æ©æ¢°å¦ç¿
æè¿Webå°èª¬ã人æ°ã§ããããæ¸ç±åããããã¢ãã¡åããããã¦ãã¾ã ä»æã®ã¢ãã¡ã§ã¯ããã³ã¸ã§ã³ã«åºä¼ããæ±ããã®ã¯ééã£ã¦ããã ãããããæ¾éããã¦ãã¦ããããã§ã(ã¢ãã¡ã«åããã¦Kindleçã®1,2å·»ãå¤ä¸ãããã¦ãã¾ã)ãã³ã¸ã§ã³ã«åºä¼ããæ±â¦
276th, 141.78pts, +0/-0 challenge Volatility: 324->324SRM 652 SRM 652 - Togetterã¾ã¨ãä¹ ãã¶ãã«åå ããããéä¸ã§ãµã³ãã«ã¢ã¦ãããããå¤ãã£ãããããã¤ãã¢ãã¦ã³ã¹ãæµãããã¿ãã¿ããæãã ã£ã ååã«å¼ãç¶ãä»åãunratedã«ãªã£ã¦ãã¾ã£â¦
ä»å¹´ã¯ããããéå¬ãããã®ã«å ¨ç¶åå ã§ããªãã£ã ç®æ¨ã«ãkaggleã«åå ãããã¨ãæ¸ãã¦ãæ°ããããã ãã©â¦â¦ ç»åèªèç³»ã ã¨ã¾ã£ããæãåºãªãã®ããªãã¨ããããæ©æ¢°å¦ç¿ã®ã³ã³ãã¯ãè¨ç·´ãã¼ã¿ãä¸ãããã¦ããã§ä½ãããäºæ¸¬ã¢ãã«ãä½ã£ã¦äºæ¸¬çµæâ¦
Machine Learning: The High Interest Credit Card of Technical Debt NIPS 2014ã®Workshopã®SE4ML: Software Engineering for Machine Learningã¨ããã®ã§çºè¡¨ãããè«æã£ã½ãã§ã Twitterã§è©±é¡ã«ãªã£ã¦ããã®ã§èªãã ã¡ã¢ã§ãä¹ ãã¶ãã«è±èªè«æãèªãã â¦
æ¨æ¥ã®è¨äºã§å¾ãããæ±æ¹ãã£ã©å士ã®é¢é£æ§ã®å¼·ã(NPMI)ã使ã£ã¦ãIsomapã¨ããææ³ã§ãã£ã©ãäºæ¬¡å ä¸ã«é ç½®ãã¦å¯è¦åãã¾ã æ±æ¹ãã£ã©ã®é¢é£æ§ã®å¼·ãããã³ãã³åç»ã®åç»æ°ã§æ¸¬ã£ã¦ã¿ã - å¯ç©æ¯ç @Scaled_Wurm æ±æ¹ãã£ã©ã®é¢é£æ§ã®å¼·ãããã³ãã³â¦
åã«æ±ºå®æ¨ã®å¯è¦åããããã¨æã£ã¦ãã£ã¦ãªãã£ãã®ã§ãã£ã¦ããã¾ã決å®æ¨ã®ã©ã¤ãã©ãªã¯ä¾ã®ãã¨ãscikit-learnã使ã pythonã®æ©æ¢°å¦ç¿ã©ã¤ãã©ãªscikit-learnã®ç´¹ä» - å¯ç©æ¯ç @Scaled_Wurm pythonã®æ©æ¢°å¦ç¿ã©ã¤ãã©ãªscikit-learnã®ç´¹ä» - å¯ç©æ¯çâ¦
42th, 725.99pts, +1/-0 challenge Volatility: 451->450Div1 復帰ãããä¹ ãã¶ãã«challengeæå Hardã¯ãã¾ãèªä¿¡ãªãã®ãæãããäºæ³éãSystem Testã§è½ã¡ã 250: RunningAroundPark æ¨ã«1ããNã®çªå·ãä»ãã¦ãã ã©ã³ãã³ã°ä¸ã«è¦ãæ¨ã®çªå·ãéä¸æâ¦
