{"id":9318,"date":"2026-08-04T12:04:33","date_gmt":"2026-08-04T03:04:33","guid":{"rendered":"https:\/\/since2020.jp\/media\/?p=9318"},"modified":"2026-08-04T12:04:33","modified_gmt":"2026-08-04T03:04:33","slug":"rl-mdp","status":"publish","type":"post","link":"https:\/\/since2020.jp\/media\/rl-mdp\/","title":{"rendered":"\u5f37\u5316\u5b66\u7fd2\u5165\u9580\uff1aMDP\uff08\u30de\u30eb\u30b3\u30d5\u6c7a\u5b9a\u904e\u7a0b\uff09\u3067\u7406\u89e3\u3059\u308b\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306e\u610f\u601d\u6c7a\u5b9a"},"content":{"rendered":"\n<h3 class=\"wp-block-heading\">\u5f37\u5316\u5b66\u7fd2\u5165\u9580\uff1aMDP\uff08\u30de\u30eb\u30b3\u30d5\u6c7a\u5b9a\u904e\u7a0b\uff09\u3067\u7406\u89e3\u3059\u308b\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306e\u610f\u601d\u6c7a\u5b9a<\/h3>\n\n\n\n<p>AI\u304c\u56f2\u7881\u306e\u4e16\u754c\u30c1\u30e3\u30f3\u30d4\u30aa\u30f3\u3092\u7834\u308a\u3001\u30ed\u30dc\u30c3\u30c8\u304c\u8ee2\u3070\u305a\u306b\u6b69\u3051\u308b\u3088\u3046\u306b\u306a\u308b\u2014\u2014\u3053\u3046\u3057\u305f\u6210\u679c\u3092\u652f\u3048\u308b\u6280\u8853\u304c\u5f37\u5316\u5b66\u7fd2\uff08Reinforcement Learning\uff09\u3067\u3059\u3002\u672c\u8a18\u4e8b\u3067\u306f\u3001\u5f37\u5316\u5b66\u7fd2\u306e\u57fa\u672c\u304b\u3089\u3001\u6570\u5b66\u7684\u306a\u67a0\u7d44\u307f\u3067\u3042\u308bMDP\uff08\u30de\u30eb\u30b3\u30d5\u6c7a\u5b9a\u904e\u7a0b\uff09\u306e\u6982\u5ff5\u3001\u305d\u3057\u3066Python\u306b\u3088\u308b\u5b9f\u88c5\u307e\u3067\u9806\u3092\u8ffd\u3063\u3066\u89e3\u8aac\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">1. \u5f37\u5316\u5b66\u7fd2\u3068\u306f\u4f55\u304b\uff1f<\/h3>\n\n\n\n<p>\u5f37\u5316\u5b66\u7fd2\u3068\u306f\u3001\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\uff08Agent\uff09\u304c\u74b0\u5883\u306e\u4e2d\u3067\u8a66\u884c\u932f\u8aa4\u3057\u306a\u304c\u3089\u3001\u6700\u9069\u306a\u884c\u52d5\u65b9\u91dd\u3092\u81ea\u5206\u3067\u5b66\u3093\u3067\u3044\u304f\u6a5f\u68b0\u5b66\u7fd2\u306e\u624b\u6cd5\u3067\u3059\u3002<\/p>\n\n\n\n<p>\u300c\u3046\u307e\u304f\u3044\u3063\u305f\u884c\u52d5\u3092\u7e70\u308a\u8fd4\u3057\u3001\u5931\u6557\u3057\u305f\u884c\u52d5\u3092\u907f\u3051\u308b\u300d\u3068\u3044\u3046\u4eba\u9593\u306e\u5b66\u7fd2\u30d7\u30ed\u30bb\u30b9\u306b\u4f3c\u305f\u4ed5\u7d44\u307f\u3067\u3059\u3002<\/p>\n\n\n\n<p>\u5f37\u5316\u5b66\u7fd2\u306e\u4e3b\u306a\u767b\u5834\u4eba\u7269\u306f\u4ee5\u4e0b\u306e3\u3064\u3067\u3059\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8<\/strong>\uff1a\u5b66\u7fd2\u30fb\u610f\u601d\u6c7a\u5b9a\u3092\u884c\u3046\u4e3b\u4f53<\/li>\n\n\n\n<li><strong>\u74b0\u5883<\/strong>\uff1a\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u304c\u884c\u52d5\u3059\u308b\u5834<\/li>\n\n\n\n<li><strong>\u5831\u916c\uff08Reward\uff09<\/strong>\uff1a\u884c\u52d5\u306e\u7d50\u