[B! algorithm] chezouã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯

chezou id:chezou

algorithmã«é–¢ã™ã‚‹chezouã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯ (25)

${{author_name}}$

{{{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}}

ä¹±æŠžãƒ‡ãƒ¼ã‚¿æ§‹é€ ã®æœ€æ–°äº‹æƒ… âˆ’MinHash ã¨ HyperLogLog ã®æœ€è¿‘ã®é€²æ©âˆ’ - iwiwiã®æ—¥è¨˜
æœ¬æ—¥ï¼ŒPFI ã‚»ãƒŸãƒŠãƒ¼ã«ã¦ã€Œä¹±æŠžãƒ‡ãƒ¼ã‚¿æ§‹é€ ã®æœ€æ–°äº‹æƒ… âˆ’MinHash ã¨ HyperLogLog ã®æœ€è¿‘ã®é€²æ©âˆ’ã€ã¨ã„ã†ã‚¿ã‚¤ãƒˆãƒ«ã§è©±ã‚’ã•ã›ã¦ã‚‚ã‚‰ã„ã¾ã—ãŸï¼Žã‚¹ãƒ©ã‚¤ãƒ‰ã¯ä»¥ä¸‹ã«ãªã‚Šã¾ã™ï¼Ž Ustream ã®éŒ²ç”»ã‚‚ã‚ã‚Šã¾ã™ï¼Ž http://www.ustream.tv/recorded/48151077 å†…å®¹ã¨ã—ã¦ã¯ï¼Œä»¥ä¸‹ã®æ“ä½œã‚’åŠ¹çŽ‡çš„ã«è¡Œã†ãŸã‚ã®é›†åˆã«é–¢ã™ã‚‹ãƒ‡ãƒ¼ã‚¿æ§‹é€ (Sketch) ã®æœ€è¿‘ã®é€²æ©ã‚’ç´¹ä»‹ã—ã¾ã—ãŸï¼Ž é›†åˆã®é¡žä¼¼åº¦ã®æŽ¨å®š (Jaccard ä¿‚æ•°) é›†åˆç•°ãªã‚Šæ•°ã®æŽ¨å®š (distinct counting) ã©ã¡ã‚‰ã‚‚é‡è¦ã‹ã¤åŸºç¤Žçš„ãªæ“ä½œã§ï¼Œb-bit MinHash ã‚„ HyperLogLog ãªã©ï¼Œæ—¢ã«å®Ÿç”¨çš„ãªæ‰‹æ³•ãŒææ¡ˆã•ã‚Œã¦ãŠã‚Šï¼Œå®Ÿéš›ã«ã‚‚ä½¿ã‚ã‚Œã¦ã„ã¾ã™ï¼Žã—ã‹ã—ï¼Œ2014 å¹´ã«ãªã£ã¦ï¼ŒOdd Sketch ã‚„ HIP Estimator ã¨ã„ã†ï¼Œã“ã‚Œã‚‰ã‚’ã•ã‚‰ã«æ”¹å–„ã™ã‚‹æ‰‹æ³•ãŒç«‹ã¦ç¶š
chezou 2014/05/30
minhash

HyperLogLog

Algorithm
ãƒªãƒ³ã‚¯
waveletè¡Œåˆ—ã§é«˜é€Ÿãªã€Œã‚‚ã—ã‹ã—ã¦å‹ã ã¡ï¼Ÿã€æ¤œç´¢ | æ ªå¼ä¼šç¤¾ã‚µã‚¤ãƒãƒ¼ã‚¨ãƒ¼ã‚¸ã‚§ãƒ³ãƒˆ
æ¥å‹™çµŒæ´ï¼š Sierã§ã®ã‚½ãƒ•ãƒˆã‚¦ã‚§ã‚¢é–‹ç™ºãƒ»å¤§æ‰‹ãƒ¡ãƒ‡ã‚£ã‚¢ã§ã®ã‚µãƒ¼ãƒ“ã‚¹é‹ç”¨ã‚’çµŒã¦2012å¹´ã‚µã‚¤ãƒãƒ¼ã‚¨ãƒ¼ã‚¸ã‚§ãƒ³ãƒˆå…¥ç¤¾ã€‚ ã‚¢ãƒ¡ãƒ¼ãƒäº‹æ¥æœ¬éƒ¨ã‚³ãƒŸãƒ¥ãƒ‹ãƒ†ã‚£ã‚µãƒ¼ãƒ“ã‚¹ã®é–‹ç™ºè²¬ä»»è€…ã‚’çµŒã¦ã€ç¾åœ¨ã¯ã‚¢ãƒ‰ãƒ†ã‚¯ã‚¹ã‚¿ã‚¸ã‚ªã§åºƒå‘Šé…ä¿¡æŠ€è¡“ã«æ³¨åŠ›ã€‚ å¥½ããªåˆ†é‡Žã¯ã‚°ãƒ©ãƒ•æŽ¢ç´¢ã¨ãƒãƒ¥ãƒ¼ãƒªãƒ³ã‚°ãƒžã‚·ãƒ³ã€‚ ã‚½ãƒ¼ã‚·ãƒ£ãƒ«ã‚µãƒ¼ãƒ“ã‚¹ã§ã¯ã€ãƒ¦ãƒ¼ã‚¶é–“ã®ã¤ãªãŒã‚Šã‚„ãƒ¦ãƒ¼ã‚¶åŒå£«ã®é¡žä¼¼æ€§ãŒã¨ã¦ã‚‚é‡è¦ã§ã™ã€‚ ã¤ãªãŒã‚Šã®è¿‘ã„ãƒ¦ãƒ¼ã‚¶ã‚„è‡ªåˆ†ã¨ä¼¼ã¦ã„ã‚‹ãƒ¦ãƒ¼ã‚¶ã‚’ã€Œã‚‚ã—ã‹ã—ã¦å‹ã ã¡ï¼Ÿã€ã¨ã‚µã‚¸ã‚§ã‚¹ãƒˆã™ã‚‹ã“ã¨ã§ãƒ¦ãƒ¼ã‚¶é–“ã®ã¤ãªãŒã‚Šã‚’ä¼¸å±•ã•ã›ã‚‹ã“ã¨ãŒã§ãã¾ã™ã€‚ ãã“ã§ã€ãƒ¦ãƒ¼ã‚¶ã®ã€Œã¤ãªãŒã‚Šã€å…·åˆãŒä¼¼ã¦ã„ã‚‹ãƒ¦ãƒ¼ã‚¶ã‚’ã€Œå‹ã ã¡ã‹ã‚‚ã—ã‚Œãªã„ãƒ¦ãƒ¼ã‚¶ã€ã¨ã—ã¦ã‚µã‚¸ã‚§ã‚¹ãƒˆã‚’è¡Œã†ã“ã¨ã‚’è€ƒãˆã¾ã—ãŸã€‚ ã—ã‹ã—ã€Œã¤ãªãŒã‚Šã€ã®ãƒ‡ãƒ¼ã‚¿ã¨ã„ã†ã®ã¯ãƒ¦ãƒ¼ã‚¶æ•°ã®ãƒ™ã‚ä¹—ã§ã‚ã‚‹ãŸã‚ã€å®¹é‡ãŒå¤§ãããªã‚Šã‚„ã™ã„æ€§è³ªãŒã‚ã‚Šã¾ã™ã€‚ å³ã¡ã€ã€Œã¤ãªãŒã‚Šã€é¡žä¼¼åº¦ã®ç®—å‡ºã«ã¯æ™‚é–“ãŒã‹ã‹ã‚‹ã€ã¨ã„ã†ã“ã¨ã§ã™ã€‚ ã“ã®ã€Œã¤ãªãŒã‚Šã€é¡žä¼¼åº¦ç®—å‡º
chezou 2014/05/19
algorithm