AIZU ONLINE JUDGE(AOJ)ã¨ãã競æããã°ã©ãã³ã°ã®åé¡ãå ¬éãã¦ãããµã¤ããããã AIZU ONLINE JUDGE: Programming Challenge 1年以ä¸åã«AOJãå°ãã ããã£ã¦ããé ã«ãAOJã®åé¡ãæ¨è¦ããã¹ã¯ãªãããæ¸ãããã®ã®ãç¹ã«å ¬éã使ç¨ãããã«æ¾ç½®ãã¦ãâ¦
åã«LIBSVMã®éã¿ã®è¦æ¹ãæ¸ããã LIBSVMã®ã¢ãã«ãã¡ã¤ã«ã«ã¯ãµãã¼ããã¯ã¿ã¼ã®éã¿(å対åé¡ã®éã¿)ãæ¸ãã¦ããã ãã§åç¹å¾´éã®éã¿(主åé¡ã®éã¿)ãæ±ããããã«ã¯è¨ç®ãããå¿ è¦ããã£ãã LIBLINEARã®æ¹ã¯ããç°¡åã§ã¢ãã«ãã¡ã¤ã«ã«åç¹å¾´éã®éã¿â¦
ãã©ã½ã³ãããã¨ããã®ã¯10æ¥éãããã®æéã§åé¡ãåæãã³ã¼ããæ¸ãã¦ã¹ã³ã¢ã競ã競æã§ãã ä»åã®ã§3åãã®ãã©ã½ã³ãããåå ã åºç¤ã¯ä¸ä½ã«ããããã®ã§ãããçµç¤ã¯å¤±éãã¦ãã¾ãã¾ããorzé ä½è¡¨ http://community.topcoder.com/longcontest/stâ¦
https://crowdsolving.jp/node/629/summary 1ä½ãã5ä½ã«ãªã£ã人ã®ææ³ãå ¬éããã¦ãã¾ã ãã¨1ä½ã ã£ãç§ã®ã¤ã³ã¿ãã¥ã¼ãè¼ã£ã¦ãã¿ããã§ã(æ¥ããããã®ã§ç¢ºèªãã¦ãªãã§ãã èªåã®ææ³ã®å¤§éæãªèª¬æ ä¸ã®è¨äºã«ç´°ããæ¸ãã¦ããã¨æããã§å¤§éæã«â¦
ã¯ã©ã½ã«(CrowdSolving)ã®ç¬¬3åã³ã³ãã«åå ä¸(éä¸çµé: 1/16ä½) - å¯ç©æ¯ç @Scaled_Wurm éä¸çµéã¯ä¸ã®è¨äºã¿ãããªæãã§ããããæçµçµæã1ä½ã§ãã ããã£ã¬ã³ã¸ã³ã³ããè¨äºéã®ãªã³ã¯æ¨å® | CrowdSolving ãã®ã³ã³ãã«éãã¦ããªã³ã¯äºæ¸¬ã«é¢ããâ¦
ã¾ã æéã®éä¸ã§ãã以ä¸ã®ã«åå ãã¦ã¾ã ããã£ã¬ã³ã¸ã³ã³ããè¨äºéã®ãªã³ã¯æ¨å® | ã¯ã©ã½ã« | CrowdSolving è¨äºã®éã«ãªã³ã¯ãããããäºæ¸¬ããã¿ã¹ã¯ã§ãä¸ä½ã«ãªãã¨ä¸å¿è³éãåºããããã§ãéå¦ç¿ã£ã½ãããã(ï¼)ãä»®é ä½è¡¨ä¸ã§ã¯1ä½ãªã®ã§å¬ããâ¦
å¤å¤é(å¤æ¬¡å )æ£è¦åå¸ã®KLãã¤ãã¼ã¸ã§ã³ã¹ã®æ±ãæ¹ - EchizenBlog-Zwei ä¸ã®è¨äºãèªãã§åå¼·ã«ãªã£ãã®ã§ãããæ°å¼ãããã¹ãã§èªã¿ã¥ããã£ãã®ã¨ãå¤å¤éã§ãªã1次å ã®æ£è¦åå¸ã®å°åºã®æ®µéã§ããããããªãã£ãã®ã§èª¿ã¹ã¦è¨äºã«ã¾ã¨ãã¾ãã 注æ æ°â¦
以åLIBSVMã§ç¹å¾´éã®éã¿ãè¦ãæ¹æ³ã«ã¤ãã¦è¨äºã§URLãç´¹ä»ããã®ã§ããããªã³ã¯å ã®è¨äºã¨ã³ã¼ãããªããªã£ã¦ããã¿ãããªã®ã§ãæ¹ãã¦è¨äºã«ãã¦ããã¾ãã SVMã§ã®ç¹å¾´éã®éã¿ é常ã«åç´åãã¦èª¬æããã¨ãç·å½¢ã«ã¼ãã«ã®SVMã¯æ¬¡ã®ãããªå¼ã®ç¬¦å·ã®æ£â¦
