JavaScript: Past, Present, and Future - NDC Porto 2020
JavaScript: Past, Present, and Future - NDC Porto 2020
GPT4ãç»å ´ãã¦ChatGPTãçãä¸ãã£ã¦ã¾ããã æ¬è¨äºã¯ãGPTï¼ãæ¯ããTransformerã¨ããä»çµã¿ï¼ãChatGPTã¦ã¼ã¶ã«ã¨ã£ã¦åããããã説æãããã®è½åã¨éçã«ã¤ãã¦ã®è¦éããããããããã¨ãã趣æ¨ã«ãªãã¾ãã å°ãé·ããªããããªã®ã§ãå ¨é¨ã§è¨äºã3åã«åãããã¨æãã¾ãã ï¼ï¼ï¼å¤§ã¾ããªèæ¯ã¨æ¦è¦ï¼æ¬è¨äº ï¼ï¼ï¼GPTã®è½åã¨å¯è½æ§ï¼å®éã®ä½¿ç¨ä¾ã¨Transformerã®ä»çµã¿ãè¸ã¾ãã¦èª¬æ ï¼ï¼ï¼GPTã®éçã¨æªæ¥å±æï¼Transformerã®ä»çµã¿ãæã¤éçã¨ç 究ã®é²å±ãäºæ³ GPT3ã¨4ã®éãï¼ ãã¼ã¯ã³é·ã¨ã¯ä½ãï¼ ã¾ããããã話ãå§ãã¾ããGPT-3ã¯ããã©ã¡ã¼ã¿æ°ã750ååï¼850GBã®å®¹éãé£ãï¼ã§ãã¼ã¯ã³é·ã4097(GPT-3.5)ã§ãããGPT-4ã¯ããã©ã¡ã¼ã¿æ°ã¯éå ¬éã§ãã¼ã¯ã³é·ã¯32768ã§ãã®ã§ãã¡ããã©8åã«ãªãã¾ãã ãã¦ããã¼ã¯
ãWorkshop OT 2023 æé©è¼¸éã¨ãã®å¨è¾º â æ©æ¢°å¦ç¿ããç±åå¦çæé©åã¾ã§ãã§ç¨ããã¹ã©ã¤ãã§ã
æ¦è¦ ãã®è¬ç¾©ã§ã¯ã深層å¦ç¿ã表ç¾å¦ç¿ã®ææ°æè¡ã«ã¤ãã¦ã主ã«æ師ããã»æ師ãªã深層å¦ç¿ãåãè¾¼ã¿ææ³ãè·é¢å¦ç¿ãç³ã¿è¾¼ã¿ã»å帰åãã¥ã¼ã©ã«ãããã¯ã¼ã¯ããããããç¨ããã³ã³ãã¥ã¼ã¿ãã¸ã§ã³ãèªç¶è¨èªå¦çãé³å£°èªèãªã©ã®å¿ç¨æè¡ãåãæ±ãã¾ãã ãã®è¬ç¾©ã¯ãDS-GA 1001 Intro to Data Scienceããããã¯å¤§å¦é¢ã¬ãã«ã®æ©æ¢°å¦ç¿ç§ç®ããã§ã«å±¥ä¿®ãã¦ãããã¨ãåæã¨ãã¾ãã è¬ç¾© å¡ä¾: ð¥ ã¹ã©ã¤ã, ð Jupyter notebook, ð¥ YouTubeãããª. é± å½¢å¼ ã¿ã¤ãã« è³æ
Deep Learning for ENGINEERï¼Eè³æ ¼ï¼ã¨ã¯ ãã®è¨äºã¯2020å¹´1ææç¹ã®æ å ±ã§ãããã¨ããäºæ¿ãã ããã ç¾å¨ããã®æ¥æ¬ã«ããã¦AIã®ã¹ãã«ã証æããå ¬çè³æ ¼ã¨ãã¦ä¸è¨ãæãããã¾ãã JDLA Deep Learning for GENERALï¼Gæ¤å®ï¼ JDLA Deep Learning for ENGINEERï¼Eè³æ ¼ï¼ ã§ãã ä½ç½®ä»ãã¨ãã¦ã¯ãGæ¤å®ãAIé¢é£æè¡ã«é¢ããä¸è¬æé¤ãåã試é¨ã Eè³æ ¼ãAIé¢é£æè¡ã«é¢ããçè«ãç解ããAIãå®è£ ã§ããã¹ãã«ãåã試é¨ã ã¨ããç解ã§ãã 詳細ã¨ããããæ£ç¢ºãªå®ç¾©ã«ã¤ãã¦ã¯ãæ¥æ¬ãã£ã¼ãã©ã¼ãã³ã°åä¼ï¼JDLAï¼ã®ãµã¤ãããåç §ãã ããã Eè³æ ¼ã®åé¨æ¹æ³ã¨JDLAèªå®ããã°ã©ã ã¨ãããããã§ãããããã¯Eè³æ ¼ã®ãã¨ãä¸å¿ã«è©±ãé²ãã¦ããããã¨æãã¾ãã Eè³æ ¼ãåé¨ãããã£ã³ã¹ã¯å¹´ã«2åããã2æã¨8
div.hs-menu-wrapper > ul > li" data-pacnav-mobile-width="820"> AISIA AIã®å®ç¨å å¤è¦³æ¤æ»ã·ã¹ãã ããã° ã¯ããã« æ©æ¢°å¦ç¿ã«ã¯ãæ師ããå¦ç¿ãããæ師ãªãå¦ç¿ãããå¼·åå¦ç¿ãã¨ãã3ã¤ã®å¦ç¿æ¹æ³ãããã¾ããããã¦ããã®èå¾ã«ã¯ãå帰ãããåé¡ãããã¯ã©ã¹ã¿ãªã³ã°ããªã©ã®çµ±è¨å¦ãããã解ãæ±ããæ¹æ³ã¨ãã¦ã決å®æ¨ããããµãã¼ããã¯ã¿ã¼ãã·ã¼ã³ãããï½å¹³åæ³ããªã©å¤ãã®ã¢ã«ã´ãªãºã ãããã¾ãã ãå¦ç¿æ¹æ³ãã¨ãçµ±è¨å¦ãã¨ãã¢ã«ã´ãªãºã ãããã£ãããã®ä¸è§é¢ä¿ã¯ã©ããªã£ã¦ããã®ã§ãããããã¾ãããå¦ç¿æ¹æ³ãã¨ãçµ±è¨å¦ãã®é¢ä¿ããç´è§£ãã¦ã¿ã¾ãã æ©æ¢°å¦ç¿æ³ã¨çµ±è¨å¦ ã¾ãã¯å³1ãã覧ãã ããããæ師ããå¦ç¿ãããæ師ãªãå¦ç¿ãããå¼·åå¦ç¿ãã¨ãã3ã¤ã®å¦ç¿æ¹æ³ã¨ãå帰ããåé¡ããã¯ã©ã¹ã¿ãªã³ã°ãã¨ãã£ãçµ±è¨å¦ã®é¢ä¿ãããã¨å³
ãã¤ãããã¾ã§ãï¼ åã¯ãã¾ã深層å¦ç¿ã«é¢ãã¦è¨äºãæ¸ããã¨ã¯ãªãã®ã§ããï¼ã¡ãã£ã¨æ°ã«ãªã£ãè«æããã£ãã®ã§ç´¹ä»ãã¾ãï¼ [1711.00165] Deep Neural Networks as Gaussian Processes è«æã¯Google Brainã®ç 究è ãã«ãããã®ã§ï¼NIPS2017 Bayesian Deep Learning WorkshopICLR2018ã«acceptããã¦ãã¾ãï¼å®ã¯æ·±å±¤å¦ç¿ãã¬ã¦ã¹éç¨ï¼Gaussian processï¼ã§æ§ç¯ããã®ã¯ãã®è«æãååºã§ã¯ãªãã®ã§ããï¼è«æã§ã¯ãã¤ãºå¦ç¿ï¼æ·±å±¤å¦ç¿ï¼ã«ã¼ãã«æ³ãç°¡ç¥ãã¤å æ¬çã«èª¬æãã¦ããå 容ã«ãªã£ã¦ããã®ã§é常ã«åèã«ãªãã¾ãï¼ ãã¦ï¼ã深層å¦ç¿ã¯ã¬ã¦ã¹éç¨ãã¨ããã®ã¯ã¡ãã£ã´ã宣ä¼çãªã¿ã¤ãã«ã«ãéãã¦ãã¾ã£ãã®ã§ããï¼ããå°ãã ãæ£ç¢ºã«è«æã®è¦ç¹ãã¾ã¨ããã¨æ¬¡ã®ããã«ãªãã¾ãï¼ èæ¯ åä¸
