{"id":444,"date":"2023-09-12T14:30:37","date_gmt":"2023-09-12T06:30:37","guid":{"rendered":"http:\/\/lingbo.online\/?p=444"},"modified":"2023-09-13T22:52:35","modified_gmt":"2023-09-13T14:52:35","slug":"%e5%bc%ba%e5%8c%96%e5%ad%a6%e4%b9%a0%e5%85%a5%e9%97%a8%ef%bc%88%e5%9f%ba%e7%a1%80%e7%90%86%e8%ae%ba%ef%bc%89","status":"publish","type":"post","link":"https:\/\/lingbo.online\/index.php\/uncategorized\/%e5%bc%ba%e5%8c%96%e5%ad%a6%e4%b9%a0%e5%85%a5%e9%97%a8%ef%bc%88%e5%9f%ba%e7%a1%80%e7%90%86%e8%ae%ba%ef%bc%89\/","title":{"rendered":"\u5f3a\u5316\u5b66\u4e60\u5165\u95e8\uff08\u57fa\u7840\u7406\u8bba\uff09"},"content":{"rendered":"<h3>\u5f3a\u5316\u5b66\u4e60\u4ecb\u7ecd<\/h3>\n<p>\u5f3a\u5316\u5b66\u4e60\uff08\u82f1\u8bed\uff1aReinforcement learning\uff0c\u7b80\u79f0RL\uff09\u662f\u673a\u5668\u5b66\u4e60\u4e2d\u7684\u4e00\u4e2a\u9886\u57df\uff0c\u5f3a\u8c03\u5982\u4f55\u57fa\u4e8e\u73af\u5883\u800c\u884c\u52a8\uff0c\u4ee5\u53d6\u5f97\u6700\u5927\u5316\u7684\u9884\u671f\u5229\u76ca\u3002\u5f3a\u5316\u5b66\u4e60\u662f\u9664\u4e86\u76d1\u7763\u5b66\u4e60\u548c\u975e\u76d1\u7763\u5b66\u4e60\u4e4b\u5916\u7684\u7b2c\u4e09\u79cd\u57fa\u672c\u7684\u673a\u5668\u5b66\u4e60\u65b9\u6cd5\u3002\u4e0e\u76d1\u7763\u5b66\u4e60\u4e0d\u540c\u7684\u662f\uff0c\u5f3a\u5316\u5b66\u4e60\u4e0d\u9700\u8981\u5e26\u6807\u7b7e\u7684\u8f93\u5165\u8f93\u51fa\u5bf9\uff0c\u540c\u65f6\u4e5f\u65e0\u9700\u5bf9\u975e\u6700\u4f18\u89e3\u7684\u7cbe\u786e\u5730\u7ea0\u6b63\u3002\u5176\u5173\u6ce8\u70b9\u5728\u4e8e\u5bfb\u627e\u63a2\u7d22\uff08\u5bf9\u672a\u77e5\u9886\u57df\u7684\uff09\u548c\u5229\u7528\uff08\u5bf9\u5df2\u6709\u77e5\u8bc6\u7684\uff09\u7684\u5e73\u8861\uff0c\u5f3a\u5316\u5b66\u4e60\u4e2d\u7684\u201c\u63a2\u7d22-\u5229\u7528\u201d\u7684\u4ea4\u6362\uff0c\u5728\u591a\u81c2\u8001\u864e\u673a\uff08\u82f1\u8bed\uff1amulti-armed bandit\uff09\u95ee\u9898\u548c\u6709\u9650MDP\u4e2d\u7814\u7a76\u5f97\u6700\u591a\u3002<\/p>\n<p>\u5176\u7075\u611f\u6765\u6e90\u4e8e\u5fc3\u7406\u5b66\u4e2d\u7684\u884c\u4e3a\u4e3b\u4e49\u7406\u8bba\uff0c\u5373\u6709\u673a\u4f53\u5982\u4f55\u5728\u73af\u5883\u7ed9\u4e88\u7684\u5956\u52b1\u6216\u60e9\u7f5a\u7684\u523a\u6fc0\u4e0b\uff0c\u9010\u6b65\u5f62\u6210\u5bf9\u523a\u6fc0\u7684\u9884\u671f\uff0c\u4ea7\u751f\u80fd\u83b7\u5f97\u6700\u5927\u5229\u76ca\u7684\u4e60\u60ef\u6027\u884c\u4e3a\u3002\u8fd9\u4e2a\u65b9\u6cd5\u5177\u6709\u666e\u9002\u6027\uff0c\u56e0\u6b64\u5728\u5176\u4ed6\u8bb8\u591a\u9886\u57df\u90fd\u6709\u7814\u7a76\uff0c\u4f8b\u5982\u535a\u5f08\u8bba\u3001\u63a7\u5236\u8bba\u3001\u8fd0\u7b79\u5b66\u3001\u4fe1\u606f\u8bba\u3001\u4eff\u771f\u4f18\u5316\u3001\u591a\u667a\u80fd\u4f53\u7cfb\u7edf\u3001\u7fa4\u4f53\u667a\u80fd\u3001\u7edf\u8ba1\u5b66\u4ee5\u53ca\u9057\u4f20\u7b97\u6cd5\u3002\u5728\u8fd0\u7b79\u5b66\u548c\u63a7\u5236\u7406\u8bba\u7814\u7a76\u7684\u8bed\u5883\u4e0b\uff0c\u5f3a\u5316\u5b66\u4e60\u88ab\u79f0\u4f5c\u201c\u8fd1\u4f3c\u52a8\u6001\u89c4\u5212\u201d\uff08approximate