graph
ãƒªãƒ³ã‚¯
https://15799.courses.cs.cmu.edu/fall2013/static/papers/stonebraker-sigmod-record-2013.pdf
chezou 2014/02/16
parallel LSH with Julia and SciDB

julia

algorithm

scidb
ãƒªãƒ³ã‚¯
K-means hashing (CVPR'13) ã¨ãƒãƒƒã‚·ãƒ³ã‚°å‘¨ã‚Š
K-means hashing (CVPR'13) ã®è«–æ–‡è§£èª¬ã¨ã€é–¢é€£ã™ã‚‹ iterative quantization ã‚„ optimized product quantization ã®ç´¹ä»‹ã€æœ€è¿‘ã®hashingç³»è«–æ–‡ãƒªã‚¹ãƒˆã€‚
chezou 2014/02/16
hashing

algorithm
ãƒªãƒ³ã‚¯
åœ§ç¸®æƒ…å ±å‡¦ç†ãƒŽã‚¹ã‚¹ãƒ¡ - An Encouragement of Compression å‚æœ¬æ¯”å‘‚å¿— (PDF)
åœ§ç¸®æƒ…å ±å‡¦ç†ãƒŽã‚¹ã‚¹ãƒ¡ An Encouragement of Compression å‚æœ¬æ¯”å‘‚å¿— ãƒ‡ãƒ¼ã‚¿ã‚’ã©ã‚Œã ã‘å°ã•ãã§ãã‚‹ã‹ã¨ã„ã†æ ¹æºçš„ãªå•ã„ã¯ï¼Œç¾åœ¨ã‚‚ãƒ‡ãƒ¼ã‚¿åœ§ç¸®ã®ä¸»è¦ãƒ†ãƒ¼ãƒžã§ã‚ã‚Šç¶šã‘ã¦ã„ã‚‹ï¼Žä¸€æ–¹ã§ï¼Œ ãƒ‡ãƒ¼ã‚¿åœ§ç¸®ã¯æ™‚ä»£ã¨ã¨ã‚‚ã«æ–°ã—ã„ä¾¡å€¤ã‚’ç²å¾—ã—ã¦ããŸï¼Žä¾‹ãˆã°ï¼Œæ–‡å—åˆ—ãƒ‡ãƒ¼ã‚¿ã‚’åœ§ç¸®ã™ã‚‹ã“ã¨ã§é«˜é€Ÿæ¤œç´¢ã™ã‚‹ç†è«–ãŒ 1990 å¹´ä»£ã«ææ¡ˆã•ã‚Œï¼Œæ§˜ã€…ãªåˆ†é‡Žã§æ´»ç”¨ã•ã‚Œã¦ã„ã‚‹ï¼Žãã—ã¦ç¾ä»£ã¯ï¼Œåœ§ç¸®ã—ãªã‘ã‚Œã°ãªã‚‰ãªã„å·¨å¤§ãªãƒ‡ãƒ¼ã‚¿ï¼Œåœ§ç¸®ãƒ‡ãƒ¼ ã‚¿ã«é«˜é€Ÿã‚¢ã‚¯ã‚»ã‚¹ã™ã‚‹ãŸã‚ã®ç†è«–ï¼Œãã‚Œã‚’å®Ÿç¾ã™ã‚‹ãƒãƒ¼ãƒ‰ã‚¦ã‚§ã‚¢ã®å…¨ã¦ãŒãã‚ã£ã¦ã„ã‚‹ï¼Žãã“ã§ï¼Œæ–‡å—åˆ—åœ§ç¸®ã«ã¤ã„ã¦è§£ èª¬ã—ï¼Œåˆå¦è€…ã®é“æ¨™ã¨ã—ãŸã„ï¼Ž ã‚ãƒ¼ãƒ¯ãƒ¼ãƒ‰ï¼šãƒ‡ãƒ¼ã‚¿åœ§ç¸®ï¼Œæ–‡å—åˆ—ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ï¼Œæ–‡æ³•åœ§ç¸®ï¼Œã‚¹ãƒˆãƒªãƒ¼ãƒ ãƒ‡ãƒ¼ã‚¿ ï¼Žã¯ ã˜ ã‚ã« ã©ã‚“ãªå¤§ããªæƒ… ç¸®å‰å¾Œã®ã‚µã‚¤ã‚ºã®æ¯”ãŒ 1% æœªæº€ã¨ã„ã†ã®ã¯æå¤±ãŒãªã„å¯ é€†åœ§ç¸®ã®æ¡ä»¶ã®ä¸‹ã§ã¯é€šå¸¸è€ƒãˆã‚‰ã‚Œãªã„å€¤ã§ã‚ã‚‹ï¼Ž ã•ã¦å‰ç½®ããŒ
chezou 2014/01/15
algorithm
ãƒªãƒ³ã‚¯
SPYSEEã®ã¤ãªãŒã‚Šãƒžã‚¤ãƒ‹ãƒ³ã‚°ã®ã¯ãªã—ã€‚ - TMBã®ãŠã¼ãˆãŒã
ã‚ªãƒ¼ãƒžÃ—ã‚¯ãƒƒã‚¯ãƒ‘ãƒƒãƒ‰å‹‰å¼·ä¼šã«å‚åŠ ã—ã¾ã—ãŸ ã”ã¯ã‚“ãŒç¾Žå‘³ã—ã‹ã£ãŸã€‚ ã¾ãˆãŠã http://spysee.jp/ã®ãªã‹ã®ã²ã¨ã§ã™ã€‚ ãƒ•ãƒãƒ³ãƒˆã‚¨ãƒ³ãƒ‰ã‚„ã‚¤ãƒ³ãƒ•ãƒ©ç³»ã¯ã‚·ãƒ£ãƒƒãƒãƒ§ãƒ¼ã‚„id:amachangãŒã‚„ã£ã¦ã„ã‚‹ã®ã§ã€ãã‚Œä»¥å¤–ã®ã¨ã“ã‚ã‚„ã£ã¦ã¾ã™ã€‚ä¸»ã«ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã€‚ã¤ãªãŒã‚Šã®æŠ½å‡ºæ‰‹æ³•ã‚„åŒå§“åŒåå‡¦ç†æ‰‹æ³•ã‚’é–‹ç™ºã—ã¾ã—ãŸã€‚ æ™‚ã€…ã€ãªã‹ã®ã²ã¨ã¨ã—ã¦ã„ã‚ã‚“ãªä¼šåˆã«å‡ºæ²¡ã—ã¦ã¾ã™ã€‚ãã®ãŸã³ã«ã€ ã€Œã¤ãªãŒã‚Šã©ã†ã‚„ã£ã¦ã§ãã¦ã‚“ã®ãƒ¼ï¼Ÿã€ ã€ŒåŒå§“åŒåã©ã†ãªã£ã¦ã‚“ã®ãƒ¼ï¼Ÿã€ ãªã©èžã‹ã‚Œã¾ã™ãŒã€è©³ç´°ã«ç”ãˆãŸã“ã¨ã¯ã‚ã‚Šã¾ã›ã‚“ã§ã—ãŸã€‚about SPYSEEçš„ãªè©±ã¯IVSã®Launch Pad(å‹•ç”»)ãªã©ã§è©±ã—ãŸã“ã¨ã¯ã‚ã‚Šã¾ã™ãŒã€ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã®è©³ã—ã„ã¨ã“ã‚ã¾ã§ã¯æ™‚é–“ãªãã¦è©±ã—ã¦ãŠã‚Šã¾ã›ã‚“ã€‚ ã•ã¦å…ˆæ—¥ã€ã‚ªãƒ¼ãƒžÃ—ã‚¯ãƒƒã‚¯ãƒ‘ãƒƒãƒ‰åˆåŒå‹‰å¼·ä¼š ã‚’é–‹å‚¬ã—ã¾ã—ãŸã€‚ãã“ã§ãŠæ™‚é–“ã„ãŸã ãã€ã€ŒSPYSEEã®ã¤ãªãŒã‚Šãƒžã‚¤ãƒ‹ãƒ³ã‚°æ‰‹æ³•ã€ã¨ã„ã†é¡Œç›®ã§è¬›æ¼”ã•ã›
chezou 2013/11/08
algorithm

