Browse images created by the Artbreeder community.
2021.05.18 ãã¯ããã¸ã¼ãæ´»ç¨ãã横æµDeNAãã¤ã¹ã¿ã¼ãºé¸æã®å¼·åæ½ç #ã³ã³ãã¥ã¼ã¿ã¼ãã¸ã§ã³ #ãã¼ã¿ãµã¤ã¨ã³ã¹ æ±ãã¦ããèª²é¡ éçã®ãã¼ã¿ã¯å¤å²ã«æ¸¡ãåæããé£ããã åæãæ´»ç¨ãããã¨ã¯ãããã«é£ããã å®ç¾ããçµæ æå±ãå°éæ§ã®ç°ãªãå¤æ§ãªã¡ã³ãã¼ãä¸ä¸¸ã¨ãªã£ã¦å®ç¨çãªã½ãªã¥ã¼ã·ã§ã³ãéçºã è¿å¹´ãã¹ãã¼ãåéã§ãããã¯ããã¸ã¼ãæ´»ç¨ãããã¼ã¿åæã大ããé²å±ãã¦ãã¾ãã横æµDeNAãã¤ã¹ã¿ã¼ãºå ã§ããå¾¹åºãããã¼ã¿åæããã¼ã å¼·åããè¡ã£ã¦ãã¾ãã æè¡çãªå°éæ§ã ãã§ã¯é£ãããéçãã¼ã¿ã®æ´»ç¨ 試åçµæã®è©³ç´°ãã¼ã¿ããã©ãã¯ãã³ï¼é«æ§è½å¼¾é測å®å¨ï¼ãªã©ã®ã»ã³ãµã¼ãã¼ã¿ã試åæ åãã¼ã¿ãªã©ããã¼ã¿ã®ç¨®é¡ã¨éãæ¥æ¿ã«å¢ããã¼ã¿åæã¯å¹´ã è¤éåãã¦ãã¾ãã æ¨ä»ãä»ã¾ã§ã®ä»®èª¬æ¤è¨¼ã«åºã¥ããã¼ã¿åæãããæ©æ¢°å¦ç¿ã«ããäºæ¸¬ã¢ãã«ãé«åº¦ãªæ å解æçã®AIæè¡ãæ´»ç¨ããã
2019/09/30ã«DeNAããã®ãªãã£ã¹ããåããã¦Data Pipeline Casual Talk Vol.4ãéå¬ãã¾ããã dpct.connpass.com DeNAããã®ç¨æãã¦ãã ãã£ããã´ãã¼ã¯ã®:Dãæããããªã ã©ã¤ã¹å¼å½ããã¤ã¹ã¿ã¼ãºBeerãé常ã«ããã£ãã§ãããããã¨ããããã¾ããã :Dãªã ã©ã¤ã¹å¼å½ ãã¤ã¹ã¿ã¼ãºBeer ä»åãããã°æ ãå ¬åããã®ã§ãããæ®å¿µãªããã¨ã³ããªã¯ãªãã£ãã®ã§ã主å¬ã®å¤ãã¨ãã¦ãå½æ¥ã®ã¬ãã¼ããæ¸ãã¾ãã ãªã¼ããã³ã°ãã¼ã¯è³æ ãã¤ãè¡ãªã£ã¦ãããªã¼ããã³ã°ã¹ã©ã¤ãã§ããä»åã¯ãã¼ã¿ãµã¤ã¨ã³ãã£ã¹ãå±æ§ãæ¸ã£ã¦ãåºç¤ããã¼ã¿ã¨ã³ã¸ãã¢ãå¤ãç³ãè¾¼ãã§ããã¾ããã ã¾ããä»åã®ä¼ã«æå¾ ãããã¨ã¨ãã¦ãå ·ä½çãªãã¤ãã©ã¤ã³ã®ãã¼ã«ã¨ãããããå ¨ä½çãªé¨åãæå¾ ãã¦ããæ§åã伺ãã¾ããã digdagä¸å¿ã®çæ´» æåã¯DeNAã®å°å£ã
