æ¬è¨äºã¯ãCyberAgent Advent Calendar 2022 19æ¥ç®ã®è¨äºã§ãã ç®æ¬¡ ã¯ããã« åé¡è¨å® å調ãã£ã«ã¿ãªã³ã°ã®ããã®ç·å½¢ã¢ãã« iALS EASE é¢é£ããéç·å½¢ã¢ãã« å®åæ´»ç¨ ãããã« ã¯ããã« ã¡ãã£ã¢ DSCæå±ã®æ©æ¢°å¦ç¿ã¨ã³ã¸ãã¢ã§ãã¿ããã«ã®æ¨è¦ã·ã¹ãã ãæ å½ãã¦ããæ©çª (@runnlp)ã§ãã æè¿ãæ¨è¦ã·ã¹ãã ã触ãå§ãã¾ãããæ¨è¦ææ³ã¯ãå調ãã£ã«ã¿ãªã³ã°ãã³ã³ãã³ããã¼ã¹ããã¤ããªãããªã©æ§ã ã§ãããä»åã¯æããä»ã«è³ãã¾ã§é·ã使ããã¦ããå調ãã£ã«ã¿ãªã³ã°ã«ã¤ãã¦ã§ãã å調ãã£ã«ã¿ãªã³ã°ã§ã¯Deepç³»ã®ã¢ãã«ãããããåºãä¸ã§ãRecSys2022ã§çºè¡¨ãããè«æã§ã¯10年以ä¸åãã使ç¨ããã¦ããç·å½¢ã¢ãã«(iALS)ãDeepç³»ã®ã¢ãã«ã«å¹æµããçµæã§ããã¨å ±åããã¦ããèå³æ·±ãã§ããã¾ããæ¨è¦ã·ã¹ãã ãéçºããã«ããã£ã¦ãåé¡è¨
æ°çæé©å Advent Calendar 2022 ã®è¨äºã§ãã ä½ã®è©±ãã¨è¨ã㨠Pythonã§ã¯ãããæ°çæé©å âã±ã¼ã¹ã¹ã¿ãã£ã§ã¢ããªã³ã°ã®ã¹ãã«ã身ã«ã¤ãããâ ä½è :岩永äºé,ç³åé¿å¤ª,西æç´æ¨¹,ç°ä¸ä¸æ¨¹ãªã¼ã 社Amazon ä¸è¨ã®æ¸ç±ã®ç¬¬7ç« ã§ã¯ã次ã®ãããªåé¡ãåãæ±ã£ã¦ãã¾ãã ç´°ããç¹ã¯æ¸ç±ã«è²ãã¾ãããã¾ããçãã¼ã¿ã¨ãã¦æ¬¡ã®ãããªãã¼ã¿ãä¸ãããã¾ãã ããã¯ãããã·ã§ããã³ã°ãµã¤ãã®å©ç¨å±¥æ´ãéè¨ãã¦å¾ããããã®ã§ãããã¦ã¼ã¶ã¼ãåãååãé²è¦§ããåæ°ï¼freqï¼ã¨ããã®ååãæå¾ã«é²è¦§ããã®ãä½æ¥åãï¼rcenï¼ã®2ã¤ã®å¤ããããã®ã¦ã¼ã¶ã¼ã次ã«ãµã¤ãã«ãã£ã¦ããæã«ãå度ããã®ååãé²è¦§ãã確çï¼probï¼ãå®ç¸¾ãã¼ã¹ã§è¨ç®ãããã®ã§ããå®ç¸¾ãã¼ã¹ã®ãã¼ã¿ãªã®ã§ãã¬ã¿ã¬ã¿ããã°ã©ãã«ãªã£ã¦ãã¾ãããçè«çã«ã¯ã ã»freq ã大ããã»ã© prob ã¯å¤§ãããª
æè¿ï¼éåã³ã³ãã¥ã¼ã¿ã®è©±é¡ããã¥ã¼ã¹ãæ°èã§è¦ããããã¨ãå¢ãã¦ãã¾ããï¼ ãã®ä¸ã§æ°ã«ãªã£ã¦ããã®ãï¼çµåãæé©åã¨éåã³ã³ãã¥ã¼ã¿ï¼ç¹ã«éåã¢ãã¼ãªã³ã°ï¼ã«é¢ããæªããè¨èª¬ï¼ç§èªèº«ã¯ï¼å¤å ¸ã³ã³ãã¥ã¼ã¿ã§ã®ï¼çµåãæé©åã®ç 究ããã£ã¦ãã¦ï¼éåã³ã³ãã¥ã¼ã¿ãç 究ãã¦ããããã§ã¯ãªãã®ã§ããï¼ãããã«ããã¯ã¡ãã£ã¨ã»ã»ã»ã¨æãè¨èª¬ãä½åãè¦ããã¦ãã¾ããï¼ æè¿ã®ãéåãã«å¯¾ããéç±ã¶ãã¯åã¾ããã®ã§ï¼ããããæªããè¨èª¬ãåºã¾ãã®ã¯å°ããã®ã§ãï¼ãã§ã«Twitterä¸ã«ã¯ï¼âçµåãæé©åã¯ä»ã®ã³ã³ãã¥ã¼ã¿ã§ã¯è§£ããªãâã¨ãâã§ãéåãªãä¸ç¬ã§è§£ããâã¨ããåéãããã¦ãã¾ã£ã¦ãã人ãå¤æ°è¦ããã¾ã*1ï¼ ãããã«å±æ©æãè¦ãã¦ããã®ã§ï¼ãã®å ´ã§ãã¡ãã¨ææãã¦ãããã¨ã«ãã¾ããï¼ ä»åç£æ¥(TL;DR) âå¤å ¸ã³ã³ãã¥ã¼ã¿ã¯çµåãæé©åã解ããªãâ â å¤å ¸ã³ã³ãã¥ã¼ã¿ã§çµåãæé©åã解
ãµã¼ãã¹çµäºã®ãç¥ãã ãã¤ãYahoo! JAPANã®ãµã¼ãã¹ããå©ç¨ããã ãèª ã«ãããã¨ããããã¾ãã ã客æ§ãã¢ã¯ã»ã¹ããããµã¼ãã¹ã¯æ¬æ¥ã¾ã§ã«ãµã¼ãã¹ãçµäºãããã¾ããã ä»å¾ã¨ãYahoo! JAPANã®ãµã¼ãã¹ããæ顧ãã ããã¾ãããããããããé¡ããããã¾ãã
ããã«ã¡ã¯ãã¡ãã£ã¢ãã¸ãã¯åæé¨ã®ç±³ç° (@mathetake) ã§ãã ä»æ¥ã¯Gunosy社ã¨KDDI社ãå ±åã§éå¶ãããã¥ã¼ã¹ãã¹ã¨ãããã¥ã¼ã¹ã¢ããªã±ã¼ã·ã§ã³ã§ä½¿ããã¦ããé¢é£è¨äºæ¨è¦ã®ã¢ã«ã´ãªãºã ã«ã¤ãã¦æ¸ãããã¨æãã¾ãã ç¹ã«ãç´åå¹´åã«ç§ãå°å ¥ãKPIã®æ¹åã«æåããæ°ããã¢ã«ã´ãªãºã ã¨ãããã§ã³ã¢ã¨ãªãè¿ä¼¼è¿åæ¢ç´¢(Approximate Nearest Neighbor search)ã®æè¡ã«ã¤ãã¦è¿°ã¹ã¾ãã é¢é£è¨äºæ¨è¦ã¨ã¯ ãã®è¨äºã§ç´¹ä»ããé¢é£è¨äºæ¨è¦ã¨ã¯ããç¹å®ã®ãã¥ã¼ã¹ã«é¢é£ãããã¥ã¼ã¹ãæ¨è¦ãããã¨ãã§ãã ããå ·ä½çã«ã¯ãç¹å®ã®è¨äºãã¯ãªãã¯ããå¾ã«è¨äºé²è¦§ç»é¢ãä¸ã«ã¹ã¯ãã¼ã«ããã¨ç»å ´ãããããããè¨äºãã®æ ã«å¯¾ãã¦ãé¢é£ãããã¥ã¼ã¹ãæ¤ç´¢ãã¦è¡¨ç¤ºãããã¨ãæãã¾ã: ãã®ãããªæ ãè¨ç½®ããã¦ããäºã¯ä¸è¬çãªã¢ããªã±ã¼ã·ã§ã³ã«ããã¦ããèªç¶ã§ãããæ¨è¦ã·
巨大ãªéåã«å¯¾ãã¦ãå®æ°ã¡ã¢ãª&å®æ°æéã§è¿ä¼¼å¤ãè¨ç®ã§ããã確ççãã¼ã¿æ§é ã®ç´¹ä»ã¹ã©ã¤ãã§ãã æ¬ã¹ã©ã¤ãã¯ãæ ªå¼ä¼ç¤¾ã¨ãã»ã³ã¼ãã®ç¤¾å åå¼·ä¼(2018/08/30)ã«ã¦ä½¿ç¨ããããã®ã§ãã
ã¢ã«ã´ãªãºã ã¯ãä½ããã®åé¡ã解決ããæé ããæããã¢ã«ã´ãªãºã ã®è¯ãããã§ã½ããã¦ã¨ã¢ã®æ§è½ã決ã¾ãã¨è¨ã£ã¦ãéè¨ã§ã¯ãªããç§ãã¡ã®çæ´»ã¯ãé«åº¦ãªã¢ã«ã´ãªãºã ã§å®è£ ãããã½ããã¦ã¨ã¢ã«æ¯ãããã¦ãããã¨ã¬ãã¼ã¿ã¼ãä¿¡å·æ©ã®å¶å¾¡ã½ãããä¾ã«ã身è¿ãªã¢ã«ã´ãªãºã ã®ä¸ç«¯ãè¦ã¦ãããæã¯ãã¤ãã¨ã¬ãã¼ã¿ã¼åã大æ¸æ»ããã¿ã³ãæ¼ãã¦ããªããªãæ¥ãªãââããããªçµé¨ã¯èª°ããããã ãããã©ãããã°ã¨ã¬ãã¼