SPYSEE
ãƒªãƒ³ã‚¯
GitHub - Mendeley/mrec: A recommender systems development and evaluation package by Mendeley
mrec is a Python package developed at Mendeley to support recommender systems development and evaluation. The package currently focuses on it em similarity and other methods that work well on implicit feedback, and on experimental evaluation. Why another package when there are already some really good software projects implementing recommender systems? mrec tries to fill two small gaps in the curre
chezou 2013/10/12
Mendeleyã®æŽ¨è–¦ãƒ©ã‚¤ãƒ–ãƒ©ãƒªã€‚ã‚ã¨ã§ãŸã‚ãã†

algorithm
ãƒªãƒ³ã‚¯
ãƒã‚±ãƒƒãƒˆã‚½ãƒ¼ãƒˆ - Wikipedia
ãƒã‚±ãƒƒãƒˆã‚½ãƒ¼ãƒˆï¼ˆè‹±: bucket sortï¼‰ã¯ã€ã‚½ãƒ¼ãƒˆã®ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã®ä¸€ã¤ã€‚ãƒã‚±ãƒ„ã‚½ãƒ¼ãƒˆã€ãƒ“ãƒ³ã‚½ãƒ¼ãƒˆï¼ˆè‹±: bin sortï¼‰ãªã©ã¨ã‚‚ã„ã†ã€‚ãƒã‚±ãƒ„æ•° k å€‹ä½¿ã£ãŸå ´åˆã€ã‚ªãƒ¼ãƒ€ãƒ¼ã¯O(n + k)ã¨ãªã‚Šã€ã‚½ãƒ¼ãƒˆã™ã‚‹è¦ç´ æ•°nã¨k ã‚’ç„¡é–¢ä¿‚ã«ã§ãã‚‹å ´åˆç·šå½¢æ™‚é–“ã‚½ãƒ¼ãƒˆã¨ãªã‚‹ãŒã€è¦ç´ é–“ã®å…¨é †åºé–¢ä¿‚ã‚’ç”¨ã„ã‚‹ã‚½ãƒ¼ãƒˆã¨ã¯ç•°ãªã‚Šã€ã‚ãƒ¼ã®å–ã‚Šã†ã‚‹å€¤ãŒkç¨®é¡žã§ã‚ã‚‹ã€ã¨ã„ã†å…¥åŠ›ã«ã‚ˆã‚Šå¼·ã„åˆ¶é™ã‚’è¦æ±‚ã™ã‚‹ã‚½ãƒ¼ãƒˆã§ã‚ã‚‹ã€‚ ãƒã‚±ãƒƒãƒˆã‚½ãƒ¼ãƒˆã®æ¦‚å¿µ æ•´åˆ—ã—ãŸã„ãƒ‡ãƒ¼ã‚¿ã®å–ã‚Šã†ã‚‹å€¤ãŒmç¨®é¡žã§ã‚ã‚‹ã¨ãã€må€‹ã®ãƒã‚±ãƒ„ã‚’ç”¨æ„ã—ã¦ãŠãã€å€¤ã”ã¨ã«1å€‹ã®ãƒã‚±ãƒ„ã‚’å¯¾å¿œã¥ã‘ã‚‹ã€‚å…ƒã®ãƒ‡ãƒ¼ã‚¿åˆ—ã‚’èµ°æŸ»ã—ã¦ã€å„ãƒ‡ãƒ¼ã‚¿ã‚’å¯¾å¿œã™ã‚‹ãƒã‚±ãƒ„ã«å…¥ã‚Œã¦ã„ãã€‚ã“ã®å‡¦ç†ãŒçµ‚ã‚ã£ãŸå¾Œã€æ•´åˆ—ã—ãŸã„é †åºã«å¾“ã£ã¦ãƒã‚±ãƒ„ã‹ã‚‰å€¤ã‚’å–ã‚Šå‡ºã›ã°ã€ãƒ‡ãƒ¼ã‚¿ã‚’ã‚½ãƒ¼ãƒˆã™ã‚‹ã“ã¨ãŒã§ãã‚‹ã€‚ å®‰å®šã‚½ãƒ¼ãƒˆã‚’å®Ÿç¾ã™ã‚‹ãŸã‚ã«ã¯ã€åŒã˜ãƒã‚±ãƒ„ã«å…¥ã£ã¦ã„ã‚‹ãƒ‡ãƒ¼ã‚¿ã¯å…¥ã‚ŒãŸã¨ãã¨åŒã˜é †åºã§å–ã‚Šå‡ºã™å¿…è¦ãŒã‚ã‚‹ã€‚é †
chezou 2013/09/30
"ã‚¹ãƒªãƒ¼ãƒ—ã‚½ãƒ¼ãƒˆ ãƒã‚±ãƒƒãƒˆã‚½ãƒ¼ãƒˆã®ãƒã‚±ãƒ„ã‚’ãƒ¡ãƒ¢ãƒªç©ºé–“ã®ä»£ã‚ã‚Šã«æ™‚é–“ã«ç½®ãæ›ãˆãŸã‚‚ã®ã§ã‚ã‚‹"