IT Leaders ããã ï¼ ãã¯ããã¸ã¼ä¸è¦§ ï¼ AI ï¼ å¸å ´åå ï¼ ãAIéèªå交æ¸ãã©ãããã©ã¼ã ãã®ãã¹ãããããIICãæ¿èªâNECãæ²é»æ°ãªã©ãææ¡ AI AIè¨äºä¸è¦§ã¸ [å¸å ´åå] ãAIéèªå交æ¸ãã©ãããã©ã¼ã ãã®ãã¹ãããããIICãæ¿èªâNECãæ²é»æ°ãªã©ãææ¡ 2019å¹´8æ22æ¥(æ¨)IT Leadersç·¨éé¨ ãªã¹ã NECãç¬ãã©ã¦ã³ãã¼ãã¡ã¼ï¼Fraunhofer IOSBï¼ãã«ãã¯ãéå½ã®KETIï¼Korea Electronics Technology Instituteï¼ãæ²é»æ°å·¥æ¥ãè±ç°éåããã³ç£æ¥æè¡ç·åç 究æï¼ç£ç·ç ï¼ã¯2019å¹´8æ21æ¥ãAIéèªå交æ¸ãã©ãããã©ã¼ã ï¼Negotiation Automation Platformï¼ããã¤ã³ãã¹ããªã¢ã«ã»ã¤ã³ã¿ã¼ãããã»ã³ã³ã½ã¼ã·ã¢ã ï¼IICï¼ããæ¿èªãããã¨çºè¡¨ãããAIéã®èªå
XGBoostã®åãã«æè¿æ°ãã¤ããã®ã§ãããã¡ãã£ã¨è©³ããç¥ãããã¨æã£ã¦ä»¥ä¸ã®è«æãèªã¿ã¾ããã XGBoost: A Scalable Tree Boosting System ãã£ãããªã®ã§ãç°¡åã«ã¾ã¨ãã¦ã¿ããã¨æãã¾ããããã¨æã£ã¦ãããçµæ§ãªéã«ãªã£ã¦ããã¾ããã ä½ãééãçãããã¾ãããã³ã¡ã³ãããã¦ããã ããã@kefism ã¸é£çµ¡ããã¦ãã ããã¨å¬ããã§ãã XGBoostã¨ã¯ åºæ¬çã«å é¨ã§è¡ããã¦ãããã¨ã¯æ±ºå®æ¨ãè¤æ°åä½ãã¨ãããã¨ã§ãããããããã®ä½ãæ¹ã«ç¹å¾´ãããã¾ãã ããã§è¨å·ãå®ç¾©ãã¦ããã¾ãããã 説æå¤æ°ã次å ã¨ãã¦ãç®çå¤æ°ããã¨ãã¾ããããã§ã¯ãã¼ã¿ã®æ°ã§ããã¾ããã¼ã¿ã«å¯¾ããäºæ¸¬å¤ãã¨ãã¾ãã ã¾ã決å®æ¨ãï¼ã¤æ§ç¯ãã¾ããããã¨ãã®æ±ºå®æ¨ã使ç¨ãã¦äºæ¸¬ãã§ããããã«ãªãã¾ããï¼ã¤ç®ã®æ±ºå®æ¨ããå¾ããããã¼ã¿ã«å¯¾ããäºæ¸¬å¤ãã¨ãã¾ãããããã®ã¨