679c\u3068\u3057\u3066\u74b0\u5883\u304b\u3089\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u3078\u4e0e\u3048\u3089\u308c\u308b\u5373\u6642\u5831\u916c\uff08\u30b9\u30ab\u30e9\u30fc\u5024\uff09<\/li>\n<\/ul>\n\n\n\n<p>\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306f\u74b0\u5883\u304b\u3089\u72b6\u614b\u3092\u53d7\u3051\u53d6\u308a\u3001\u884c\u52d5\u3092\u9078\u629e\u3057\u3001\u305d\u306e\u7d50\u679c\u3068\u3057\u3066\u5831\u916c\u3092\u53d7\u3051\u53d6\u308a\u307e\u3059\u3002\u3053\u306e\u7e70\u308a\u8fd4\u3057\u3092\u901a\u3058\u3066\u3001\u7d2f\u7a4d\u5831\u916c\u304c\u6700\u5927\u306b\u306a\u308b\u3088\u3046\u306a\u884c\u52d5\u65b9\u91dd\uff08\u30dd\u30ea\u30b7\u30fc\uff09\u3092\u5b66\u7fd2\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">2. \u5f37\u5316\u5b66\u7fd2\u306e\u5b9f\u7528\u4f8b<\/h3>\n\n\n\n<p>\u5f37\u5316\u5b66\u7fd2\u306f\u3001\u3055\u307e\u3056\u307e\u306a\u5206\u91ce\u3067\u5b9f\u7528\u5316\u304c\u9032\u3093\u3067\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<h5 class=\"wp-block-heading\">1. \u30b2\u30fc\u30e0AI<\/h5>\n\n\n\n<p>DeepMind\u304c\u958b\u767a\u3057\u305fAlphaGo\u306f\u3001\u56f2\u7881\u306e\u4e16\u754c\u30c1\u30e3\u30f3\u30d4\u30aa\u30f3\u3092\u7834\u3063\u305f\u5f37\u5316\u5b66\u7fd2\u30d9\u30fc\u30b9\u306eAI\u3067\u3059\u3002\u307e\u305f\u3001OpenAI Five\u306fDota 2\u3067\u30d7\u30ed\u30b2\u30fc\u30de\u30fc\u30c1\u30fc\u30e0\u306b\u52dd\u5229\u3057\u307e\u3057\u305f\u3002<\/p>\n\n\n\n<h5 class=\"wp-block-heading\">2. \u30ed\u30dc\u30c3\u30c8\u5236\u5fa1<\/h5>\n\n\n\n<p>\u30ed\u30dc\u30c3\u30c8\u30a2\u30fc\u30e0\u306e\u628a\u6301\u52d5\u4f5c\u3084\u4e8c\u8db3\u6b69\u884c\u30ed\u30dc\u30c3\u30c8\u306e\u6b69\u884c\u5236\u5fa1\u306a\u3069\u3001\u8907\u96d1\u306a\u52d5\u4f5c\u306e\u81ea\u52d5\u7fd2\u5f97\u306b\u5f37\u5316\u5b66\u7fd2\u304c\u4f7f\u308f\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<h5 class=\"wp-block-heading\">3. \u81ea\u52d5\u904b\u8ee2<\/h5>\n\n\n\n<p>\u8eca\u4e21\u304c\u30d6\u30ec\u30fc\u30ad\u30fb\u52a0\u901f\u30fb\u30cf\u30f3\u30c9\u30eb\u64cd\u4f5c\u3092\u3069\u306e\u30bf\u30a4\u30df\u30f3\u30b0\u3067\u884c\u3046\u3079\u304d\u304b\u3092\u3001\u30b7\u30df\u30e5\u30ec\u30fc\u30b7\u30e7\u30f3\u74b0\u5883\u3067\u5f37\u5316\u5b66\u7fd2\u306b\u3088\u308a\u5b66\u7fd2\u3055\u305b\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<h5 class=\"wp-block-heading\">4. \u63a8\u85a6\u30b7\u30b9\u30c6\u30e0<\/h5>\n\n\n\n<p>Netflix\u3084YouTube\u306a\u3069\u306e\u52d5\u753b\u63a8\u85a6\u306b\u304a\u3044\u3066\u3082\u3001\u30e6\u30fc\u30b6\u30fc\u306e\u8996\u8074\u884c\u52d5\u3068\u3044\u3046\u5831\u916c\u3092\u3082\u3068\u306b\u6700\u9069\u306a\u63a8\u85a6\u3092\u5b66\u7fd2\u3059\u308b\u5f37\u5316\u5b66\u7fd2\u304c\u5fdc\u7528\u3055\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<h5 