algorithm
ãƒªãƒ³ã‚¯
ã‚¦ã‚§ãƒ¼ãƒ–ãƒ¬ãƒƒãƒˆè¡Œåˆ—ã®rankã‚’2å€é«˜é€ŸåŒ–ã™ã‚‹æ¡ˆã«ã¤ã„ã¦è‘—è€…ã‹ã‚‰ã‚³ãƒ¡ãƒ³ãƒˆã‚’ã„ãŸã ã„ãŸ - EchizenBlog-Zwei
ä¸€å¹´ãã‚‰ã„å‰ã«ã‚¦ã‚§ãƒ¼ãƒ–ãƒ¬ãƒƒãƒˆè¡Œåˆ—ã®rankè¨ˆç®—ã‚’2å€é«˜é€ŸåŒ–ã™ã‚‹æ–¹æ³•ã‚’æ€ã„ã¤ã„ãŸã€‚ è©³ç´°ã¯DSIRNLPç™ºè¡¨è³‡æ–™(http://ja.scribd.com/doc/102636443/Wavelet-Matrix)ã®P.56ä»¥é™ã€‚ æœ¬å½“ã«ã‚¤ã‚±ã¦ã„ã‚‹ã®ã‹è‡ªä¿¡ãŒãªã‹ã£ãŸã®ã§ã‚¦ã‚§ãƒ¼ãƒ–ãƒ¬ãƒƒãƒˆè¡Œåˆ—è«–æ–‡ã®first authorã§ã‚ã‚‹F. Claudeæ°ã«"ã‚¢ãƒ‰ãƒã‚¤ã‚¹ãŠé¡˜ã„ã—ã¾ã™"ã¨é€ã£ãŸãƒ¡ãƒ¼ãƒ«ã«è¿”ä¿¡ãŒã‚ã£ãŸã€‚ "I guess that if your alphabet is not bit, this might be a very good alternative to pointer-based wavelet trees! :D"ã¨ã®ã“ã¨ã€‚ã‚ã§ãŸã—ã‚ã§ãŸã—ã€‚
chezou 2013/09/29
ã‚¦ã‚§ãƒ¼ãƒ–ãƒ¬ãƒƒãƒˆ

algorithm
ãƒªãƒ³ã‚¯
å¤šè…•ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆ ãƒ†ã‚¹ãƒˆ - ã‚¢ãƒŠãƒªãƒ†ã‚£ã‚¯ã‚¹ ãƒ˜ãƒ«ãƒ—
Google ã‚¢ãƒŠãƒªãƒ†ã‚£ã‚¯ã‚¹ ã‚¦ã‚§ãƒ–ãƒ†ã‚¹ãƒˆã®åŸºç›¤ã‚’æˆã™çµ±è¨ˆæ‰‹æ³•ã«ã¤ã„ã¦èª¬æ˜Žã—ã¾ã™ã€‚Google ã‚¢ãƒŠãƒªãƒ†ã‚£ã‚¯ã‚¹ã§ã¯ã€ã‚¦ã‚§ãƒ–ãƒ†ã‚¹ãƒˆã®æ‰‹æ³•ã¨ã—ã¦å¤šè…•ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆæ–¹å¼ã‚’æŽ¡ç”¨ã—ã¦ã„ã¾ã™ã€‚å¤šè…•ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆ ãƒ†ã‚¹ãƒˆã«ã¯ã€æ¬¡ã®ã‚ˆã†ãªç‰¹å¾´ãŒã‚ã‚Šã¾ã™ã€‚ æœ€ã‚‚åˆ©ç›Šã®å¤§ãã„é¸æŠžè‚¢ã®ç‰¹å®šã‚’ç›®æ¨™ã¨ã™ã‚‹ ãƒ©ãƒ³ãƒ€ãƒ åˆ†å¸ƒãŒãƒ†ã‚¹ãƒˆã®é€²è¡Œã¨ã¨ã‚‚ã«æ›´æ–°ã•ã‚Œã‚‹ ã€Œå¤šè…•ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆï¼ˆmulti-armed banditï¼‰ã€ã¨ã„ã†åå‰ã¯ã€ãã‚Œãžã‚Œã«ç•°ãªã‚‹è¦‹è¾¼ã¿é…å½“çŽ‡ãŒè¨å®šã•ã‚ŒãŸã€ã€ŒOne-armed banditï¼ˆç‰‡è…•ã®ç›—è³Šï¼‰ã€ã¨ã„ã†ã‚¹ãƒãƒƒãƒˆ ãƒžã‚·ãƒ³ãŒè¤‡æ•°ä¸¦ã‚“ã§ã„ã‚‹çŠ¶æ³ã‚’æ¨¡ã—ãŸä»®èª¬ãƒ†ã‚¹ãƒˆã¨ã„ã†æ„å‘³ã‚’æŒã£ã¦ã„ã¾ã™ã€‚ã‚¹ãƒãƒƒãƒˆ ãƒžã‚·ãƒ³ã®ãƒ—ãƒ¬ã‚¤ãƒ¤ãƒ¼ã¯ã€æœ€ã‚‚è¦‹è¾¼ã¿é…å½“çŽ‡ãŒé«˜ã„ã‚¹ãƒãƒƒãƒˆ ãƒžã‚·ãƒ³ã‚’è¦‹ã¤ã‘å‡ºã™å¿…è¦ãŒã‚ã‚‹ä¸€æ–¹ã§ã€åˆ©ç›Šã‚’æœ€å¤§åŒ–ã™ã‚‹å¿…è¦ã‚‚ã‚ã‚Šã¾ã™ã€‚ã“ã®çŠ¶æ³ã§ã¯ã€ã“ã‚Œã¾ã§ã®é…å½“çŽ‡ãŒæœ€ã‚‚å„ªã‚Œã¦ã„ã‚‹ãƒžã‚·ãƒ³ã®ã¿ã‚’ãƒ—ãƒ¬ã‚¤ã™ã‚‹ã‹ã€ãã‚Œã¨ã‚‚ã•ã‚‰ã«é…å½“çŽ‡
chezou 2013/08/15
algorithm