ã¾ã gzipã§æ¶èãï¼ç¥ï¼ 2016å¹´ã人é¡ãå¾ ã¡æãã§ãããgzipãå§åããOSSå§ç¸®ãã¼ã«zstd(Zstandard)ããªãªã¼ã¹ãããã«ãé¢ãããããªããããã¾ã話é¡ã«ãªã£ã¦ããªãã¦å¯ããã®ã§ãã¡ããããããã®è³ãããæ¯è¼è¨äºãæ¸ãã¾ãããå§ç¸®ãã¼ã«ã®ã«ã¿ãã°çã«çºãã¦ããã ãããã¨æãã¾ãã ã¯ããã« ï¼ãã®è¨äºã§è¨ãï¼å§ç¸®ãã¼ã«ã¨ã¯ä½ã å§ç¸®ãã¼ã«ã¨ããå¼ã³åã¯æ£ç¢ºã§ã¯ãªãï¼ã¯ãï¼ã§ããå¹³ããè¨ãã°ãgzipãbzip2ãxzãlz4ãªã©ã§ããã人ã«ãã£ã¦ã¯ãtarã®è£å´ã¨ãã¦ãã使ã£ã¦ãªãã¦ãèãããã¨ããªãããããã¾ããããããããã¨ãã¯ã¾ãgzipã®manpageã¨ãèªãã§ãã ããã ãããããããããã¼ã«ãä½ã¨å¼ã¹ã°ããã®ãããããªãã®ã§ãããã§ã¯å§ç¸®ãã¼ã«ã¨å¼ãã§ãã¾ãã ãããããã§ãããã¢ã¼ã«ã¤ãã§ã¯ããã¾ãããã¢ã¼ã«ã¤ãã¨ã¯å®æ ãä¸ã¤ã®ãã¡ã¤ã«ã«ãªã£ã¦ãããã©ã«
Algorithms of Recommender Systems ⨠http://www.kamishima.net/ â© Release: 2016-09-26 21:53:16 +0900; 9645c3b i 2007 11 [ 07] 2008 1 [ 08a] 2008 3 [ 08b] 3 (1) (3) GitHub https://github.com/tkamishima/recsysdoc TYPO GitHub pull request issues I II III IV V ii J. Riedl J. Herlocker GroupLens WWW iii ð¥ ð ð± ð î° ð¥ ð¦ ð ð ð± ð² ð ð î {1, ⦠, ð} î {1, ⦠, ð} îð¦ ð¦ î ð¥ x ð ðð¥ð¦ ð¥ ð¦ Ì ðð¥
ã¡ããã©äºå¹´ãããåï¼æ©æ¢°å¦ç¿ã§çãã¯ãã«ã®å§ç¸®ã«æ å ±æ¤ç´¢ã§ãã使ãããæ´æ°åã®å§ç¸®æè¡ã使ããã¨ãæ¤è¨ãããã¨ããã£ãï¼ãªã³ã©ã¤ã³å¦ç¿ã§ãã£ãã·ã¥ãå®è£ ãã¦ã¿ã - ny23ã®æ¥è¨ï¼ï¼ãã®ã¨ãã¯ï¼ãªã³ã©ã¤ã³ã§å§ç¸®ã Disk ã«ä¿åï¼å§ç¸®ãããã¯ãã«ã¯é½ã«ã¡ã¢ãªã«ç½®ããèªãï¼OS ã«ä»»ããï¼ã¨ããå®è£ ã§ï¼ï¼Disk IO ã®ãªã¼ãã¼ãããã大ããï¼å§ç¸®ããããã°ä½ã使ã£ã¦ã大差ãªãã¨ãã身ãèããªãçµè«ã«ãªã£ãï¼çµå±2è¡ã§æ¸ããæãåç´ãª Variable byte code ãæ¡ç¨ï¼ï¼ ãã以éã¯æ´æ°åå§ç¸®ã¢ã«ã´ãªãºã ã«é¢ããç¥èã NewPFD ãããã§æ¢ã¾ã£ã¦ããã®ã ãã©ï¼ã¤ãå æ¥ï¼ç¾æç¹ã§æéã®å§ç¸®ã¢ã«ã´ãªãºã ã®ææ¡ï¼ããæ°å¹´ã®ä¸»ãªæ´æ°åå§ç¸®ã¢ã«ã´ãªãºã ï¼Simple-8b (J. Software Pract. Exper. 2010), VSEncoding (CIKM 20