class=\"wp-block-heading\">5. \u91d1\u878d\u30fb\u30c8\u30ec\u30fc\u30c7\u30a3\u30f3\u30b0<\/h5>\n\n\n\n<p>\u682a\u5f0f\u5e02\u5834\u306b\u304a\u3051\u308b\u58f2\u8cb7\u30bf\u30a4\u30df\u30f3\u30b0\u306e\u6700\u9069\u5316\u306b\u3082\u3001\u5f37\u5316\u5b66\u7fd2\u304c\u7814\u7a76\u30fb\u6d3b\u7528\u3055\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">3. MDP\uff08\u30de\u30eb\u30b3\u30d5\u6c7a\u5b9a\u904e\u7a0b\uff09\u3068\u306f\uff1f<\/h3>\n\n\n\n<p>\u5f37\u5316\u5b66\u7fd2\u306e\u554f\u984c\u3092\u6570\u5b66\u7684\u306b\u5b9a\u5f0f\u5316\u3059\u308b\u305f\u3081\u306e\u6700\u3082\u57fa\u672c\u7684\u306a\u67a0\u7d44\u307f\u304c**MDP\uff08Markov Decision Process\uff1a\u30de\u30eb\u30b3\u30d5\u6c7a\u5b9a\u904e\u7a0b\uff09**\u3067\u3059\u3002<\/p>\n\n\n\n<p>MDP\u306f\u4ee5\u4e0b\u306e4\u3064\u306e\u8981\u7d20\u3067\u5b9a\u7fa9\u3055\u308c\u307e\u3059\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>S\uff08\u72b6\u614b\u96c6\u5408 \/ State\uff09<\/strong>\uff1a\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u304c\u53d6\u308a\u3046\u308b\u72b6\u614b\u306e\u96c6\u5408<\/li>\n\n\n\n<li><strong>A\uff08\u884c\u52d5\u96c6\u5408 \/ Action\uff09<\/strong>\uff1a\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u304c\u9078\u629e\u3067\u304d\u308b\u884c\u52d5\u306e\u96c6\u5408<\/li>\n\n\n\n<li><strong>R\uff08\u5831\u916c\u95a2\u6570 \/ Reward\uff09<\/strong>\uff1a\u72b6\u614b\u3068\u884c\u52d5\u306e\u30da\u30a2\u306b\u5bfe\u3057\u3001\u5373\u6642\u5831\u916c\u3092\u8fd4\u3059\u95a2\u6570<\/li>\n\n\n\n<li><strong>P\uff08\u9077\u79fb\u78ba\u7387 \/ Transition Probability\uff09<\/strong>\uff1a\u884c\u52d5\u306b\u3088\u3063\u3066\u72b6\u614b\u304c\u3069\u3046\u5909\u5316\u3059\u308b\u304b\u306e\u78ba\u7387<\/li>\n<\/ul>\n\n\n\n<p>\u300c\u30de\u30eb\u30b3\u30d5\u300d\u3068\u3044\u3046\u540d\u524d\u306f\u3001\u6b21\u306e\u72b6\u614b\u304c\u73fe\u5728\u306e\u72b6\u614b\u3068\u884c\u52d5\u3060\u3051\u3067\u6c7a\u307e\u308a\u3001\u904e\u53bb\u306e\u5c65\u6b74\u306b\u4f9d\u5b58\u3057\u306a\u3044\u3068\u3044\u3046\u300c\u30de\u30eb\u30b3\u30d5\u6027\u300d\u306b\u7531\u6765\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">4. \u72b6\u614b\u30fb\u884c\u52d5\u30fb\u5831\u916c\u30fb\u72b6\u614b\u9077\u79fb\u306e\u30b5\u30a4\u30af\u30eb<\/h3>\n\n\n\n<p>MDP\u306b\u304a\u3051\u308b\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306e\u884c\u52d5\u30b5\u30a4\u30af\u30eb\u306f\u6b21\u306e\u3088\u3046\u306b\u8868\u305b\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img decoding=\"async\" width=\"840\" height=\"229\" src=\"https:\/\/since2020.jp\/media\/wp-content\/uploads\/2026\/08\/image-31.png\" alt=\"\" class=\"wp-image-9319\" srcset=\"https:\/\/since2020.jp\/media\/wp-content\/uploads\/2026\/08\/image-31.png 840w, https:\/\/since2020.jp\/media\/wp-content\/uploads\/2026\/08\/image-31-300x82.png 