bandit
ãƒªãƒ³ã‚¯
ç¬¬ï¼’ï¼˜å›žåŠ©æ•™ã®ä¼š
ä»Šå›žã¯æ•°ç†æƒ…å ±å¦ç¬¬ï¼‘ç ”ç©¶å®¤ã®æœ¬å¤šæ·³ä¹Ÿã•ã‚“ã«ç™ºè¡¨ã—ã¦ã„ãŸã ãã¾ã—ãŸã€‚ã‚¿ã‚¤ãƒˆãƒ«ã¯ã€Œå¤šè…•ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆå•é¡Œã«ãŠã‘ã‚‹æ¼¸è¿‘æœ€é©æˆ¦ç•¥ã«ã¤ã„ã¦ã€ã§ã™ã€‚æœ¬å¤šã•ã‚“ã¯æ•°ç†4ç ”ã§ä¿®å£«ã‚’ã•ã‚Œã€ãã®å¾Œæ•°ç†1ç ”ã§åšå£«å·ã‚’ä»Šå¹´3æœˆã«æ‰€å¾—ã—æ˜¥ã‹ã‚‰1ç ”ã§åŠ©æ•™ã‚’ã•ã‚Œã¦ã„ã¾ã™ã€‚æœ¬å¤šã•ã‚“ã¯æƒ…å ±ç†è«–ã¨çµ±è¨ˆçš„æ©Ÿæ¢°å¦ç¿’ã®ä¸¡æ–¹ã«èˆˆå‘³ã‚’ãŠæŒã¡ã§ã€ä»Šå›žã®è©±ã¯ä¸»ã«å¾Œè€…ã®å•é¡Œã‚’æ‰±ã£ã¦ã„ã¾ã™ã€‚ å¤šè…•ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆå•é¡Œã¨ã„ã†ã®ã¯ã€ã‚®ãƒ£ãƒ³ãƒ–ãƒ©ãƒ¼ãŒã‚¹ãƒãƒƒãƒˆãƒžã‚·ãƒ³ã§é‡‘å„²ã‘ã‚’ç‹™ã†å ´åˆã«ã€ $k$å°ã‚ã‚‹æ€§èƒ½ã®ç•°ãªã‚‹ãƒžã‚·ãƒ³ã®ä¸ã‹ã‚‰$n$å›ž($n>> k$)ãƒ—ãƒ¬ã‚¤ã™ã‚‹ä¸ã§å„ãƒžã‚·ãƒ³ã®å ±é…¬ã®æœŸå¾…å€¤ã‚’æŽ¨å®šã—ã€ãã‚Œã«å¿œã˜ã¦å„å°ã®ãƒ—ãƒ¬ã‚¤å›žæ•°ã‚’å ±é…¬ãŒã§ãã‚‹ã ã‘å¤§ãããªã‚‹ã‚ˆã†ã«æ±ºã‚ã‚‹å•é¡Œã§ã™ã€‚ä»®å®šã¨ã—ã¦ã¯å„ãƒžã‚·ãƒ³ã®å ±é…¬ãŒã‚ã‚‹ç¢ºçŽ‡åˆ†å¸ƒã®é›†åˆã«å±žã™ã‚‹ã“ã¨ã€ã¾ãŸãã®é›†åˆï¼ˆä¾‹ãˆã°ãƒ™ãƒ«ãƒŒãƒ¼ã‚¤åˆ†å¸ƒã€ã¾ãŸã¯æ£è¦åˆ†å¸ƒã§ã‚ã‚‹ã“ã¨ï¼‰ã¯æ—¢çŸ¥ã§ã‚ã‚‹ã“ã¨ã§ã™ã€‚å¿œç”¨ä¾‹ã¯ã‚®ãƒ£ãƒ³ãƒ–ãƒ«ã®ä»–ã«ã„ãã¤ã‚‚ã‚ã‚‹ãã†ã§ã™ã€‚
chezou 2013/08/07
å¤šè…•ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒ‰ã®ãŠè©±

