€‚Googleã‚„Yahoo!ã®ã‚¤ãƒ¡ãƒ¼ã‚¸æ¤œç´¢ã¯ã€ã‚¯ã‚¨ãƒªã«ã‚ãƒ¼ãƒ¯ãƒ¼ãƒ‰ã‚’å…¥ã‚Œã¦ã‚ãƒ¼ãƒ¯ãƒ¼ãƒ‰ã«é–¢é€£ã—ãŸç”»åƒã‚’æ¤œç´¢ã—ã¾ã™ãŒã€é¡žä¼¼ç”»åƒæ¤œç´¢ã§ã¯ã‚¯ã‚¨ãƒªã«ç”»åƒã‚’ä¸Žãˆã‚‹ã®ãŒç‰¹å¾´çš„ã§ã™ã€‚ã“ã®åˆ†é‡Žã¯ã€Content-Based Image Retrieval (CBIR)ã¨å‘¼ã°ã‚Œã¦ãŠã‚Šã€æœ€æ–°ã®ã‚µãƒ¼ãƒ™ã‚¤è«–æ–‡ï¼ˆDatta,2008ï¼‰ã‚’èªã‚€ã¨1990å¹´ä»£å‰åŠã¨ã‘ã£ã“ã†æ˜”ã‹ã‚‰ç ”ç©¶ã•ã‚Œã¦ã¾ã™ã€‚ æœ€æ–°ã®æ‰‹æ³•ã§ã¯ã€è‰²ã€å½¢çŠ¶ã€ãƒ†ã‚¯ã‚¹ãƒãƒ£ã€ç‰¹å¾´ç‚¹ãªã©ã•ã¾ã–ã¾ãªç‰¹å¾´é‡ã‚’ç”¨ã„ã¦é¡žä¼¼åº¦ã‚’åˆ¤å®šã™ã‚‹ãã†ã§ã™ãŒã€ä»Šå›žã¯ã€ã‚‚ã£ã¨ã‚‚ç°¡å˜ãªã€Œè‰²ã€ã‚’ç”¨ã„ãŸé¡žä¼¼ç”»åƒæ¤œç´¢ã‚’å®Ÿé¨“ã—ã¦ã¿ã¾ã™
yass 2013/02/11
opencv

image

similarity

histogram
ãƒªãƒ³ã‚¯
1