人工ç¥è½ï¼AIï¼æ´»ç¨ãé²ããå¤ãã®ä¼æ¥ããAIã®ããã©ãã¯ããã¯ã¹ãåã«é ãæ©ã¾ãã¦ãããã深層å¦ç¿ãã使ãã¨é«åº¦ãªåæãã§ããä¸æ¹ã§ãAIã®å¤ææ ¹æ ãä¸éæã«ãªãããã人å½ãªã©ãå·¦å³ããåéã§ã¯ä½¿ãã¥ãããããã§æ³¨ç®ãããã®ããXAIï¼=説æå¯è½ãªAIï¼ãã ãNECãå¯å£«éã ãã§ãªããä¸çã®ITä¼æ¥ãæè¡éçºãå éããããç¾æç¹ã§AIãå ¨é¢çã«å°å ¥ããã®ã¯é£ãããããã大æåå¦ã¡ã¼ã«ã¼ã®äººäº
R&Dãã¼ã ã®ä¸æå¥å¤ªéã§ãããããã®è¨äºã«ç¶ã2度ç®ã®ç»å ´ã«ãªãã¾ããããã¦ã5æ30æ¥ã§ã¢ã©ãµã¼ãè¿ãã¾ããã ä»åã¯æ·±å±¤å¦ç¿ã«ããã¦èã§ããããæéãããããã¼ã¿ã»ããä½æãèªååãããã¨ãè¨äºã«ãã¦ã¿ã¾ããã èæ¯ ãã¼ã¿ã½ã¼ã¹ ãã¼ã¿ã»ãã å¦ç¿ æ¨è« ãããã« èæ¯ ã³ã³ãã¥ã¼ã¿ã¼ã°ã©ãã£ãã¯ã¹(CG)ã§ä½æãããã®ãã¼ã¿ã»ããã®ç 究ã«é¢ãã¦ã¯NVIDIA Researchãããç»å ´ãããã¨ããããä¸ä¾ã¨ãã¦ä»¥ä¸ã®è«æãåºããã¦ãã¾ãã arxiv.org ããã§ã¯ç¾å®ä¸çã®ç»åã使ããã¨ãªãå¦ç¿ãããçµæãFaster R-CNNã§ç©ä½æ¤åºãã§ããã¨ããå 容ã触ãããã¦ãã¾ãã å¦ç¿ã®æ¹æ³ã¨ãã¦æ師ãªãå¦ç¿ãå¼±æ師ããå¦ç¿ãæ師ããå¦ç¿ãªã©æ§ã ãªææ³ãææ¡ããã¦ãã¾ããããããã«å¼·ã¿å¼±ã¿ã¯ããã©ãä»ã®ã¨ããæ©ãå¦ç¿ãåæãã精度çã«ãè¯ãçµæãåºãã¦ãããã®ã¯æ師ããå¦ç¿ã§ã
â»ããã¯ã¬ãã¼ãè¨äºã«ããåã®å人çãªãã¼ãããã®ã¾ã¾å ¬éãããã®ã§ãã誤å誤æ¤ãåéããããå¯è½æ§ãããã®ã§ãäºæ¿ãã ããã NLP Innovationï¼èªç¶è¨èªå¦çã®ã¤ããã¼ã·ã§ã³ï¼ Dr. Ming Zhouï¼å¨æï¼ããã³ã»ã¸ã§ã¦æ°ï¼ Ming Zhou at Microsoft Research ãã¤ã¯ãã½ãããªãµã¼ãã¢ã¸ã¢ï¼MSRAï¼å¯æé·ãACL ï¼ã³ã³ãã¥ã¼ã¿è¨èªï¼èªç¶è¨èªå¦çç 究ã®å½éå¦ä¼ï¼ä¼é·ãä¸å½è¨ç®æºå¦ä¼ï¼CCFï¼ä¸å½æ å ±æè¡å§å¡ä¼å§å¡é·ãä¸å½æ å ±å¦çå¦ä¼çäºã 1989å¹´ãä¸å½åã®CEMT-I ä¸è±æ©æ¢°ç¿»è¨³ã·ã¹ãã éçºã«ã¤ãã¦ä¸å½æ¿åºå 端ç§å¦æè¡è³ã 1998å¹´ãèåãªä¸æ¥æ©æ¢°ç¿»è¨³ã½ããã¦ã§ã¢ J-Beijing ãæ¥æ¬ã«ã¦éçºã MSRA ã«ã¦BingãOfficeãWindowsãAzure çã§ç¨ãããã¦ããèªç¶è¨èªå¦çæè¡ãéçºãããã£ãããããã§ããã·ã£