ã¢ã¤ããã¼ãç¡æAIããã°ã©ãã³ã°å¦ç¿ãµã¼ãã¹ãAidemyãããªãªã¼ã¹ï¼æ·±å±¤å¦ç¿ãèªç¶è¨èªå¦çãæ°å¤äºæ¸¬ãªã©ãå¦ç¿å¯è½ ã¢ã¤ããã¼ã¯ããªã³ã©ã¤ã³ã§AIããã°ã©ãã³ã°ãå¦ç¿ã§ãããµã¼ãã¹ãAidemyããå ¬éãããPythonå ¥éã®ä»ã深層å¦ç¿ãèªç¶è¨èªå¦çãæ°å¤äºæ¸¬ãªã©å ¨11è¬åº§ãç¡æã§å©ç¨ã§ããã ã¢ã¤ããã¼ã¯2017å¹´12æ20æ¥ãæè¡è åãã®AIï¼äººå·¥ç¥è½ï¼ããã°ã©ãã³ã°ãç¡æã§å¦ã¹ããµã¼ãã¹ãAidemyããå ¬éãããWebãã©ã¦ã¶ã§å©ç¨ãããªã³ã©ã¤ã³è¬åº§ãªã®ã§ãå©ç¨è ã¯æå ã®PCã«ããã°ã©ãã³ã°ç°å¢ãæ§ç¯ããã«å¦ã¹ãã Aidemyã¯ã深層å¦ç¿ãèªç¶è¨èªå¦çãæ°å¤äºæ¸¬ãªã©å ¨11è¬åº§ãç¡æã§æä¾ãAIããã°ã©ãã³ã°ã«åºãå©ç¨ããã¦ããPythonè¨èªããNumPyãPandasãscikit-learnãTensorFlowã¨ãã£ãPythonåãã®ã©ã¤ãã©ãªã«ã¤ãã¦ãå¦ã¹ããç
ï¼»éå ±ï¼½Googleãæ©æ¢°å¦ç¿ã«æé©åãããCloud TPUãçºè¡¨ãããããããã¯ãã¾ãAIãã¡ã¼ã¹ããã¼ã¿ã»ã³ã¿ã¼ã®æ§ç¯ã«åãããã£ã¦ããããGoogle I/O 2017 Googleã¯5æ17æ¥ï¼æ¥æ¬æé5æ18æ¥æªæï¼ããéå¬ä¸ã®å¹´æ¬¡ã¤ãã³ããGoogle I/O 2017ãã«ã¦ãæ©æ¢°å¦ç¿ã«æé©åãã第äºä¸ä»£ã®TensorFlow Processing Unitï¼TPUï¼ã¨ãããç¨ããã¯ã©ã¦ããµã¼ãã¹ãCloud TPUããçºè¡¨ãã¾ããã ããããã¯ãã¾ãAIãã¡ã¼ã¹ããã¼ã¿ã»ã³ã¿ã¼ãä½ã£ã¦ãã Google CEO Sundar Pichaiæ°ã ãããããæ¤ç´¢äºæ¥ãå§ããã¨ãããããã¹ã±ã¼ã«ãããããã«ã¯ããã¾ã§ã®è¨ç®æ©ã®ã¢ã¼ããã¯ãã£ãè¦ç´ãå¿ è¦ããã£ããã ããããããããã¯ãã®ããã«ãã¼ã¿ã»ã³ã¿ã¼ãæåããæ§ç¯ãããã¨ã«ãªã£ãã®ã ã ããã¦ãã¾ãæ©æ¢°å¦ç¿ãAIã®é²åã¯ãµã
AIããã°ã©ãã³ã°ã®å°éç¥èããªããã©ãã¢ããã«ã®Siriãã¢ãã¾ã³ã®Alexaã®ãããªãã¼ã½ãã«ã¢ã·ã¹ã¿ã³ããä½ãããéçºè ã¸ãæ軽ã«è©¦ããWebãµã¼ãã¹APIã¨ãã¬ã¼ã ã¯ã¼ã¯ãç´¹ä»ãã¾ãã 人工ç¥è½ããã¼ã½ãã«ã¢ã·ã¹ã¿ã³ãããã£ãããããã¯æ¥éã«æ®åãã¦ãã¾ãããSiriããAlexaããCortanaããOk Googleãã¨ããè¨èã§äººãWebã«ã¤ãªãã§ããããã¼ã½ãã«ã¢ã·ã¹ã¿ã³ãããå¢ãç¶ããIoTï¼Internet of