scikit-learn(sklearn)ã®æ¥æ¬èªã®å ¥éè¨äºãããã¾ããªããªã¼ã¨æã£ã¦æ¸ãã¾ããã ã©ã¡ããã£ã¦ããã¨ãã使ãæ©è½ã®ç´¹ä»çãªæãã§ãã è±èªãèªããæ¹ã¯å ¬å¼ã®ãã¥ã¼ããªã¢ã«ãããããã§ãã scikit-learnã¨ã¯ï¼ scikit-learnã¯ãªã¼ãã³ã½ã¼ã¹ã®æ©æ¢°å¦ç¿â¦
æ©æ¢°å¦ç¿ã®ãã¼ã¿ã¨ãã¦ç¹å¾´éãä½ãã¨ãã®æ³¨æç¹ãæ©ããã¨ãªã©ãã¡ã¢ã£ã¦ããã¾ãããééããªã©ãå«ã¾ãã¦ããããããã¾ããã åºæ¬çãªå 容ã§ãã®ã§èª¿ã¹ãã°ãã£ã¨é©åãªããæ¹ãããã¨æãã¾ãã ã«ãã´ãªã«ã«ã»ãã¼ã¿ ã«ãã´ãªã«ã«ã»ãã¼ã¿ã¨ããã®ã¯â¦
CrowdSolving | ãã¼ã¿åæã»äºæ¸¬ã¢ãã«ä½æã®ã³ã³ããµã¤ãã®ã³ã³ããçµãã£ãã®ã§ãçµæã¨ãã¢ããã¼ãã¨ãæ¸ãã¨ãã¾ã以åæ¸ããè¨äºâ CrowdSolvingã«åå ä¸ - ãããã¯æ©æ¢°å¦ç¿é¢é£ã®ã¡ã¢ - å¯ç©æ¯ç @Scaled_Wurm ã¿ã¹ã¯ ã³ã³ãã³ã販売ãµã¼ãã¹ã®ä¼å¡â¦
çè¡åã¨ã¯ çè¡åã¯æåã®ã»ã¨ãã©ãã¼ãã§ããè¡åã®ãã¨ã§ãã ãã¨ãã°ãææ¸ã«ç»å ´ããåèªã®é »åº¦ãæ°ãããããã¨ãããã£ãè¡åã«ãªãã¾ãã ä»ã«ãçãªã°ã©ãã®é£æ¥è¡åã¯çè¡åã«ãªãã¾ãã ãããã£ãè¡åã¯é常ã®è¡å(å¯è¡å)ã使ãããããå°ãªâ¦
æ©æ¢°å¦ç¿ã¨ã³ã¸ã§ã¤å¢ã¨ãã¦ãCrowdSolving | ãã¼ã¿åæã»äºæ¸¬ã¢ãã«ä½æã®ã³ã³ããµã¤ãã«åå ãã¦ãæã¨ãã«èª¿ã¹ããã¨ãªã©ã®ã¡ã¢ã ã¾ã çµãã£ã¦ãªãã§ããããããããããã¨ãæãã¤ããªãã®ã§ã¨ããããã¾ã¨ãã¨ãã¾ããã ä¸å¿ã¾ã ã¢ãã«ã¨ãç¹å¾´éã®â¦
åã¾ã§Rã使ã£ã¦ããã®ã§ããããã£ããPythonã使ã£ã¦ããã®ã ããå ¨é¨ç§»è¡ããããªã¼ã¨æã£ã¦è²ã åå¼·ãã¦ã¾ãã numpyã§csvãèªã¿è¾¼ã numpy.genfromtxté¢æ°ãnumpy.loadtxté¢æ°ã使ãã°ç°¡åã«èªã¿è¾¼ããã¿ããã§ãã import numpy as np data = np.genfrâ¦
LIBSVM -- A Library for Support Vector Machinesã¨ãLIBLINEAR -- A Library for Large Linear Classificationã使ã£ã¦äºå¤åé¡ã®åé¡ã§éãã§ãã¾ããã ãã®æã«ããã«ããããªãã£ããã¨ã®ã¡ã¢ã Accuracy以å¤ã®å°ºåº¦ãåºåãã æ®éLIBSVMãLIBLINEARãâ¦
æ©æ¢°å¦ç¿ãçµ±è¨ã§ã¯"cross-validation"ã¨ããææ³ã使ããã¦ãã¾ãã 詳細ã¯ä»¥ä¸ã®URLãªã©ãè¦ã¦ããããã¨ããã®ã§ããããã¼ã¿ãåå²ãä¸é¨()ãè©ä¾¡ç¨ããã®ä»()ãè¨ç·´ç¨ã¨ãã¦æ§è½ãè©ä¾¡ãããã¨ããã®ãè©ä¾¡ç¨ãè¨ç·´ç¨ããå ¥ãæ¿ãã¦éãã«ã¤ãã¦è¡ããâ¦