ä¸ççã«éç±ããAIéçºç«¶äºããã®ãªãã§æ¥æ¬ã®AIéçºã¯å¨åé ãã«ãªã£ã¦ããã¨åº¦ã ææããã¦ãã¾ãããæ¬é£è¼ã¯ãå²ç¢AIã®ç 究éçºãè¡ãç¦åæºæ°ã®èæ¸ããã¯ããã¸ã¼ã»ãã¡ã¼ã¹ããï¼ææ¥æ°èåºçï¼ããä¸é¨ãæç²ããçè ãå²ç¢AIéçºãç¶ãããªãã§æããå½å AIæ¥çãæ±ããåé¡ç¹ã«ã¤ãã¦ç´¹ä»ãã¦ããã¾ããä»åã¯å²ç¢ã¨é¢ä¿ã®æ·±ãæ¥æ¬ããå²ç¢AIç 究ã§ç±³å½ã«é ãã«ã¨ã£ãçç±ã«ã¤ãã¦èãã¦ããã¾ãã æ¥æ¬ã®AIç 究è ã«ãããä¸ä»£ãã極端ã«å°ãªãçç± æ¨è«AIãã¤ãããã¨ãã第äºä¸ä»£ã³ã³ãã¥ã¼ã¿ã§è¡ãããå²ç¢ã®ç 究ã¯ãããã¾ã§ãµãããã¸ã§ã¯ãã®æ±ãã«ãããªãã£ããå²ç¢ãããã¸ã§ã¯ãã®ä¸å¿ã«ãªã£ã¦ããããéãã¨ã³ãã£ã³ã°ã«ãã©ãã¤ããã®ã§ã¯ãªãããããæãã¨æ¯ãããã ã²ã¼ã ã«åã¤ã¨ããã¢ããã¼ãã§AIãç 究ããæ´å²ã¯å¤ãã1950å¹´ã«ã¯ããã§ã«ã³ã³ãã¥ã¼ã¿ã»ãã§ã¹ã®è«æãç»å ´ãã¦ãããããã¦ãæããA
1ï¼æ¥æ¬èä½æ¨©æ³47æ¡ã®7ãé©ç¨ãããã¨èä½ç©ã®ç¡æå©ç¨ãé©æ³ã«ãªãã¨ãããã¨ã ããå¤å½ã®ãµã¼ãä¸ã§å¦ç¿ä½æ¥ãè¡ã£ã¦ããã®47æ¡ã®7ã¯é©ç¨ãããã®ã ããã¯ãèä½ç©ã®å©ç¨è¡çºã®æºæ æ³ã®åé¡ï¼ããå©ç¨è¡çºã«ã©ãã®å½ã®æ³å¾ãé©ç¨ããããã®åé¡ï¼ã§ãããèä½æ¨©æ³ã«ã¤ãã¦ã¯ãèä½ç©ã®ãå©ç¨è¡çºå°ãã«ãããæ³å¾ãé©ç¨ãããã¨ããã¦ãã¾ãã ãã ãç¹ã«ããããå©ç¨ããå©ç¨è¡çºã®å ´åãã©ãããå©ç¨è¡çºå°ãã§ãããã®è§£éã¯é£ããåé¡ã§ãã 1ã¤ã®èãæ¹ã¨ãã¦ã¯ãããµã¼ãã®æå¨å°ããå©ç¨è¡çºå°ã§ããã¨ãããã®ã§ãããAå½ã«ãã人ãBå½ã«æå¨ãã¦ãããµã¼ããå©ç¨ãã¦å¦ç¿è¡çºãè¡ã£ãå ´åã«Aå½ã®èä½æ¨©æ³ãé©ç¨ãããã®ããBå½ã®èä½æ¨©æ³ãé©ç¨ããããã¯é£ããåé¡ã§ãã ãã ããæ¥æ¬å½å ã«ãµã¼ãããããæ¥æ¬å½å ã«ãã人ãåãµã¼ããå©ç¨ãã¦ãã¼ã¿ã®ãã¦ã³ãã¼ããã©ãã«ä»ããå¦ç¿è¡çºãè¡ããã±ã¼ã¹ã§ããã°ãæ¥æ¬èä½æ¨©æ³