dynamic programming\uff0cADP\uff09\u3002\u5728\u6700\u4f18\u63a7\u5236\u7406\u8bba\u4e2d\u4e5f\u6709\u7814\u7a76\u8fd9\u4e2a\u95ee\u9898\uff0c\u867d\u7136\u5927\u90e8\u5206\u7684\u7814\u7a76\u662f\u5173\u4e8e\u6700\u4f18\u89e3\u7684\u5b58\u5728\u548c\u7279\u6027\uff0c\u5e76\u975e\u662f\u5b66\u4e60\u6216\u8005\u8fd1\u4f3c\u65b9\u9762\u3002\u5728\u7ecf\u6d4e\u5b66\u548c\u535a\u5f08\u8bba\u4e2d\uff0c\u5f3a\u5316\u5b66\u4e60\u88ab\u7528\u6765\u89e3\u91ca\u5728\u6709\u9650\u7406\u6027\u7684\u6761\u4ef6\u4e0b\u5982\u4f55\u51fa\u73b0\u5e73\u8861\u3002<\/p>\n<p>\u5728\u673a\u5668\u5b66\u4e60\u95ee\u9898\u4e2d\uff0c\u73af\u5883\u901a\u5e38\u88ab\u62bd\u8c61\u4e3a\u9a6c\u5c14\u53ef\u592b\u51b3\u7b56\u8fc7\u7a0b\uff08Markov decision processes\uff0cMDP\uff09\uff0c\u4e0b\u9762\u5c31\u5c06\u4ecb\u7ecd\u7ecf\u5178\u7684\u9a6c\u5c14\u79d1\u592b\u51b3\u7b56\u8fc7\u7a0b\u53ca\u76f8\u5e94\u7684\u52a8\u6001\u89c4\u5212\u7b97\u6cd5\u3002<\/p>\n<h3>\u9a6c\u5c14\u79d1\u592b\u51b3\u7b56\u8fc7\u7a0b\u4e0e\u5f3a\u5316\u5b66\u4e60\u52a8\u6001\u89c4\u5212\u7b97\u6cd5<\/h3>\n<p>\u9a6c\u5c14\u79d1\u592b\u51b3\u7b56\u8fc7\u7a0b\uff08Markov Decision Process\uff09\u53ef\u4ee5\u8ba4\u4e3a\u662f\u4e00\u4e2a<span class=\"katex-eq\" data-katex-display=\"false\">&lt;S,A,P,r,\\gamma&gt;<\/span>\u7684\u4e94\u5143\u7ec4\uff0c\u5176\u4e2d\uff1a<\/p>\n<ul>\n<li><span class=\"katex-eq\" data-katex-display=\"false\">S<\/span>\u662f\u72b6\u6001\u7684\u96c6\u5408<\/li>\n<li><span class=\"katex-eq\" data-katex-display=\"false\">A<\/span>\u662f\u52a8\u4f5c\u7684\u96c6\u5408<\/li>\n<li><span class=\"katex-eq\" data-katex-display=\"false\">\\gamma<\/span>\u662f\u6298\u6263\u56e0\u5b50<\/li>\n<li><span class=\"katex-eq\" data-katex-display=\"false\">r(s,a)<\/span>\u662f\u5956\u52b1\u51fd\u6570<\/li>\n<li><span class=\"katex-eq\" data-katex-display=\"false\">P(s&#039;|s,a)<\/span>\u662f\u72b6\u6001\u8f6c\u79fb\u51fd\u6570\uff0c\u8868\u793a\u5728\u72b6\u6001s\u6267\u884c\u52a8\u4f5ca\u540e\u8f6c\u79fb\u5230s'\u7684\u6982\u7387\u3002<\/li>\n<\/ul>\n<p>\u5b9a\u4e49\u72b6\u6001\u4ef7\u503c\u51fd\u6570\u4e3a\u5728\u5f53\u524d\u72b6\u6001\u53ca\u7b56\u7565\u4e0b\uff0c\u4f9d\u636e\u7b56\u7565\u6240\u9009\u62e9\u7684\u52a8\u4f5c\u53ef\u4ee5\u83b7\u5f97\u7684\u671f\u671b\u5956\u52b1,\u5e76\u5bf9\u4e4b\u540e\u7684\u72b6\u6001\u8fdb\u884c<span