{"id":40434,"date":"2024-11-04T01:51:42","date_gmt":"2024-11-04T01:51:42","guid":{"rendered":"https:\/\/atmokpo.com\/w\/?p=40434"},"modified":"2024-11-26T06:39:11","modified_gmt":"2024-11-26T06:39:11","slug":"51-%eb%aa%a8%eb%8d%b8-%ea%b8%b0%eb%b0%98-%ea%b0%95%ed%99%94%ed%95%99%ec%8a%b5model-based-rl%ea%b3%bc-%eb%aa%a8%eb%8d%b8-%ed%94%84%eb%a6%ac-%ea%b0%95%ed%99%94%ed%95%99%ec%8a%b5%ec%9d%98-%eb%b9%84","status":"publish","type":"post","link":"https:\/\/atmokpo.com\/w\/40434\/","title":{"rendered":"51. \ubaa8\ub378 \uae30\ubc18 \uac15\ud654\ud559\uc2b5(Model-Based RL)\uacfc \ubaa8\ub378 \ud504\ub9ac \uac15\ud654\ud559\uc2b5\uc758 \ube44\uad50, \ubaa8\ub378 \uae30\ubc18 RL\uc758 \uc7a5\uc810\uacfc \ub2e8\uc810"},"content":{"rendered":"

<\/p>\n

\uac15\ud654\ud559\uc2b5(reinforcement learning, RL)\uc740 \uc5d0\uc774\uc804\ud2b8\uac00 \ud658\uacbd\uacfc \uc0c1\ud638\uc791\uc6a9\ud558\uba74\uc11c \ucd5c\uc801\uc758 \ud589\ub3d9 \uc815\ucc45\uc744 \ud559\uc2b5\ud558\ub294 \uacfc\uc815\uc785\ub2c8\ub2e4. \uac15\ud654\ud559\uc2b5\uc758 \uc8fc\uc694 \uae30\ubc95 \uc911 \ud558\ub098\ub294 \ubaa8\ub378 \uae30\ubc18 \uac15\ud654\ud559\uc2b5(Model-Based RL)\uacfc \ubaa8\ub378 \ud504\ub9ac \uac15\ud654\ud559\uc2b5(Model-Free RL)\uc785\ub2c8\ub2e4. \uc774 \uae00\uc5d0\uc11c\ub294 \ub450 \ubc29\uc2dd\uc758 \ucc28\uc774\uc810, \uc7a5\uc810\uacfc \ub2e8\uc810, \uadf8\ub9ac\uace0 \ubaa8\ub378 \uae30\ubc18 \uac15\ud654\ud559\uc2b5\uc758 \uc608\uc81c\ub97c \ub2e4\ub8e8\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n

1. \ubaa8\ub378 \uae30\ubc18 \uac15\ud654\ud559\uc2b5\uc774\ub780?<\/h2>\n

\ubaa8\ub378 \uae30\ubc18 \uac15\ud654\ud559\uc2b5\uc740 \uc5d0\uc774\uc804\ud2b8\uac00 \ud658\uacbd\uc758 \ubaa8\ub378\uc744 \ud559\uc2b5\ud558\ub294 \ubc29\ubc95\uc785\ub2c8\ub2e4. \ud658\uacbd\uc758 \uc0c1\ud0dc\uc640 \ud589\ub3d9\uc758 \uacb0\uacfc\ub97c \uc608\uce21\ud560 \uc218 \uc788\ub294 \ubaa8\ub378\uc744 \uad6c\ucd95\ud558\uc5ec \uc774\ub97c \ud1b5\ud574 \ucd5c\uc801\uc758 \uc815\ucc45\uc744 \ucc3e\uace0, \ud544\uc694 \uc2dc \uc2dc\ubbac\ub808\uc774\uc158\uc744 \ud1b5\ud574 \ud589\ub3d9\uc744 \uacb0\uc815\ud569\ub2c8\ub2e4.<\/p>\n

1.1 \ubaa8\ub378 \uae30\ubc18 RL\uc758 \uacfc\uc815<\/h3>\n
    \n
  1. \ud658\uacbd \ubaa8\ub378 \ud559\uc2b5:<\/strong> \uc5d0\uc774\uc804\ud2b8\ub294 \ud658\uacbd\uc758 \ub3d9\uc801 \ud2b9\uc131\uc744 \ud559\uc2b5\ud569\ub2c8\ub2e4. \uc774\ub97c \ud1b5\ud574 \uc0c1\ud0dc \uc804\uc774\uc640 \ubcf4\uc0c1\uc744 \uc608\uce21\ud560 \uc218 \uc788\ub294 \ubaa8\ub378(P(s’|s,a)<\/em>\uc640 R(s,a)<\/em>)\uc744 \uad6c\ucd95\ud569\ub2c8\ub2e4.<\/li>\n
  2. \uc815\ucc45 \ud559\uc2b5:<\/strong> \ud559\uc2b5\ud55c \ubaa8\ub378\uc744 \uc0ac\uc6a9\ud558\uc5ec \ucd5c\uc801\uc758 \uc815\ucc45\uc744 \ucc3e\uc2b5\ub2c8\ub2e4. \uc8fc\ub85c \uac00\uce58 \ud568\uc218\ub098 \uc815\ucc45\uc744 \ucd5c\uc801\ud654\ud558\ub294 \ubc29\ubc95\uc744 \uc0ac\uc6a9\ud569\ub2c8\ub2e4.<\/li>\n
  3. \ud589\ub3d9 \uc120\ud0dd:<\/strong> \ud658\uacbd \ubaa8\ub378\uc744 \ubc14\ud0d5\uc73c\ub85c \ud589\ub3d9\uc744 \uc120\ud0dd\ud558\uace0, \uc774\ub97c \ud1b5\ud574 \uc2e4\uc81c \ud658\uacbd\uacfc \uc0c1\ud638\uc791\uc6a9\ud558\uc5ec \uc0c8\ub85c\uc6b4 \ub370\uc774\ud130\ub97c \uc218\uc9d1\ud569\ub2c8\ub2e4.<\/li>\n<\/ol>\n