AIã»äººå·¥ç¥è½EXPOã®ææ³ æ¥å ´è ãå¤ã ã¹ã¼ã姿ãå¤ã ããè¦ãã¨ãã£ããããããªã å社ã®å±ç¤ºå 容ãã³ãããçãã»ã©ä¼¼ã¦ãã Chatbotãæ¸ã£ã¦GUIã®AIéçºãã¼ã«ãå¢ãã GRID社ã®ãã¼ã¹ãã§ãããã SIãã³ãã®ç¡çããAIã«å¯ãã¾ããæãã¹ã´ã¤ 人æè²æã¨ãªãã·ã§ã¢ããã¬ã³ãã£ã½ã â ãã¹ã¯ãã»ã¢ãã©ã¤ãº (@maskedanl) April 5, 2019 ãããªä¸»è¦³ã¾ã¿ããªææ³ãããã伸ã³ã¦ãã¾ã£ãã®ã§ãã¡ãã£ã¨çé¢ç®ã«AIã»äººå·¥ç¥è½EXPOã®ææ³ãæ¸ãã¦ã¿ãã â å ¨ä½ã®å°è±¡æ¥å ´è ã¯ã¹ã¼ã姿ãå¤ããå¹´é½¢ã¯30-50代ã¨ãã£ãã¨ããã6-70代ã¨æãã人ããããããã±é²æ¢AIãæ¢ãã¦ããã®ããèå¾ã®è³éãAIã«æè³ããã®ãããã ã®å±ç¤ºä¼ããã¢ãªã®ãã¯ããããªãã æ¥å ´è ã¯æ¥ç¨®ã®ã«ã¼ããä¸ãã¦ãããã製é ããéä¿¡ã»ITãã6å²ç¨åº¦ãæ®ãã¯éèã»ç¤¾ä¼ã¤ã³ãã©ã»å»çã»å°é
Keras FAQ: Kerasã«é¢ãããããã質å Kerasãå¼ç¨ããã«ã¯ï¼ KerasãGPUã§åããã«ã¯ï¼ Kerasããã«ãGPUã§åããã«ã¯ï¼ "sample","batch"ï¼"epoch" ã®æå³ã¯ï¼ Keras modelãä¿åããã«ã¯ï¼ training lossãtesting lossãããã¯ããã«å¤§ããã®ã¯ãªãï¼ ä¸éã¬ã¤ã¤ã¼ã®åºåãå¾ãã«ã¯ï¼ ã¡ã¢ãªã«è¼ããªã大ããã®ãã¼ã¿ãæ±ãã«ã¯ï¼ validation lossãæ¸ããªããªã£ãã¨ãã«å¦ç¿ãä¸æããã«ã¯ï¼ validation splitã¯ã©ã®ããã«å®è¡ããã¾ããï¼ è¨ç·´æã«ãã¼ã¿ã¯ã·ã£ããã«ããã¾ããï¼ åepochã®training/validationã®lossãaccuracyãè¨é²ããã«ã¯ï¼ ã¬ã¤ã¤ã¼ã "freeze" ããã«ã¯ï¼ stateful RNNãå©ç¨ããã«ã¯ï¼ Sequentialã¢ã