300w, https:\/\/since2020.jp\/media\/wp-content\/uploads\/2026\/08\/image-31-768x209.png 768w\" sizes=\"(max-width: 840px) 100vw, 840px\" \/><\/figure>\n\n\n\n<p>\u3053\u306e\u30b5\u30a4\u30af\u30eb\u3092\u5177\u4f53\u4f8b\u3067\u8003\u3048\u3066\u307f\u307e\u3057\u3087\u3046\u3002<\/p>\n\n\n\n<p><strong>\u4f8b\uff1a\u8ff7\u8def\u3092\u89e3\u304f\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u72b6\u614b\uff08State\uff09<\/strong>\uff1a\u73fe\u5728\u3044\u308b\u30de\u30b9\u76ee\u306e\u4f4d\u7f6e\uff08\u4f8b\uff1a\u4f4d\u7f6e0\u3001\u4f4d\u7f6e1\u3001\u4f4d\u7f6e2 \u306a\u3069\uff09<\/li>\n\n\n\n<li><strong>\u884c\u52d5\uff08Action\uff09<\/strong>\uff1a\u5de6\u30fb\u53f3\u3078\u306e\u79fb\u52d5<\/li>\n\n\n\n<li><strong>\u5831\u916c\uff08Reward\uff09<\/strong>\uff1a\u30b4\u30fc\u30eb\u306b\u5230\u9054\u3057\u305f\u3089 +1\u3001\u305d\u308c\u4ee5\u5916\u306f 0<\/li>\n\n\n\n<li><strong>\u9077\u79fb\uff08Transition\uff09<\/strong>\uff1a\u884c\u52d5\u306b\u5f93\u3063\u3066\u6b21\u306e\u30de\u30b9\u3078\u79fb\u52d5<\/li>\n<\/ul>\n\n\n\n<p>\u3053\u306e\u30b5\u30a4\u30af\u30eb\u3092\u7e70\u308a\u8fd4\u3059\u3053\u3068\u3067\u3001\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306f\u30b4\u30fc\u30eb\u3078\u306e\u6700\u77ed\u7d4c\u8def\uff08\u6700\u9069\u30dd\u30ea\u30b7\u30fc\uff09\u3092\u5b66\u7fd2\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">5. Python\u306b\u3088\u308bMDP\u306e\u7c21\u5358\u306a\u5b9f\u88c5<\/h3>\n\n\n\n<p>\u5148\u306bQ\u5b66\u7fd2\u306e\u6838\u5fc3\u3067\u3042\u308b<strong>Q\u5024<\/strong>\u3092\u8aac\u660e\u3057\u307e\u3059\u3002Q\u5024\u3068\u306f\u300c\u3042\u308b\u72b6\u614b\u3067\u305d\u306e\u884c\u52d5\u3092\u9078\u3093\u3060\u3068\u304d\u3001\u5c06\u6765\u306b\u308f\u305f\u3063\u3066\u5f97\u3089\u308c\u308b\u5831\u916c\u306e\u671f\u5f85\u7dcf\u548c\u300d\u306e\u3053\u3068\u3067\u3059\u3002\u5024\u304c\u9ad8\u3044\u307b\u3069\u300c\u826f\u3044\u9078\u629e\u300d\u3092\u610f\u5473\u3057\u3001\u5b66\u7fd2\u3092\u901a\u3058\u3066\u5f90\u3005\u306b\u66f4\u65b0\u3055\u308c\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u3053\u3053\u3067\u306f\u3001\u30b7\u30f3\u30d7\u30eb\u306a\u300c1\u6b21\u5143\u30b0\u30ea\u30c3\u30c9\u4e16\u754c\u300d\u3067Q\u5b66\u7fd2\uff08Q-Learning\uff09\u3092\u5b9f\u88c5\u3057\u307e\u3059\u3002\u30de\u30b90\u301c4\u304c\u4e26\u3093\u3067\u304a\u308a\u3001\u30de\u30b94\u304c\u30b4\u30fc\u30eb\u3067\u3059\u3002\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306f\u6bce\u56de\u30de\u30b90\u304b\u3089\u30b9\u30bf\u30fc\u30c8\u3057\u3001\u53f3\u306b\u9032\u3093\u3067\u30b4\u30fc\u30eb\u3092\u76ee\u6307\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img decoding=\"async\" width=\"739\" height=\"190\" src=\"https:\/\/since2020.jp\/media\/wp-content\/uploads\/2026\/08\/image-32.png\" alt=\"\" class=\"wp-image-9320\" srcset=\"https:\/\/since2020.jp\/media\/wp-content\/uploads\/2026\/08\/image-32.png 