class=\"katex-eq\" data-katex-display=\"false\">\\gamma\\textless1<\/span>\u7684\u6298\u6263\uff0c\u8fd9\u6837\u505a\u7684\u76ee\u7684\u4e3b\u8981\u6709\u4e24\u70b9\uff1a<\/p>\n<ol>\n<li>\u662f\u4e3a\u4e86\u8868\u793a\u5f53\u524d\u72b6\u6001\u7684\u5956\u52b1\u6bd4\u672a\u6765\u72b6\u6001\u8981\u91cd\u8981\u3002<\/li>\n<li>\u662f\u4e3a\u4e86\u786e\u4fdd<strong>\u72b6\u6001\u4ef7\u503c\u51fd\u6570\u6536\u655b<\/strong>\u3002<\/li>\n<\/ol>\n<p>\u6b64\u65f6\u53ef\u4ee5\u5199\u51fa\u72b6\u6001\u4ef7\u503c\u51fd\u6570\uff08<strong>\u8fde\u7eed\u5f62\u5f0f<\/strong>\uff09\uff1a<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/lingbo.online\/wp-content\/uploads\/2023\/09\/pic_34-1.png\" alt=\"\" \/><br \/>\n\u4ee5\u53ca\u52a8\u4f5c\u4ef7\u503c\u51fd\u6570:<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/lingbo.online\/wp-content\/uploads\/2023\/09\/pic_35-1.png\" alt=\"\" \/><\/p>\n<p>\u8fdb\u4e00\u6b65\u53ef\u4ee5\u5199\u51fa\u6700\u4f18\u72b6\u6001\u4ef7\u503c\u51fd\u6570\uff08\u8fd9\u91cc\u7684\u6700\u4f18\u72b6\u6001\u4ef7\u503c\u51fd\u6570\u4e0e\u6700\u4f18\u52a8\u4f5c\u4ef7\u503c\u51fd\u6570\u90fd\u662f\u76f8\u5bf9\u4e8e\u7b56\u7565\u800c\u8a00\u7684\uff0c\u5373\u5728\u8be5\u6700\u4f18\u7b56\u7565\u4e0b\u6240\u5bf9\u5e94\u7684\u72b6\u6001\u4ef7\u503c\u51fd\u6570\uff08\u6216\u52a8\u4f5c\u4ef7\u503c\u51fd\u6570\uff09\u5176\u5728\u4efb\u610f\u72b6\u6001\uff08\u72b6\u6001\u52a8\u4f5c\u5bf9\uff09\u4e0b\u7684\u503c\u90fd\u4e0d\u5c0f\u4e8e\u5176\u4ed6\u7b56\u7565\uff0c\u6b64\u65f6\u5c31\u5c06\u6700\u4f18\u7b56\u7565\u4e0b\u6240\u5bf9\u5e94\u7684\u72b6\u6001\u4ef7\u503c\u51fd\u6570\uff08\u52a8\u4f5c\u4ef7\u503c\u51fd\u6570\uff09\u79f0\u4e3a\u6700\u4f18\u7684\uff09\uff1a<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/lingbo.online\/wp-content\/uploads\/2023\/09\/pic_36-1.png\" alt=\"\" \/><\/p>\n<p>\u4ee5\u53ca\u6700\u4f18\u52a8\u4f5c\u4ef7\u503c\u51fd\u6570\uff1a<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/lingbo.online\/wp-content\/uploads\/2023\/09\/pic_37-1.png\" alt=\"\" \/><\/p>\n<h4>\u7b56\u7565\u63d0\u5347\u5b9a\u7406<\/h4>\n<p>\u7b56\u7565\u63d0\u5347\u5b9a\u7406\u662f\u6307\uff1a\u82e5\u6709\u4e00\u4e2a\u786e\u5b9a\u6027\u7b56\u7565<span class=\"katex-eq\" data-katex-display=\"false\">\\pi&#039;<\/span>,\u5728\u6bcf\u4e2a\u72b6\u6001s\u9009\u62e9\u4ef7\u503c\u6700\u5927\u7684\u52a8\u4f5c\uff1a<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/lingbo.online\/wp-content\/uploads\/2023\/09\/pic_38-1.png\" alt=\"\" \/><\/p>\n<p>\u5373\u5728\u4efb\u610f\u4e00\u4e2a\u72b6\u6001s\u4e0b\u5747\u6709\uff1a<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/lingbo.online\/wp-content\/uploads\/2023\/09\/pic_39.png\" alt=\"\" \/><\/p>\n<p>\u4e8e\u662f\u5728\u4efb\u610f\u72b6\u6001s\u4e0b\u5747\u6709\uff1a<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/lingbo.online\/wp-content\/uploads\/2023\/09\/pic_40.png\" alt=\"\" \/><\/p>\n<p><img