è«æææ³: "TCSST: Transfer Classification of Short & Sparse Text Using External Data"(CIKM 2012)
TCSST: Transfer Classification of Short & Sparse Text Using External Data æ¦è¦ Twitterãªã©ã®ãã¤ã¯ãããã°ãæ¤ç´¢ã¨ã³ã¸ã³ã®ã¹ãããããã¬ãã¥ã¼ãªã©ã®çãã¹ãã¼ã¹ãªæç« ã«å¯¾ãã¦ãå¤é¨ã®ãªã½ã¼ã¹ã使ã£ã¦è»¢ç§»å¦ç¿ãè¡ããã¨ã§ç²¾åº¦ã®åä¸ããããé¢â¦
"Adversarial Support Vector Machine Learning" æ¦è¦ æµå¯¾çå¦ç¿ã®è©±ã spamæ¤åºãä¾µå ¥æ¤åºãªã©ã®ç¸æããã§ãã¯ãããããããã¨ãã¦é常ã«è¦ãããããã¨ããå ´åã®æ©æ¢°å¦ç¿ã See: æµå¯¾çå¦ç¿ - æ©æ¢°å¦ç¿ã®ãæ±é·ºã®æWikiã å è¡ç 究ã§ã¯æªæã®ããæµå¯¾â¦
ãè¨èªå¦çã®ããã®æ©æ¢°å¦ç¿å ¥é (èªç¶è¨èªå¦çã·ãªã¼ãº)ã(ããããé«ææ¬)ã§ç 究室ã®å¾è¼©ãåå¼·ä¼ããã¦ãã¦ãèªåã§ãä¸ç¬ããããªãã£ãã®ã§ã¡ã¢ã PLSAã¨ãPLSIã¨ãã¢ã¹ãã¯ãã¢ãã«ã¨ãååããããããã£ã¦ããããããªãã§ãã åæ確ç ææ¸ããåâ¦
以åãµã¶ã¨ããã®ãããããäºæ¸¬åé¡ã®ãµã¼ãã¤ãããæã«ãµã¶ã¨ããã¨ããªãã¥ã¢ã®ã¸ã£ã³ã±ã³ãã¼ã¿ããã¦ã³ãã¼ãããã¹ã¯ãªãããæ¸ãããã¦æ¾ç½®ãã¦ããã®ãè¦ã¤ããã®ã§ãå®æããã¦å ¬éãã¾ãããã¥ã¢ãã¼ã¹ã®ãã¼ã¿ã¯30å¼±ããããããªãã®ã§ãäºæ¸¬ã¨â¦
Support Vector Machinesãç¨ãããéæ³å°å¥³ã¾ã©ãâãã®ã«ã人ç©äºæ¸¬ã¢ãã« - Yuta.Kikuchiã®æ¥è¨ âã®è¨äºãé¢ç½ãã£ãã®ã¨ãSVMã触ã£ããã¨ããªãã£ãã®ã§libsvmããããã£ã¦éãã ã¡ã¢ã ã¯ãã¹ããªãã¼ã·ã§ã³ libsvmã§ã®ã¯ãã¹ããªãã¼ã·ã§ã³ã¯"svm-traiâ¦
Multi-Domain Learning: When Do Domains Matter? æ¦è¦ ãã«ããã¡ã¤ã³ã©ã¼ãã³ã°(MDL)ã«ãããããã¤ãã®çåã«å¯¾ããæ¤è¨¼ï¼ MDLã«ãã精度ä¸æã¯å¿ ããããã¡ã¤ã³ã®å½±é¿ã§ã¯ãªãï¼ã¢ã³ãµã³ãã«ã®å¹æã§ã¯ãªããï¼ ãã¡ã¤ã³ã©ãã«ãã©ã³ãã ã«å ¥ãæ¿ãã¦ãâ¦
Polarity Inducing Latent Semantic Analysis æ¦è¦ ææ¸-åèªè¡åä¸ã§cosé¡ä¼¼åº¦ãªã©ã使ãã¨ï¼æå³çã«ä¼¼ãåèªå士ãè¦ã¤ãããã¨ãã§ããï¼ ããããã®å ´åã®ãæå³çã«ä¼¼ããã¨ããã®ã¯ãé¡ç¾©èªãããã¯åæèªãã§ããï¼ãåæèªãã ããè¦ã¤ãããå ´åã«â¦