739w, https:\/\/since2020.jp\/media\/wp-content\/uploads\/2026\/08\/image-32-300x77.png 300w\" sizes=\"(max-width: 739px) 100vw, 739px\" \/><\/figure>\n\n\n\n<p>\u306a\u304a\u3001GAMMA\u30fbALPHA\u30fbEPISODES\u30fbEPSILON_END\u306a\u3069\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u306f\u3042\u304f\u307e\u3067\u4e00\u4f8b\u3067\u3059\u3002\u4eca\u56de\u306e\u30dd\u30a4\u30f3\u30c8\u306f\u03b5-decay\u306e\u5c0e\u5165\u3067\u3001\u63a2\u7d22\u7387\u3092\u5e8f\u76e4\u306e1.0\u304b\u3089\u7d42\u76e4\u306e0.05\u3078\u7dda\u5f62\u306b\u4e0b\u3052\u308b\u3053\u3068\u3067\u3001\u5e8f\u76e4\u306e\u5341\u5206\u306a\u63a2\u7d22\u3092\u4fdd\u8a3c\u3057\u3066\u3044\u307e\u3059\u3002\u5404\u5024\u3092\u81ea\u7531\u306b\u5909\u3048\u3066\u52d5\u304b\u3057\u3066\u307f\u308b\u3068\u3001\u5b66\u7fd2\u306e\u6319\u52d5\u306e\u9055\u3044\u304c\u898b\u3048\u3066\u9762\u767d\u3044\u3068\u601d\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>import numpy as np\n\n# \u74b0\u5883\u306e\u8a2d\u5b9a\nNUM_STATES = 5       # \u72b6\u614b\u6570\uff08\u30de\u30b90\u301c4\uff09\nNUM_ACTIONS = 2      # \u884c\u52d5\u6570\uff080:\u5de6\u79fb\u52d5, 1:\u53f3\u79fb\u52d5\uff09\nGOAL_STATE = 4       # \u30b4\u30fc\u30eb\u72b6\u614b\nGAMMA = 0.9          # \u5272\u5f15\u7387\uff08\u5c06\u6765\u306e\u5831\u916c\u3092\u3069\u308c\u3060\u3051\u91cd\u8996\u3059\u308b\u304b\uff09\nALPHA = 0.1          # \u5b66\u7fd2\u7387\nEPSILON_START = 1.0   # \u63a2\u7d22\u7387\u306e\u521d\u671f\u5024\uff08\u5e8f\u76e4\u306f\u5b8c\u5168\u30e9\u30f3\u30c0\u30e0\u63a2\u7d22\uff09\nEPSILON_END = 0.05    # \u63a2\u7d22\u7387\u306e\u6700\u7d42\u5024\uff08\u7d42\u76e4\u306f\u5b66\u7fd2\u7d50\u679c\u3092\u6d3b\u7528\uff09\nEPISODES = 500       # \u5b66\u7fd2\u30a8\u30d4\u30bd\u30fc\u30c9\u6570\n\n# Q\u30c6\u30fc\u30d6\u30eb\u306e\u521d\u671f\u5316\uff08\u72b6\u614b\u00d7\u884c\u52d5 \u2192 Q\u5024\uff09\nQ = np.zeros((NUM_STATES, NUM_ACTIONS))\n\ndef get_reward(state):\n    return 1.0 if state == GOAL_STATE else 0.0\n\ndef step(state, action):\n    if action == 0:  # \u5de6\u79fb\u52d5\n        next_state = max(0, state - 1)\n    else:            # \u53f3\u79fb\u52d5\n        next_state = min(NUM_STATES - 1, state + 1)\n    reward = get_reward(next_state)\n    done = (next_state == GOAL_STATE)\n    return next_state, reward, done\n\n# Q\u5b66\u7fd2\u306e\u30e1\u30a4\u30f3\u30eb\u30fc\u30d7\nfor episode in range(EPISODES):\n    state = 0  # \u30b9\u30bf\u30fc\u30c8\u4f4d\u7f6e\n    # \u30a8\u30d4\u30bd\u30fc\u30c9\u304c\u9032\u3080\u306b\u3064\u308c\u63a2\u7d22\u7387\u3092\u7dda\u5f62\u306b\u4e0b\u3052\u308b\n    epsilon = EPSILON_END + (EPSILON_START - EPSILON_END) * (1 - episode \/ EPISODES)\n\n    for _ in range(100):\n        # \u03b5-greedy\u6cd5\u3067\u884c\u52d5\u9078\u629e\n        if