Thingsï¼ã¢ãã®ã¤ã³ã¿ã¼ãããï¼ã¯ãããããã¾ãã¾ãå½ããåã«ãªã£ã¦ãã¾ãã2017å¹´ãã»ã¼ãã¹ã¦ã®ã¡ãã»ã³ã¸ã£ã¼ããã°ã©ã ã¨ã¹ãã¼ããã©ã³OSã«ã¯ãã£ãããããããã¼ã½ãã«ã¢ã·ã¹ã¿ã³ããç¨æããã¦ãã¾ãã ã人工ç¥è½ãã®å®ç¾©ã«ã¤ãã¦ã¯è°è«ã®ä½å°ãããã¾ããã誰ã§ãä»®æ³ã¢ã·ã¹ã¿ã³ããèªç±ã«ä½¿ããä¸çã®å¹éããç§ãã¡ã¯ç®ã®å½ããã«ãã¦ããã®ã§ãã 幸ããªãã¨
ãã£ã¼ãã©ã¼ãã³ã°å®è·µå ¥é ï½Â Kerasã©ã¤ãã©ãªã§ç»åèªèãã¯ããããï¼ ãã£ã¼ãã©ã¼ãã³ã°ï¼æ·±å±¤å¦ç¿ï¼ã«èå³ãããã©ããªããªãæéããªãã¦ãã¨ããæ¹ã®ããã«ãã³ã¼ããåãããªãããããã£ã¨è©¦ãã¦æ触ãã¤ããã§ããããããã«ã解説ãã¾ãã ã¯ããã¾ãã¦ãå®®æ¬åªä¸ã¨ç³ãã¾ãã æè¿ãªã«ãã¨è©±é¡ã®å¤ããã£ã¼ãã©ã¼ãã³ã°ï¼æ·±å±¤å¦ç¿ãdeep learningï¼ãã¨ã³ã¸ãã¢Hubã®èªè ã®æ¹ã§ããèå³ãã人ã¯å¤ãã®ã§ã¯ãªãã§ããããã ãããããã£ã¼ãã©ã¼ãã³ã°ã«ã¤ãã¦å¨ãã®ã¨ã³ã¸ãã¢ã«èãã¦ã¿ãã¨ã ããªããé£ãããã ããªããªãæéããªãã¦ãã©ãããå§ããã°è¯ãããåãããªãã ãä¸å試ãã¦ã¿ããã ãã©ãåå¿è åããã¥ã¼ããªã¢ã«ï¼MNISTãªã©ï¼ãåããã¦æ«æãã¡ãã£ããã ããã ã¨ãã声ãèããã¦ãã¾ãã ããã§ï¼ ãã®è¨äºã§ã¯ãããããæ¹ã対象ã«ããã£ã¼ãã©ã¼ãã³ã°ãããã£ã¨è©¦ãã¦æ触ã
tech.kayac.com Advent Calendar 2015 14æ¥ç®æ å½ã® @_somtd ã§ãã æè¿æãã£ãããå¯ãã£ããã©ããããããªæ°åãã¤ã¥ãã¾ããã è¥ãã¨ã³ã¸ãã¢ãã©ãã©ãå¢ãã¦ç«å ´ãå±ããæ¨ä»ãæ°åã®å¤åã«æè»ã«å¯¾å¿ãã身ä½ã ãã¯å¥åº·ã§ãããã¨é¡ã£ã¦ãã¾ãã ãããªä¸ãä»åã®ã¢ããã³ãã«ã¬ã³ãã¼ã§ã¯æ©æ¢°å¦ç¿ãã¤ãã£ã¦ãä»æ¥ãªã«çã¦ãï¼ããæãã¦ãããããã®ã¢ãã¯ã試ãã¾ãã åé¡ãç°¡åã«ããããã«ï¼æéããªãã¦ãªã¢ã«ãªãã¼ã¿ãç¨æã§ããªãã£ãã®ã§ï¼ä»¥ä¸ã®ãã¼ã¿ã»ãããç¨æãã¾ãã ãã¼ã¿ 100æ¥åã®ããªã«ãçããï¼ãã¨ãããã¼ã¿ sample_wear.txt temperature:æ°æ¸© rainfall:é水確ç outer:ã¢ã¦ã¿ã¼ï¼ã³ã¼ãçï¼ ï¼0:çãªã/1:ã¸ã£ã±ãã/2:ãã¤ãã³ãã¼ã«ã¼/3:ãã¦ã³ã¸ã£ã±ããï¼ tops:ãããï¼ä¸çï¼ ï¼0:T
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}}