ãã®è¨äºã¯2017å¹´ã«æ¸ãããã®ã¾ã¾ã ã£ãããã°è¨äºã®ä¾é¤ã§ãã ä»2018å¹´å¹´å§ãªã®ã§ãNIPSåå å ±åãç®ç«ã¤ä¸ã4ã¶æé ããããã§ããæ¸ãããã ã£ãICML2017ã®åå é²ãã²ã¨ã¾ãæ稿ãã¦ããã¾ãã ãããããã㯠ã¢ãã«ãã¼ã¹æ·±å±¤å¼·åå¦ç¿ ã½ããæé© é層æ§ï¼ãªãã·ã§ã³ï¼ ï¼å ±é ¬æ å ±ãªãã§ã®ï¼ç³»çµ±ç«ã£ãæ¢ç´¢ ä¸è²«å¦ç¿ (end-to-end learning) å¦ç¿æã¨ç°ãªãç°å¢ã¸ã®æ±åï¼transfer/zero-shotï¼ å®ç¨çãªæ¢ç´¢ï¼å®å ¨ã»å ¬å¹³ï¼ ããã«ããæ示 æ¹çãªãåå¦ç¿ã§ã®ãã¤ã¢ã¹ã»ããªã¢ã³ã¹ å人çã«é¢ç½ãã£ãçºè¡¨ã»ãããã¯5é¸ ãã¥ã¼ããªã¢ã« ã½ããæé© ã«ãã´ãªã«ã«DQN ãã¼ã¿æ¹ç PVFã«ãããªãã·ã§ã³ã®çºè¦ ç·æ¬ [PR] éç¿ å¼·åå¦ç¿ 8æä¸æ¬ã«ã·ããã¼ã§è¡ãããICML2017ã«åå ãã¦ãã¾ãã ç§èªèº«ã¯èªç¶è¨èªçæã®ã¯ã¼ã¯ã·ã§ããã§çºè¡¨ããã¦ã
{{#tags}}- {{label}}
{{/tags}}