ã¢ã©ã¤ï¼ãã£ã¼ãã©ã¼ãã³ã°ã®æ¼ç®éåæ¸æè¡ãä¿æ ãã£ã¼ãã©ã¼ãã³ã°ãæ©æ¢°å¦ç¿ã§äºæ¸¬ã¢ãã«ã®æ§ç¯ãããã°ãã¼ã¿ã®è§£æãè¡ã£ã¦ããä¼æ¥ã ãã¼ã¿åæã»äºæ¸¬ã«åºã¥ãæé©ãªã¢ã¯ã·ã§ã³ãææ¡ããã½ãªã¥ã¼ã·ã§ã³ããã£ã¼ãã©ã¼ãã³ã°ãæ´»ç¨ãã¦ç»åã»ã»ã³ãµã¼ããå¾ããããã¼ã¿ã§äººãç©ãèå¥ããã½ãªã¥ã¼ã·ã§ã³ããã£ã¼ãã©ã¼ãã³ã°å°ååã½ãªã¥ã¼ã·ã§ã³ãªã©ããã®ãµã¼ãã¹ã¨ãã¦æããããã ã¢ã©ã¤ãéçºãããã£ã¼ãã©ã¼ãã³ã°ãå°ååããããã®ãã£ã¼ãã©ã¼ãã³ã°æ¼ç®éåæ¸æè¡ã¯ãMicrosoft Innovation Award 2018ãã®æåªç§è³ãåè³ãã¦ããã https://www.araya.org/ ã¨ã¤ã·ã³ã°ï¼æ©æ¢°å¶å¾¡ã«ç¹åããAIã¢ã«ã´ãªãºã ãæ¡ç¨ æ©æ¢°ã¸ã®çµã¿è¾¼ã¿ãåæã¨ããAIã¢ã«ã´ãªãºã ãéçºã»æä¾ãã¦ããä¼æ¥ãã³ã¢æè¡ãDBTï¼Deep Binary Treeï¼ãã¯å°å¤éãã¼ã¿ãé«éã»é«
ãªãã¨ï¼AI(人工ç¥è½)ãæãã裸婦ç»ãä¸ççãªè¸è¡è³ã§ã°ã©ã³ããªãç²å¾ãå°è±¡çãªè²ä½¿ãã¨ç使ãã«é«ãè©ä¾¡ã è¨äºã®æ¬æã«ã¹ããã ã¤ã®ãªã¹ã»ãã³ãã³ã§æè³å¼ã®è¡ããã第7åã«ã¼ã¡ã³è³(Lumen Prize)ã§ãAI(人工ç¥è½)ã®æãã裸婦ç»ãéè³ã«è¼ããããã ã ä¸ççãªè¸è¡è³ã§AIã«ããä½åãã°ã©ã³ããªãç²å¾ããã®ã¯åãã¦ã®ãã¨ãªã®ã ã¨ãã ææããã®ã¯å½éçã«è©ä¾¡ã®é«ããã¸ã¿ã«ã¢ã¼ãã£ã¹ããããªãªã»ã¯ãªã³ã²ãã³ããã§ãä½ååã¯ãã¶ã»ãããã£ã¼ãºã»ãµã³(The Butcherâs Son)ãã¨ããã ããªãªããã¯ãã®ä½åãã人éã¨ããåå¨ããã¥ã¼ã©ã«ãããã¯ã¼ã¯(ç¥çµå路網)ã¨ãã¦è§£éãããã®ãã ã¨èª¬æãã¦ããã æ£ç¶ã®å³å½¢ãã段éãçµã¦è£¸å©¦ç»ã¸ ãã¶ã»ãããã£ã¼ãºã»ãµã³ãã¯ãGANs(æµå¯¾ççæãããã¯ã¼ã¯ / æ師ãªãæ©æ¢°å¦ç¿ã§ä½¿ç¨ããã人工ç¥è½ã¢ã«ã´ãªãºã ã®ä¸ç¨®)ã§ä¸é£ã®å¦ç¿ã
ã©ã³ãã³ã°
ã©ã³ãã³ã°
ã©ã³ãã³ã°
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}}