    2. \ubaa8\ub378 \ud504\ub9ac \uac15\ud654\ud559\uc2b5\uc774\ub780?<\/h2>\n

    \ubaa8\ub378 \ud504\ub9ac \uac15\ud654\ud559\uc2b5\uc740 \ud658\uacbd\uc758 \ubaa8\ub378\uc744 \uad6c\ucd95\ud558\uc9c0 \uc54a\uace0, \uc9c1\uc811\uc801\uc778 \uacbd\ud5d8\uc744 \uae30\ubc18\uc73c\ub85c \uc815\ucc45\uc744 \ud559\uc2b5\ud558\ub294 \ubc29\ubc95\uc785\ub2c8\ub2e4. \uc77c\ubc18\uc801\uc73c\ub85c \uac12\uc744 \ud3c9\uac00\ud558\ub294 \ubc29\ubc95(state-value function, action-value function)\uc744 \ud1b5\ud574 \ucd5c\uc801\uc758 \uc815\ucc45\uc744 \uacb0\uc815\ud569\ub2c8\ub2e4.<\/p>\n

    2.1 \ubaa8\ub378 \ud504\ub9ac RL\uc758 \uc7a5\uc810\uacfc \ub2e8\uc810<\/h3>\n

    \ubaa8\ub378 \ud504\ub9ac \uac15\ud654\ud559\uc2b5\uc740 \ubaa8\ub378\uc744 \uc9c1\uc811 \ud559\uc2b5\ud558\uc9c0 \uc54a\uae30 \ub54c\ubb38\uc5d0 \uc801\uc6a9\uc774 \uac04\ud3b8\ud558\uace0, \ubcf5\uc7a1\ud55c \ud658\uacbd\uc5d0\uc11c\ub3c4 \uc27d\uac8c \uc0ac\uc6a9\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uadf8\ub7ec\ub098 \ub9ce\uc740 \ub370\uc774\ud130\ub97c \ud544\uc694\ub85c \ud558\uace0, \ud0d0\uc0c9\uacfc \ud65c\uc6a9\uc758 \uade0\ud615\uc744 \uc798 \ub9de\ucdb0\uc57c \uacb0\uacfc\ub97c \uc5bb\uc744 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n

    3. \ubaa8\ub378 \uae30\ubc18 VS \ubaa8\ub378 \ud504\ub9ac \uac15\ud654\ud559\uc2b5<\/h2>\n

    \ub450 \ubc29\uc2dd\uc740 \uc0c1\ubc18\ub41c \uc811\uadfc\ubc95\uc744 \uac00\uc9c0\uace0 \uc788\uc73c\uba70, \uc0c1\ud669\uc5d0 \ub530\ub77c \uc7a5\ub2e8\uc810\uc774 \uc788\uc2b5\ub2c8\ub2e4. \ub2e4\uc74c \ud45c\ub294 \ub450 \ubc29\uc2dd\uc758 \ube44\uad50\uc785\ub2c8\ub2e4.<\/p>\n\n\n\n\n\n\n
    \ud2b9\uc9d5<\/th>\n\ubaa8\ub378 \uae30\ubc18 \uac15\ud654\ud559\uc2b5<\/th>\n\ubaa8\ub378 \ud504\ub9ac \uac15\ud654\ud559\uc2b5<\/th>\n<\/tr>\n
    \ubaa8\ub378 \uc0ac\uc6a9<\/td>\n\ud658\uacbd \ubaa8\ub378 \uc0ac\uc6a9<\/td>\n\ubaa8\ub378 \ubbf8\uc0ac\uc6a9<\/td>\n<\/tr>\n
    \ub370\uc774\ud130 \ud6a8\uc728\uc131<\/td>\n\uc0c1\ub300\uc801\uc73c\ub85c \ud6a8\uc728\uc801<\/td>\n\uc0c1\ub300\uc801\uc73c\ub85c \ube44\ud6a8\uc728\uc801<\/td>\n<\/tr>\n
    \ud6c8\ub828 \uc2dc\uac04<\/td>\n\uc0c1\ub300\uc801\uc73c\ub85c \ube60\ub984<\/td>\n\uc0c1\ub300\uc801\uc73c\ub85c \ub290\ub9bc<\/td>\n<\/tr>\n
    \ubcf5\uc7a1\ud55c \ud658\uacbd<\/td>\n\uc131\ub2a5 \uc800\ud558 \uac00\ub2a5\uc131<\/td>\n\uc0c1\ub300\uc801\uc73c\ub85c \uc6d0\ud65c\ud568<\/td>\n<\/tr>\n<\/table>\n

    4. \ubaa8\ub378 \uae30\ubc18 \uac15\ud654\ud559\uc2b5\uc758 \uc7a5\uc810\uacfc \ub2e8\uc810<\/h2>\n

    4.1 \uc7a5\uc810<\/h3>\n