tensorflow ã® rnn ã®ãã¥ã¼ããªã¢ã«ã¯ã¡ãã£ã¨è¤éã§ä½ããã£ã¦ãã®ãããã¨ã¯ããããªãã®ã§ãç°¡åãªãã¤ãä½ã£ã¦åããã¦ã¿ãã ããããRNNã¨ã¯ï¼ åºåãã¾ãå ¥åã«ä½¿ããããããããããªå³ãè¦ããã©å ·ä½çã«ä½ãèµ·ãã¦ããã®ãï¼ æ·±å±¤å¦ç¿ã«ããèªç¶è¨èªå¦ç ã¨ããæ¬ã ã¨ã ã¨ãããã¨ãªã®ã§ã1層ã§å ¥åãåºåã2åã®ã¨ãã¯ãä¸å³ã£ã½ãæãã«ãªãã¯ãã ä»»æã®é·ãã®å ¥å (ä»åã¯å ¥åç´ åã2åãªã®ã§é·ã4ãªã [ [3,2], [4,4], [1,0], [9,9] ] ã¿ãããªæã) ã«ã¤ãã¦ã æå¾ã®å ¥å(ä¸ã®ä¾ã§ã¯ [9,9])ã渡ãããã¨ãã®åºå(ã«softmaxã¨ãããããã®)ã§èª¤å·®é¢æ°ã®å¤ãè¨ç®ãã¦ããããæå°åããããã«Wã調æ´ãã¦ãããã¨ã§å¦ç¿ãã¦ããã (å¿ è¦ãªãåå ¥åã«å¯¾ããåºåãä¿åãã¦ããã¦ãããã§èª¤å·®é¢æ°ãè¨ç®ãã¦ãããã¿ãã) å ¥åã®ç´ åæ°ã1ã¤ã§æ´»æ§å
{{#tags}}- {{label}}
{{/tags}}