Amazon Redshiftã§ã¯ãã¼ãã«ãä½æããéã®è¦ç´ ã¨ãã¦å¹¾ã¤ããã¤ã³ããããã¾ãã1.ã¨5.ã«ã¤ãã¦ã¯ä¸è¬çãªRDBMSã§ãç¨ãããããããªæ¦å¿µã§ãããæ®ãã®2ã4ã«ã¤ãã¦ã¯Redshiftç¹æã®è¨å®ã¨ãªãã¾ããå½ã¨ã³ããªã§ã¯ãã®ä¸ã®ããã¼ã¿ã®åãã¨ãåå§ç¸®ã¿ã¤ããã«ã¤ãã¦ããã®æ¦è¦ã¨ãå®è·µã§ä½¿ãããã/調ã¹ãããæ§ã«è«¸ã å人ã§æ´çããæ å ±ãªã©ãä½µãã¦æä¸ãããã¨æãã¾ãã 1.ãã¼ã¿ã®å(data types) 2.åå§ç¸®ã¿ã¤ã(column compression types) 3.åæ£ãã¼(distkey) 4.ã½ã¼ããã¼(sortkey) 5.å¶ç´(constraint) ç®æ¬¡ ãã¼ã¿å æ°å¤å æåå æ¥ä»å ãã¼ã«å åå§ç¸®ã¿ã¤ã raw bytedict delta delta32k lzo mostly8 mostly16 mostly32 runlength t
5. æ´æ°åå§ç¸®ã®ä½¿ãéæ´æ°åå§ç¸®ã®ä½¿ãé ã»ãæ¤ç´¢ã¨ã³ã¸ã³ - 転置ã¤ã³ããã¯ã¹ ( Inverted Index ) ã»ãæ¨è¦ã¨ã³ã¸ã³ - ãã¼ã¿æ§é ãã®ãã® userID : ItemID1, ItemID 2 ãã»ã»ã» word : docID1, docID2 ãã»ã»ã» userID : user1, user2 ãã»ã»ã» å§ç¸®å¯¾è±¡
ããã«ã¡ã¯ãã¤ã³ãã©ã¹ãã©ã¯ãã£ã¼é¨ SRE ã°ã«ã¼ãã®åå· ( @rrreeeyyy ) ã§ããä»æãªã¹ã¹ã¡ã®ã¢ãã¡ã¯ãã¤ã³ã¨ã³ã¸ã§ã« BREAK ã§ãã æ®æ®µã®æ¥å並ã³ã«è¶£å³ã®ä¸ç°ã¨ãã¦ããµã¼ãã®ã¢ãã¿ãªã³ã°ç°å¢ã®èª¿æ»ãæ¹åã«åãçµãã§ãã¾ãã ããã§æ¬ç¨¿ã§ã¯ãã¢ãã¿ãªã³ã°ã®ã³ã³ãã¼ãã³ãã®ä¸ã¤ã¨ãã¦å¤ããã¨ãåºæ¥ãªããæç³»åãã¼ã¿ãã¼ã¹ã®åºç¤ç¥èã«é¢ãã¦ç´¹ä»ãã¾ãã ããããæç³»åãã¼ã¿ã»æç³»åãã¼ã¿ãã¼ã¹ã¨ã¯ï¼ æç³»åãã¼ã¿ã¨ããã®ã¯ãç¹å®ã®æéãã¨ã«ä½ããã®å¤ãåå¾ããéã®ãåå¾ããä¸é£ã®å¤ãæãã¾ãã ä¾ãã°ã以ä¸ã®ãããªãã©ã¼ãããããããã¼ã¿ãªã©ã¯æç³»åãã¼ã¿ã«ãããã§ãããã timestamp1,key,value1 timestamp2,key,value2 timestamp3,key,value3 : æç³»åãã¼ã¿ãã¼ã¹ã¨ã¯ãä¸è¨ã®ãããªæç³»åãã¼ã¿ã®ä¿åã»å¦çã«
