ããã°ãã¯ï¼ ä»æ¥ã¯çµ±è¨ãæ©æ¢°å¦ç¿ã«ããã¦æãåºæ¬ã¨ãªãææ³ã§ããç·å½¢å帰ããåºçºãï¼ï¼ã¤ï¼ã¤ã¢ãã«ãå¦ç¿æ¹æ³ã«å¤æ´ãå ãã¦ããï¼æçµçã«æ·±å±¤å¦ç¿ã®åéã§é常ã«è¯ã使ããã¦ããçæã¢ãã«ã§ããå¤åãªã¼ãã¨ã³ã³ã¼ãï¼variational auto-encoderï¼VAEï¼*1*2ãå°ãã¦ããããã¨æãã¾ãï¼ 2014å¹´ã«çºè¡¨ãããVAEã¯ï¼å¾é è¿ä¼¼ãå¾ãããã®reparametrization trickãï¼å¹ççã«æ½å¨å¤æ°ãè¿ä¼¼æ¨è«ããèªèã¢ãã«ï¼recognition model, inference modelï¼ã®å°å ¥ï¼ç¢ºççå¾é æ³ï¼stochastic gradient descentï¼SGDï¼ã®2éé©ç¨ãªã©ï¼æ§ã ãªã¢ã¤ãã¢ãæ£ãã°ãããã¦ãã確çççæã¢ãã«ã§ãï¼èæ¯ã¨ãã¦ã¯ï¼å½æãã¥ã¼ã©ã«ãããã¯ã¼ã¯ãç¨ãã¦ç»åãçæããã¨ãã£ãã¢ããã¼ããï¼CNNã使ã£ãèå¥ãªã©ã¨æ¯ã¹ã¦ï¼ã
{{#tags}}- {{label}}
{{/tags}}