np.random.rand() &lt; epsilon:\n            action = np.random.randint(NUM_ACTIONS)  # \u30e9\u30f3\u30c0\u30e0\u63a2\u7d22\n        else:\n            action = np.argmax(Q&#091;state])             # \u6700\u5584\u884c\u52d5\u3092\u9078\u629e\n\n        next_state, reward, done = step(state, action)\n\n        # Q\u30c6\u30fc\u30d6\u30eb\u306e\u66f4\u65b0\uff08\u30d9\u30eb\u30de\u30f3\u65b9\u7a0b\u5f0f\uff09\n        Q&#091;state, action] += ALPHA * (\n            reward + GAMMA * np.max(Q&#091;next_state]) - Q&#091;state, action]\n        )\n\n        state = next_state\n        if done:\n            break\n\n# \u7d50\u679c\u306e\u8868\u793a\nprint(\"\u5b66\u7fd2\u5f8c\u306eQ\u30c6\u30fc\u30d6\u30eb\uff1a\")\nprint(\"\u72b6\u614b | \u5de6(Q\u5024) | \u53f3(Q\u5024) | \u6700\u9069\u884c\u52d5\")\nfor s in range(NUM_STATES):\n    best = \"\u5de6\" if np.argmax(Q&#091;s]) == 0 else \"\u53f3\"\n    print(f\"  {s}  |  {Q&#091;s,0]:.3f}  |  {Q&#091;s,1]:.3f}  |    {best}\")\n<\/code><\/pre>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n\n<h3 class=\"wp-block-heading\">6. \u5b9f\u884c\u7d50\u679c\u3068\u89e3\u8aac<\/h3>\n\n\n\n<p>\u5b9f\u884c\u3059\u308b\u3068\u4ee5\u4e0b\u306e\u3088\u3046\u306a\u51fa\u529b\u304c\u5f97\u3089\u308c\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img decoding=\"async\" width=\"213\" height=\"103\" src=\"https:\/\/since2020.jp\/media\/wp-content\/uploads\/2026\/08\/image-33.png\" alt=\"\" class=\"wp-image-9321\" \/><\/figure>\n\n\n\n<p><strong>\u7d50\u679c\u306e\u8aad\u307f\u65b9<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u72b6\u614b0\u301c3<\/strong>\uff1a\u53f3\u79fb\u52d5\u306eQ\u5024\u304c\u5de6\u79fb\u52d5\u3088\u308a\u9ad8\u304f\u306a\u3063\u3066\u3044\u307e\u3059\u3002\u53f3\u306b\u9032\u3080\u307b\u3069\u30b4\u30fc\u30eb\uff08\u72b6\u614b4\uff09\u306b\u8fd1\u3065\u304d\u9ad8\u3044\u5831\u916c\u304c\u671f\u5f85\u3067\u304d\u308b\u3053\u3068\u3092\u3001\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u304c\u5b66\u7fd2\u3057\u305f\u7d50\u679c\u3067\u3059\u3002<\/li>\n\n\n\n<li><strong>\u72b6\u614b4\uff08\u30b4\u30fc\u30eb\uff09<\/strong>\uff1a\u30b4\u30fc\u30eb\u5230\u9054\u5f8c\u306f\u30a8\u30d4\u30bd\u30fc\u30c9\u304c\u7d42\u4e86\u3059\u308b\u305f\u3081\u3001\u5de6\u53f3\u3069\u3061\u3089\u306eQ\u5024\u30820\u306e\u307e\u307e\u3067\u3059\u3002Q\u5024\u304c\u540c\u3058\u5834\u5408\u3001\u30d7\u30ed\u30b0\u30e9\u30e0\u306f\u5148\u982d\u306e\u30a4\u30f3\u30c7\u30c3\u30af\u30b9\uff08\uff1d\u5de6\uff09\u3092\u8fd4\u3059\u305f\u3081\u300c\u5de6\u300d\u3068\u8868\u793a\u3055\u308c\u3066\u3044\u307e\u3059\u304c\u3001\u30b4\u30fc\u30eb\u72b6\u614b\u3067\u306e\u884c\u52d5\u306b\u5b9f\u8cea\u7684\u306a\u610f\u5473\u306f\u3042\u308a\u307e\u305b\u3093\u3002<\/li>\n\n\n\n<li><strong>Q\u5024\u306e\u5927\u304d\u3055<\/strong>\uff1a\u30b4\u30fc\u30eb\u304b\u3089