algorithm

machine learning
ãƒªãƒ³ã‚¯
ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã«ã‚ˆã‚‹æœ€é©åŒ–æ‰‹æ³•
TOPICS Programming , Database ç™ºè¡Œå¹´æœˆæ—¥ 2013å¹´07æœˆ30æ—¥ ISBN 978-4-87311-627-3 åŽŸæ›¸ Bandit Algorithms for Website Optimization FORMAT æœ¬æ›¸ã¯ã€ã€Œå¤šè…•ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆå•é¡Œã€ã¨å‘¼ã°ã‚Œã‚‹å•é¡Œã‚’è§£ããŸã‚ã®ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã‚’ã€Webã‚µã‚¤ãƒˆã®æœ€é©åŒ–ã¨ã„ã†ä¾‹ã‚’ã‚‚ã¨ã«è§£èª¬ã™ã‚‹æ›¸ç±ã§ã™ã€‚ ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã«é–¢ã™ã‚‹åŸºæœ¬çš„ãªçŸ¥è˜ã«ã¤ã„ã¦ã€æ—¢å˜ç ”ç©¶ã«ã¤ã„ã¦ã®ç†è§£ã‚’ååˆ†ã«å¾—ã¦ã€å¤šè…•ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆå•é¡Œã«ã¤ã„ã¦ã®è³‡æ–™ã‚’è‡ªåŠ›ã§èªã‚ã‚‹ã‚ˆã†ã«ã™ã‚‹ã“ã¨ã‚’ç›®çš„ã¨ã—ã¦ã„ã¾ã™ã€‚ A/Bãƒ†ã‚¹ãƒˆã®ã‚ˆã†ãª2è€…æŠžä¸€ã§ã¯ãªãã€æ–°ã—ã„ã‚¢ã‚¤ãƒ‡ã‚¢ã®æŽ¢ç´¢ã¨ã€æ—¢å˜ã®ã‚¢ã‚¤ãƒ‡ã‚¢ã‹ã‚‰æœ€å¤§é™ã®åˆ©ç›Šã‚’å¼•ãã ã™ã¨ã„ã†çŸ›ç›¾ã™ã‚‹2ã¤ã®å•é¡Œã‚’è§£æ±ºã™ã‚‹ãŸã‚ã®ä¸€åŠ©ã¨ãªã‚‹ã§ã—ã‚‡ã†ã€‚ãªãŠæœ¬æ›¸ã¯Ebookã®ã¿ã®è²©å£²ã¨ãªã‚Šã¾ã™ã€‚ yuku_tã•ã‚“ã«ã‚ˆã‚‹æœ¬æ›¸ã®è‹±èªžç‰ˆã¨ãƒãƒ³ãƒ‡
chezou 2013/08/04
æ—¥æœ¬èªžç‰ˆå‡ºã¦ãŸã‚“ã ã€‚è²·ãŠã†

algorithm
ãƒªãƒ³ã‚¯
é«˜é€Ÿã‹ã¤çœãƒ¡ãƒ¢ãƒªãªbit vectorã€ŒsucBVã€ã‚’ä½œã‚‹
ã¯ã˜ã‚ã« å¤§è¦æ¨¡ãªãƒ‡ãƒ¼ã‚¿ã‚’æ‰±ã†ã‚¢ãƒ—ãƒªã‚±ãƒ¼ã‚·ãƒ§ãƒ³ã§ã¯ã€é€Ÿåº¦ã¨ã¨ã‚‚ã«ä½œæ¥é ˜åŸŸé‡ã‚‚å¤§ããªå•é¡Œã¨ãªã‚Šã¾ã™ã€‚ä½œæ¥é ˜åŸŸãŒãƒ¡ã‚¤ãƒ³ãƒ¡ãƒ¢ãƒªã«åŽã¾ã‚‰ãªã„å ´åˆã€ã‚¹ãƒ¯ãƒƒãƒ”ãƒ³ã‚°ãŒç™ºç”Ÿã—ã€å¤§å¹…ãªé€Ÿåº¦ä½Žä¸‹ã«ã¤ãªãŒã‚Šã¾ã™ã€‚ãã®ãŸã‚è¿‘å¹´ã€ãƒ‡ãƒ¼ã‚¿æ§‹é€ ã¯é«˜é€Ÿãªã ã‘ã§ãªãã€ä½œæ¥é ˜åŸŸé‡ãŒå°ã•ã„ã“ã¨ã‚‚æ±‚ã‚ã‚‰ã‚Œã¦ã„ã¾ã™ã€‚ä»Šå›žç´¹ä»‹ã™ã‚‹ãƒ‡ãƒ¼ã‚¿æ§‹é€ ã¯ã€Œæ“ä½œä»˜ãbit vectorï¼ˆSUCcinct Bit Vector:sucBVï¼‰ã€ã§ã™ã€‚sucBVã¯ã€åœ§ç¸®ç´¢å¼•ã‚„Succinct Data Structureãªã©ã€ãƒ‡ãƒ¼ã‚¿ã‚’ã‚³ãƒ³ãƒ‘ã‚¯ãƒˆã«è¡¨ç¾ã™ã‚‹éš›ã«é‡è¦ãªãƒ‡ãƒ¼ã‚¿æ§‹é€ ã§ã™ã€‚STLã®vector<bool>ã¨åŒæ§˜ã«ã€bitåˆ—æƒ…å ±B[0....n-1]ã‚’ä¿å˜ã—ã¾ã™ã€‚ã“ã®bitåˆ—æƒ…å ±ã¯å‰ã‚‚ã£ã¦ä¸Žãˆã‚‰ã‚Œã€å¤‰æ›´ãŒç„¡ã„ã“ã¨ã‚’å‰æã¨ã—ã¾ã™ã€‚sucBVã¯ã€æ¬¡ã®äºŒã¤ã®æ“ä½œã‚’å®šæ•°æ™‚é–“ã§ã‚µãƒãƒ¼ãƒˆã—ã¾ã™ã€‚ rank(p,bit)â€•â€•B[0...p]ä¸ã®bitï¼ˆbitã¯1
chezou 2013/05/22
programming