decoding=\"async\" src=\"https:\/\/lingbo.online\/wp-content\/uploads\/2023\/09\/pic_41.png\" alt=\"\" \/><\/p>\n<p>\u6839\u636e\u4e0a\u8ff0\u7684\u7b56\u7565\u63d0\u5347\u5b9a\u7406\u53ef\u4ee5\u5f97\u5230\u4e0b\u9762\u7684\u7b56\u7565\u8fed\u4ee3\u7b97\u6cd5\u3002<br \/>\n\u6709\u4e86\u4e0a\u8ff0\u7684\u57fa\u7840\uff0c\u63a5\u4e0b\u6765\u5c06\u4ecb\u7ecd\u57fa\u4e8e\u52a8\u6001\u89c4\u5212\u7684\u5f3a\u5316\u5b66\u4e60\uff0c\u5373\u7b56\u7565\u8fed\u4ee3\u7b97\u6cd5\u548c\u503c\u8fed\u4ee3\u7b97\u6cd5\u3002<\/p>\n<h4>\u7b56\u7565\u8fed\u4ee3\u7b97\u6cd5<\/h4>\n<p>\u7b56\u7565\u8fed\u4ee3\u7b97\u6cd5\u662f<strong>\u5e7f\u4e49\u7b56\u7565\u8fed\u4ee3\uff08generalized policy iteration\uff0cGPI\uff09<\/strong>\u6846\u67b6\u4e0b\u7684\u4e00\u79cd\u7b97\u6cd5\uff0c\u5e7f\u4e49\u7b56\u7565\u8fed\u4ee3\u662f\u6307\u4f7f\u7528<strong>\u7b56\u7565\u8bc4\u4f30<\/strong>\u548c<strong>\u7b56\u7565\u6539\u8fdb<\/strong>\u76f8\u4e92\u4f5c\u7528\u7684\u4e00\u79cd\u7b97\u6cd5\u6846\u67b6\uff0c\u51e0\u4e4e\u6240\u6709\u7684\u5f3a\u5316\u5b66\u4e60\u7b97\u6cd5\u90fd\u53ef\u4ee5\u770b\u6210\u662fGPI\uff08\u5305\u62ec\u7b56\u7565\u8fed\u4ee3\u7b97\u6cd5\u548c\u503c\u8fed\u4ee3\u7b97\u6cd5\uff09\u3002<br \/>\n<img decoding=\"async\" src=\"https:\/\/lingbo.online\/wp-content\/uploads\/2023\/09\/pic_42.png\" alt=\"\" \/><\/p>\n<p>\u7b56\u7565\u8fed\u4ee3\u7b97\u6cd5\u4f2a\u4ee3\u7801\uff1a<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/lingbo.online\/wp-content\/uploads\/2023\/09\/pic_43.jpg\" alt=\"\" \/><br \/>\n\u7b56\u7565\u8fed\u4ee3\u7684\u8fc7\u7a0b\u53ef\u4ee5\u7528\u4e0b\u56fe\u8bf4\u660e\uff1a<br \/>\n<img decoding=\"async\" src=\"https:\/\/lingbo.online\/wp-content\/uploads\/2023\/09\/pic_44.jpg\" alt=\"\" \/><br \/>\n\u53ef\u4ee5\u770b\u5230\u7b56\u7565\u8fed\u4ee3\u7b97\u6cd5\u5c31\u662f\u4e00\u4e2a\u7b56\u7565\u8bc4\u4f30\u548c\u7b56\u7565\u63d0\u5347\u7684\u8fc7\u7a0b\uff0c\u76f4\u5230\u6700\u540e\u6536\u655b\u5230\u6700\u4f18\u7b56\u7565\u3002<\/p>\n<h4>\u503c\u8fed\u4ee3\u7b97\u6cd5<\/h4>\n<p>\u7531\u7b56\u7565\u8fed\u4ee3\u7b97\u6cd5\u7684\u6b65\u9aa4\u4e8c\u53ef\u77e5\uff0c\u6211\u4eec\u9700\u8981\u6267\u884c\u5f88\u591a\u8f6e\u7b56\u7565\u8fed\u4ee3\u624d\u80fd\u4f7f\u7b56\u7565\u7684\u8bc4\u4f30\u6536\u655b\uff0c\u90a3\u4e48\u80fd\u4e0d\u80fd\u7701\u53bb\u8fd9\u4e00\u6b65\u5462\u3002\u8bd5\u60f3\u4e00\u4e0b\u8fd9\u79cd\u60c5\u51b5\uff0c\u867d\u7136\u72b6\u6001\u4ef7\u503c\u51fd\u6570\u6ca1\u6709\u6536\u655b\uff0c\u4f46\u662f\u7b56\u7565\u6240\u9009\u62e9\u7684\u52a8\u4f5c\u5374\u662f\u540c\u4e00