è¨èªå¦çå¦ä¼ç¬¬ï¼ï¼å年次大ä¼ï¼2014/3ï¼ã®ãã¥ã¼ããªã¢ã«è¬ç¾©è³æã§ãã - è¦æ¨ - ææ³å§ç¸®ã¨ã¯ï¼å ¥åããã¹ããããã³ã³ãã¯ããªæèèªç±ææ³ï¼CFGï¼ã«å¤æããå§ç¸®æ³ã®ç·ç§°ã§ããï¼ ææ³å§ç¸®ã®å¼·ã¿ã¯å§ç¸®ããã¹ããå±éãããã¨ç¡ãï¼æ¤ç´¢çã®ããã¹ãå¦çãå¹çããè¡ããç¹ã«ããï¼ é©ãã¹ããã¨ã«ãã®å¦çé度ã¯ï¼å ããã¹ãä¸ã§ã®åãå¦çãçè«çã«ï¼æã«ã¯å®éã«ãåé§ããï¼ ã¾ãè¿å¹´ï¼ã¦ã§ãã¢ã¼ã«ã¤ãããã°ï¼ã²ãã é åçã®å¤§è¦æ¨¡å®ãã¼ã¿ãé«å¹çã«å§ç¸®ã§ãããã¨ã§æ³¨ç®ãéãã¦ããï¼ ããããªããï¼ææ³å§ç¸®ã«ã¤ãã¦ã®åå¦è åãã®è§£èª¬è³æã¯ã¾ã ã¾ã å°ãªãï¼ ããã§æ¬ãã¥ã¼ããªã¢ã«ã§ã¯ï¼ææ³å§ç¸®ã®æ´å²çèæ¯ããææ°ååã¾ã§ãå¹ åºãç´¹ä»ããï¼ å ·ä½çã«ã¯ææ³å¤æã¢ã«ã´ãªãºã ï¼å§ç¸®ããã¹ãä¸ã§ã®æååãã¿ã¼ã³æ¤ç´¢ï¼ææ³å§ç¸®ã«åºã¥ãçã¡ã¢ãªãã¼ã¿æ§é çã®è§£èª¬ãè¡ãï¼Read less
ä»æ´ãªããï¼Googleã®Maglevè«æã§ææ¡ããã¦ããMaglev Hashingãæå ã§å®è£ ãã¦ã¿ãï¼ Maglev: A Fast and Reliable Software Network Load Balancer Maglev Hashingã¨ã¯ æè¬Consitent Hashã®ä¸ç¨®ï¼Maglevãã¼ããã©ã³ãµã«ããããªã¢ã«ãµã¼ãé¸æã«ä½¿ç¨ããã¦ããï¼ ä¸è¨è«æã®Section 3.4ã§è©³ç´°ã説æããã¦ããï¼NSDI'16ã§ã®çºè¡¨ã¹ã©ã¤ããä½µãã¦çºããã¨åãããããï¼ Maglev: A Fast and Reliable Software Network Load Balancer | USENIX Slide: https://www.usenix.org/sites/default/files/conference/protected-files/nsdi16_sli
A BK-tree is a tree data structure specialized to index data in a metric space. A metric space is essentially a set of objects which we equip with a distance function $d(a, b)$ for every pair of elements $(a, b)$. This distance function must satisfy a set of axioms in order to ensure itâs well-behaved. The exact reason why this is required will be explained in the âSearchâ paragraph below. The BK-
ãªãªã¼ã¹ãé害æ å ±ãªã©ã®ãµã¼ãã¹ã®ãç¥ãã
ææ°ã®äººæ°ã¨ã³ããªã¼ã®é ä¿¡
å¦çãå®è¡ä¸ã§ã
j次ã®ããã¯ãã¼ã¯
kåã®ããã¯ãã¼ã¯
lãã¨ã§èªã
eã³ã¡ã³ãä¸è¦§ãéã
oãã¼ã¸ãéã
{{#tags}}- {{label}}
{{/tags}}