algorithm
ãƒªãƒ³ã‚¯
kfka.net
The domain kfka.net may be for sale. Please send an inquiry to [email protected]
chezou 2013/05/05
algorithm
ãƒªãƒ³ã‚¯
A/Bãƒ†ã‚¹ãƒˆã‚ˆã‚Šã™ã”ã„ï¼Ÿãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã¨ã¯ä¸€ä½“ä½•è€…ã‹ - Qiita
ã‚ªãƒãƒžå¤§çµ±é ˜ã®å†é¸ã«å¤§ããå¯„ä¸Žã—ãŸã“ã¨ã§å¤§ããªæ³¨ç›®ã‚’é›†ã‚ã¦ã„ã‚‹A/Bãƒ†ã‚¹ãƒˆã€‚A/Bãƒ†ã‚¹ãƒˆã‚’å°Žå…¥ã—ãŸã€ã™ã‚‹ã“ã¨ã‚’æ¤œè¨Žã—ã¦ã„ã‚‹ã€ã¨ã„ã†é–‹ç™ºç¾å ´ã‚‚å¤šã„ã®ã§ã¯ãªã„ã ã‚ã†ã‹ã€‚ ãã‚“ãªä¸ã€Webä¸Šã§æ¬¡ã®ã‚ˆã†ãªè°è«–ã‚’è¦‹ã¤ã‘ãŸã€‚ 20 lines of code that will beat A/B testing every time Why multi-armed bandit algorithm is not â€œbetterâ€ than A/B testing ä¸€è¨€ã§ã¾ã¨ã‚ã‚‹ã¨ã€ŒA/Bãƒ†ã‚¹ãƒˆã‚ˆã‚Šãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã®æ–¹ãŒã™ã”ã„ã‚ˆã€ã€Œã„ã‚„ã„ã‚„A/Bãƒ†ã‚¹ãƒˆã®æ–¹ãŒã™ã”ã„ã—ã€ã¨ã„ã†ã“ã¨ã ã€‚ ã§ã€ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã¨ã¯ä¸€ä½“ä½•è€…ãªã®ã‹ï¼Ÿ ãã“ã§Bandit Algorithms for Website Optimization (O'REILLY)ã‚’èªã‚“ã§ã¿ãŸã€‚ãã®çµæžœåˆ†ã‹ã£ãŸã“ã¨ã‚’è¸ã¾ãˆã¦ã–ã£ãã‚Šã¨
chezou 2013/04/03
æœ¬ã‚ˆã‚‚ã†ã€‚æ´»ç”¨ã§ä½¿ã†æ–¹æ³•ã¨æŽ¢æ±‚ã§ä½¿ã†æ–¹æ³•ã®é‡ã¿ä»˜ã‘ã®å•é¡Œã«è½ã¨ã™ã®ã‹

algorithm

book
ãƒªãƒ³ã‚¯
åŠ£ãƒ¢ã‚¸ãƒ¥ãƒ© - æ©Ÿæ¢°å¦ç¿’ã®ã€Œæœ±é·ºã®æœWikiã€
åŠ£ãƒ¢ã‚¸ãƒ¥ãƒ© (submodular)â€ é›¢æ•£æœ€é©åŒ–å•é¡Œã‚’è§£ãã¨ãï¼Œç›®çš„é–¢æ•°ã«åŠ£ãƒ¢ã‚¸ãƒ¥ãƒ©æ€§ãŒã‚ã‚Œã°ï¼Œå¤šé …å¼æ™‚é–“ã§è§£ãã“ã¨ãŒã§ãã‚‹ï¼Ž æœ‰é™é›†åˆ $V$ ã«å¯¾ã—ã¦ï¼Œãã®ä»»æ„ã®éƒ¨åˆ†é›†åˆ $X\subseteq V$ ã‹ã‚‰å®Ÿæ•°ã¸ã®é–¢æ•°ã‚’ $f(X)$ ã¨ã™ã‚‹ï¼Žã“ã®é–¢æ•°ãŒåŠ£ãƒ¢ã‚¸ãƒ¥ãƒ©é–¢æ•°ã§ã‚ã‚‹ã¨ã¯ï¼Œä»»æ„ã®éƒ¨åˆ†é›†åˆ $X,Y\subseteq V$ ã«å¯¾ã—ã¦æ¬¡ã®æ€§è³ªãŒã‚ã‚‹ã“ã¨ï¼š \[f(X)+f(Y)\ge f(X\cap Y)+f(X\cup Y)\] ã“ã‚Œã¯æ¬¡ã®æ€§è³ªã¨ç‰ä¾¡ \[X\subseteq Y\subseteq V,\, x\in V\backslash Y,\;f(X\cup\{x\})-f(X)\ge f(Y\cup\{x\})-f(Y)\] åŠ£ãƒ¢ã‚¸ãƒ¥ãƒ©é–¢æ•°ã‚’æœ€å°åŒ–ã™ã‚‹éƒ¨åˆ†é›†åˆã¯ï¼Œ$|V|$ ã®5ã€œ6ä¹—ã®å¤šé …å¼æ™‚é–“ã§è§£ãã“ã¨ãŒã§ãã‚‹ï¼Ž ã•ã‚‰ã«ï¼Œ\(f(X)=f(V\ba
chezou 2013/03/10
algorithm

submodularity
ãƒªãƒ³ã‚¯
GoogleNewsã®ãƒ¬ã‚³ãƒ¡ãƒ³ãƒ‰ã®ä¸èº« - UMEko Branding
å…ˆæ—¥ã€å…¨ä½“ã‚¼ãƒŸã§ç™ºè¡¨ã—ãŸã¨ãã®å†…å®¹ã§ã™ãŒã€ã“ã“ã«ã¾ã¨ã‚ã¨ãã¾ã™ã€‚ã€‚GoogleNewsã®ãƒ¬ã‚³ãƒ¡ãƒ³ãƒ‰ã®ä¸èº«ã‚’è¿½ã£ãŸè«–æ–‡ã®è¦ç´„ã§ã™ã€‚å°‘ã—å‰ã®å…¨ä½“ã‚¼ãƒŸã§ç”¨ã„ãŸè³‡æ–™ã§ã™ã€‚ã‚½ãƒ¼ã‚¹ï¼šAbhinandan Das,Mayur Datar,Ashutosh Garg,Shyam Rajaram,"Google News Personalization: Scala ble OnlineCollaborative Filtering",WWW2007ä¸å‹‰å¼·ãªå€‹æ‰€ãŒå¤šã€…ã‚ã‚Šã¾ã™ã®ã§ã€èª¤ã£ã¦ã„ã‚‹ç®‡æ‰€ç‰ã‚ã‚Šã¾ã—ãŸã‚‰ã€æ˜¯éžã”æŒ‡æ‘˜ãã ã•ã„ã€‚ å€‹äººçš„ã«ã¯ã€æœ€è¿‘ã®ãƒ¢ãƒ‡ãƒ«ãƒ™ãƒ¼ã‚¹ã®æ‰‹æ³•ã®å‹‰å¼·ãƒ»ãŠã•ã‚‰ã„ã¨ã„ã†æ„å‘³ã§ç”¨ã„ã¦ã„ã‚‹ã®ã§ã€GoogleNewsç‹¬è‡ªã®æ‹¡å¼µãªã‚Šå®Ÿè£…ã®éƒ¨åˆ†ã®å†…å®¹ãŒçœã‹ã‚Œã¦ã„ã‚‹å ´åˆãŒã‚ã‚Šã¾ã™ã€‚ã¾ãŸã€ãƒ‡ãƒ¼ã‚¿æ§‹é€ ã‚„MapReduceã‚’ç”¨ã„ãŸè¨ˆç®—ã®ä»•çµ„ã¿ã®éƒ¨åˆ†ã¯ã€ã“ã“ã§ã¯çœç•¥ã—ã¦ã„ã¾ã™ã€‚ã€‚ä¸€å¿œã€ å…¨ä½“åƒã€€ãƒ»LSH(Lo
chezou 2013/03/10
MinHashingã¨pLSIã‚’ä½¿ã£ã¦å‹•çš„ã«å¤‰åŒ–ã™ã‚‹vectorã«å¯¾ã—ã¦recommendationã—ã¦ã‚‹ã®ã‹ï¼