\u9060\u3044\u307b\u3069\u5024\u304c\u5c0f\u3055\u304f\u306a\u3063\u3066\u3044\u307e\u3059\u3002\u5272\u5f15\u7387\uff08\u03b3=0.9\uff09\u306b\u3088\u3063\u3066\u3001\u9060\u3044\u5c06\u6765\u306e\u5831\u916c\u307b\u3069\u5c0f\u3055\u304f\u8a55\u4fa1\u3055\u308c\u308b\u305f\u3081\u3067\u3059\u3002<\/li>\n<\/ul>\n\n\n\n<p><strong>\uff08\u88dc\u8db3\uff09\u5de6\u79fb\u52d5\u306eQ\u5024\u304c\u30bc\u30ed\u3067\u306a\u3044\u7406\u7531<\/strong><br>Q\u5b66\u7fd2\u306f\u3059\u3079\u3066\u306e\uff08\u72b6\u614b, \u884c\u52d5\uff09\u30da\u30a2\u306b\u300c\u305d\u306e\u5f8c\u30b4\u30fc\u30eb\u3078\u5230\u9054\u3067\u304d\u308b\u898b\u8fbc\u307f\u306e\u5831\u916c\u300d\u3092\u5272\u308a\u5f53\u3066\u307e\u3059\u3002\u5de6\u79fb\u52d5\u3057\u3066\u3082\u5373\u5ea7\u306b\u30a8\u30d4\u30bd\u30fc\u30c9\u306f\u7d42\u308f\u3089\u305a\u3001\u305d\u306e\u5f8c\u53f3\u3078\u623b\u308c\u3070\u30b4\u30fc\u30eb\u306b\u5230\u9054\u3067\u304d\u308b\u305f\u3081\u3001\u5de6\u79fb\u52d5\u306b\u3082\u5c06\u6765\u306e\u4fa1\u5024\u304c\u751f\u307e\u308c\u307e\u3059\u3002\u305f\u3068\u3048\u3070\u72b6\u614b0\u3067\u5de6\u79fb\u52d5\u3059\u308b\u3068\u58c1\u306a\u306e\u3067\u72b6\u614b0\u306b\u7559\u307e\u308a\u307e\u3059\u304c\u3001\u305d\u306e\u5f8c\u53f3\u306b\u9032\u3081\u3070\u30b4\u30fc\u30eb\u306b\u5c4a\u304f\u305f\u3081 Q[0, \u5de6] = \u03b3 \u00d7 Q[0, \u53f3] = 0.9 \u00d7 0.729 \u2248 0.656 \u3068\u306a\u308a\u307e\u3059\u3002\u3053\u308c\u306f\u30b3\u30fc\u30c9\u306e\u30d0\u30b0\u3067\u306f\u306a\u304f\u3001\u30da\u30ca\u30eb\u30c6\u30a3\u306e\u306a\u3044\u74b0\u5883\u3067Q\u5b66\u7fd2\u304c\u6b63\u3057\u304f\u52d5\u3044\u305f\u7d50\u679c\u3067\u3059\u3002\u53f3\u79fb\u52d5\u306eQ\u5024\u304c\u5e38\u306b\u5de6\u3092\u4e0a\u56de\u3063\u3066\u3044\u308b\u305f\u3081\u3001\u6700\u9069\u884c\u52d5\u306f\u6b63\u3057\u304f\u300c\u53f3\u300d\u3068\u5b66\u7fd2\u3067\u304d\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u3053\u306e\u30b7\u30f3\u30d7\u30eb\u306a\u4f8b\u3067\u3082\u3001\u300c\u72b6\u614b\u2192\u884c\u52d5\u2192\u5831\u916c\u2192\u6b21\u306e\u72b6\u614b\u300d\u306e\u30b5\u30a4\u30af\u30eb\u3092\u7e70\u308a\u8fd4\u3059\u3060\u3051\u3067\u3001\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u304c\u6700\u9069\u306a\u7d4c\u8def\u3092\u81ea\u3089\u5b66\u7fd2\u3067\u304d\u308b\u3053\u3068\u304c\u78ba\u8a8d\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">\u307e\u3068\u3081<\/h3>\n\n\n\n<p>\u672c\u8a18\u4e8b\u306e\u5185\u5bb9\u3092\u307e\u3068\u3081\u307e\u3059\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u5f37\u5316\u5b66\u7fd2<\/strong>\uff1a\u8a66\u884c\u932f\u8aa4\u3092\u901a\u3058\u3066\u6700\u9069\u306a\u884c\u52d5\u3092\u5b66\u7fd2\u3059\u308b\u624b\u6cd5<\/li>\n\n\n\n<li><strong>MDP<\/strong>\uff1a\u72b6\u614b\u30fb\u884c\u52d5\u30fb\u5831\u916c\u30fb\u9077\u79fb\u78ba\u7387\u3067\u554f\