\u4e2a\u52a8\u4f5c\uff0c\u5373\u7b56\u7565\u4e0d\u53d1\u751f\u6539\u53d8\u3002\u8fd9\u5c31\u5f15\u5165\u4e86\u63a5\u4e0b\u6765\u8981\u4ecb\u7ecd\u7684\u7b97\u6cd5\uff0c\u503c\u8fed\u4ee3\u7b97\u6cd5\uff0c\u503c\u8fed\u4ee3\u7b97\u6cd5\u76f4\u63a5\u5bf9\u503c\u51fd\u6570\u8fdb\u884c\u64cd\u4f5c\uff0c\u5f53\u7136\u4e5f\u53ef\u4ee5\u770b\u505a\u662f\u53ea\u8fdb\u884c\u4e86\u4e00\u6b21\u7b56\u7565\u8bc4\u4f30\u7684\u7b56\u7565\u8fed\u4ee3\u7b97\u6cd5\uff0c\u5176\u5229\u7528\u7684\u662f\u6700\u4f18bellman\u65b9\u7a0b\uff1a<br \/>\n<img decoding=\"async\" src=\"https:\/\/lingbo.online\/wp-content\/uploads\/2023\/09\/pic_45.png\" alt=\"\" \/><\/p>\n<p>\u5f97\u5230\u6700\u4f18\u72b6\u6001\u4ef7\u503c\u51fd\u6570\u540e\uff0c\u6211\u4eec\u53ea\u9700\u8981\u5229\u7528\uff1a<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/lingbo.online\/wp-content\/uploads\/2023\/09\/pic_46.png\" alt=\"\" \/><br \/>\n\u5373\u53ef\u6062\u590d\u6700\u4f18\u7b56\u7565\uff08\u6ce8\u610f\u5728\u4ef7\u503c\u8fed\u4ee3\u7b97\u6cd5\u4e2d\u5e76\u4e0d\u5b58\u5728\u663e\u5f0f\u7684\u7b56\u7565\uff0c\u4f46\u6211\u4eec\u53ef\u4ee5\u4ece\u6700\u4f18\u72b6\u6001\u4ef7\u503c\u51fd\u6570\u4e2d\u6062\u590d\u6700\u4f18\u7b56\u7565\uff0c\u5176\u6536\u655b\u6027\u5982\u4e0b\uff09\uff0c\u4ef7\u503c\u8fed\u4ee3\u7b97\u6cd5\u4f2a\u4ee3\u7801\u5982\u4e0b\uff1a<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/lingbo.online\/wp-content\/uploads\/2023\/09\/pic_47.jpg\" alt=\"\" \/><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u5f3a\u5316\u5b66\u4e60\u4ecb\u7ecd \u5f3a\u5316\u5b66\u4e60\uff08\u82f1\u8bed\uff1aReinforcement learning\uff0c\u7b80\u79f0RL\uff09\u662f\u673a\u5668\u5b66\u4e60\u4e2d\u7684\u4e00\u4e2a\u9886\u57df\uff0c\u5f3a\u8c03\u5982\u4f55\u57fa\u4e8e\u73af\u5883\u800c\u884c &#8230;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"emotion":"","emotion_color":"","title_style":"","license":"","footnotes":""},"categories":[1],"tags":[],"class_list":["post-444","post","type-post","status-publish","format-standard","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/posts\/444","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/comments?post=444"}],"version-history":[{"count":11,"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/posts\/444\/revisions"}],"predecessor-version":[{"id":477,"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/posts\/444\/revisions\/477"}],"wp:attachment":[{"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/media?parent=444"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/categories?post=444"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/tags?post=444"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}