lsh

recommend

algorithm
ãƒªãƒ³ã‚¯
ãŸã¶ã‚“30åˆ†ã§ã‚ˆãã‚ã‹ã‚‹LOUDSå…¥é–€ - EchizenBlog-Zwei
IMEæœ¬ã®åŠ¹æžœã§LOUDSã®èªçŸ¥åº¦ãŒé«˜ã¾ã£ã¦ããŸæ°—ãŒã™ã‚‹ã€‚ãŒã€ä¸€æ–¹ã§é›£ã—ã„ã¨ã„ã†æ„è¦‹ã‚‚ãƒãƒ©ãƒ›ãƒ©ã‚ã‚‹æ§˜åã€‚ ã¨ã„ã†ã‚ã‘ã§LOUDSã‚’ã©ã“ã¾ã§ã‚ã‹ã‚Šã‚„ã™ãèª¬æ˜Žã§ãã‚‹ã‹ï¼Ÿã¨ã„ã†ã“ã¨ã«æŒ‘æˆ¦ã—ãŸããªã£ãŸã®ã§è¨˜äº‹ã‚’æ›¸ã„ã¦ã¿ã‚‹ã‚ˆã€‚ LOUDSã¨ã„ã†ã®ã¯æœ¨ã‚’è¡¨ã™ãƒ‡ãƒ¼ã‚¿æ§‹é€ ã€‚æœ¨ã¨ã„ã†ã®ã¯ä»¥ä¸‹ã®ã‚ˆã†ãªã‚‚ã®ã‚’æƒ³åƒã™ã‚Œã°ã‚ˆã„ã€‚ ã“ã®æœ¨ã‚’è¡¨ã™ãƒ‡ãƒ¼ã‚¿æ§‹é€ ã‚’ä½œã‚ŠãŸã„ã€‚å˜ç´”ã«è€ƒãˆã‚‹ã¨å„ãƒŽãƒ¼ãƒ‰æ¯Žã«åãƒŽãƒ¼ãƒ‰ã®IDã‚’æŒãŸã›ã¦ãŠã‘ã°ã‚ˆã„ã€‚ã¤ã¾ã‚Š 0 => {1, 4, 5} 1 => {2, 3} 2 => {} 3 => {} 4 => {} 5 => {6} 6 => {7}ã¨ã„ã†ã‚ˆã†ãªã‚‚ã®ã‚’è€ƒãˆã‚‹ã€‚ã“ã“ã§å„IDã¨åãƒŽãƒ¼ãƒ‰æ•°ã‚’å„1ãƒã‚¤ãƒˆã§ç®¡ç†ã—ãŸã¨ã—ã¦ 0 => {1, 4, 5} # 1 + 3 = 4ãƒã‚¤ãƒˆ 1 => {2, 3} # 1 + 2 = 3ãƒã‚¤ãƒˆ 2 => {} # 1 + 0 = 1ãƒã‚¤ãƒˆ 3 => {}
chezou 2012/03/18
nlp

algorithm

LOUDS
ãƒªãƒ³ã‚¯
Tree Edit Distanceã¨è‡ªç„¶è¨€èªžå‡¦ç†ã¸ã®å¿œç”¨ - Preferred Networks Tech Blog
æµ·é‡Žã§ã™ã€‚ã¡ã‚‡ã£ã¨æ™‚é–“ãŒã‚ã„ã¦ã—ã¾ã„ã¾ã—ãŸãŒã€æ˜¨å¹´ã®12æœˆã«é–‹å‚¬ã•ã‚ŒãŸNTCIR-9ã¨ã„ã†ä¼šè°ã®Recognizing Inference in TExt (RITE)ã¨ã„ã†ã‚¿ã‚¹ã‚¯ã«ã€å‰è·ã®æ–¹ã€…ã¨å…±è‘—ã§å‡ºå ´ã—ã¾ã—ãŸã€‚ Syntactic Difference Based Approach for NTCIR-9 RITE Task. Yuta Tsuboi, Hiroshi Kanayama, Masaki Ohno and Yuya Unno. NTCIR-9, 2011. [pdf] å«æ„é–¢ä¿‚èªè˜ã¨ã„ã‚ã‚Œã‚‹ã“ã®ã‚¿ã‚¹ã‚¯ã¯ã€å¤§é›‘æŠŠã«è¨€ã†ã¨ä¸Žãˆã‚‰ã‚ŒãŸ2ã¤ã®æ–‡ãŒåŒã˜æ„å‘³ã®ã“ã¨ã‚’è¨€ã£ã¦ã„ã‚‹ã‹ã©ã†ã‹åˆ¤å®šã—ãªã•ã„ã¨ã„ã†ã‚¿ã‚¹ã‚¯ã§ã™ï¼ˆåŽ³å¯†ã«ã¯ä¸€æ–¹ã‹ã‚‰ã‚‚ã†ä¸€æ–¹ãŒå¸°çµã§ãã‚‹ã‹ã®åˆ¤å®šã§ã™ï¼‰ã€‚ä»Šæ—¥ã¯ã€ãã®ä¸ã§ä½¿ã£ãŸTree Edit Distance (TED) ã«ã¤ã„ã¦è§£èª¬ã—ã¾ã™ã€‚ TEDã¯2ã¤ã®é †åºä»˜ãæœ¨ãŒ
chezou 2012/02/13
Tree Edit Distanceé¢ç™½ã„

algorithm

nlp
ãƒªãƒ³ã‚¯
Blog
Numerical optimization is at the core of much of machine learning. In this post, we derive the L-BFGS algorithm, commonly used in batch machine learning applications.
chezou 2011/10/19
onlineã®åˆ†é¡žæ¥

algorithm

classify
ãƒªãƒ³ã‚¯
1 2 æ¬¡ã®ãƒšãƒ¼ã‚¸