èæ¯ ã¯ããã¾ãã¦ãJXé信社ã§ã¤ã³ã¿ã¼ã³ããã¦ããåç°ã§ãã è¿å¹´æ·±å±¤å¦ç¿ã§ã¯ã¢ãã«ãè¥å¤§åããå¾åã«ããã¾ãã2020å¹´ã«open aiã示ããScaling Lawsï¼[2001.08361] Scaling Laws for Neural Language Modelsï¼ ã®è¡æã¯è¨æ¶ã«æ°ãããMLP-Mixerã示ããããã«ãã¢ãã«ã大ããããã°Attentionæ§é ãCNNã§ãããä¸å¿ è¦ã¨ãã説ãããã¾ããï¼[2105.01601] MLP-Mixer: An all-MLP Architecture for Visionï¼ ããã大ããªæ·±å±¤å¦ç¿ã¢ãã«ãå©ç¨ãããã¨ããã¨ããã°ãã°ä»¥ä¸ã®ãããªåé¡ã«æ©ã¾ããã¾ãã æ¨è«é度ãåé¡ã§ãããã¯ãã«å®è£ ä¸å¯è½ GPU/TPUã¯ã³ã¹ãä¸å³ãã ãããã¯ãã®æ§è³ªä¸ãããå¦çãä¸å¯è½ï¼å¹ççã«GPU/TPUãå©ç¨ã§ããªãï¼ ä¾ãã°JXé信社
{{#tags}}- {{label}}
{{/tags}}