u984c\u3092\u5b9a\u5f0f\u5316\u3059\u308b\u67a0\u7d44\u307f<\/li>\n\n\n\n<li><strong>Q\u5b66\u7fd2<\/strong>\uff1aMDP\u3092\u89e3\u304f\u305f\u3081\u306e\u4ee3\u8868\u7684\u306a\u30a2\u30eb\u30b4\u30ea\u30ba\u30e0\u3067\u3001Q\u30c6\u30fc\u30d6\u30eb\u3092\u66f4\u65b0\u3057\u306a\u304c\u3089\u6700\u9069\u30dd\u30ea\u30b7\u30fc\u3092\u7fd2\u5f97\u3059\u308b<\/li>\n<\/ul>\n\n\n\n<p>\u5f37\u5316\u5b66\u7fd2\u306f\u307e\u3060\u307e\u3060\u767a\u5c55\u9014\u4e0a\u306e\u5206\u91ce\u3067\u3042\u308a\u3001\u3088\u308a\u8907\u96d1\u306a\u74b0\u5883\u3084\u5927\u898f\u6a21\u306a\u554f\u984c\u306b\u5bfe\u5fdc\u3059\u308b\u305f\u3081\u306b\u3001\u6df1\u5c64\u5b66\u7fd2\u3068\u7d44\u307f\u5408\u308f\u305b\u305f\u6df1\u5c64\u5f37\u5316\u5b66\u7fd2\uff08Deep Reinforcement Learning\uff09\u306a\u3069\u3082\u76db\u3093\u306b\u7814\u7a76\u3055\u308c\u3066\u3044\u307e\u3059\u3002\u8208\u5473\u304c\u3042\u308b\u65b9\u306f\u3001\u305c\u3072\u305d\u3061\u3089\u3082\u8abf\u3079\u3066\u307f\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u5f37\u5316\u5b66\u7fd2\u5165\u9580\uff1aMDP\uff08\u30de\u30eb\u30b3\u30d5\u6c7a\u5b9a\u904e\u7a0b\uff09\u3067\u7406\u89e3\u3059\u308b\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306e\u610f\u601d\u6c7a\u5b9a AI\u304c\u56f2\u7881\u306e\u4e16\u754c\u30c1\u30e3\u30f3\u30d4\u30aa\u30f3\u3092\u7834\u308a\u3001\u30ed\u30dc\u30c3\u30c8\u304c\u8ee2\u3070\u305a\u306b\u6b69\u3051\u308b\u3088\u3046\u306b\u306a\u308b\u2014\u2014\u3053\u3046\u3057\u305f\u6210\u679c\u3092\u652f\u3048\u308b\u6280\u8853\u304c\u5f37\u5316\u5b66\u7fd2\uff08Reinforcement Lear [&hellip;]<\/p>\n","protected":false},"author":4,"featured_media":9322,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"content-type":"","swell_btn_cv_data":"","footnotes":"","_wp_rev_ctl_limit":""},"categories":[1251],"tags":[96,629,419],"class_list":["post-9318","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-data-analysis","tag-ai","tag-629","tag-419"],"_links":{"self":[{"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/posts\/9318","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/users\/4"}],"replies":[{"embeddable":true,"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/comments?post=9318"}],"version-history":[{"count":1,"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/posts\/9318\/revisions"}],"predecessor-version":[{"id":9323,"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/posts\/9318\/revisions\/9323"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/media\/9322"}],"wp:attachment":[{"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/media?parent=9318"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/categories?post=9318"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/since2020.jp\/media\/wp-json\/wp\/v2\/tags?post=9318"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}