{"id":2282,"date":"2019-01-07T11:25:47","date_gmt":"2019-01-07T03:25:47","guid":{"rendered":"https:\/\/yanjingang.com\/blog\/?p=2282"},"modified":"2019-01-07T12:05:32","modified_gmt":"2019-01-07T04:05:32","slug":"alphago%e8%ae%ba%e6%96%87%ef%bc%9a%e7%94%a8%e6%b7%b1%e5%ba%a6%e7%a5%9e%e7%bb%8f%e7%bd%91%e7%bb%9c%e5%92%8c%e6%a0%91%e6%90%9c%e7%b4%a2%e5%be%81%e6%9c%8d%e5%9b%b4%e6%a3%8b-mastering-the-game-of-go-with","status":"publish","type":"post","link":"https:\/\/yanjingang.com\/blog\/?p=2282","title":{"rendered":"AlphaGo\u8bba\u6587\u8bd1\u6587\uff1a\u7528\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u548c\u6811\u641c\u7d22\u5f81\u670d\u56f4\u68cb Mastering the game of Go with deep neural networks and tree search"},"content":{"rendered":"

\u524d\u8a00\uff1a<\/h1>\n

\u56f4\u68cb\u7684\u82f1\u6587\u662f the game of Go\uff0c\u6807\u9898\u7ffb\u8bd1\u4e3a\uff1a\u300a\u7528\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u548c\u6811\u641c\u7d22\u5f81\u670d\u56f4\u68cb\u300b\u3002<\/p>\n

\u5728AlphaGo<\/a>\u7684\u5f71\u54cd\u4e4b\u4e0b\uff0c\u5168\u793e\u4f1a\u5bf9\u4eba\u5de5\u667a\u80fd\u7684\u5173\u6ce8\u8fdb\u4e00\u6b65\u63d0\u5347\u30023\u670812\u65e5\uff0cAlphaGo \u7b2c\u4e09\u6b21\u51fb\u8d25\u674e\u4e16\u77f3\u3002\u57283\u670815\u65e5\u603b\u6bd4\u5206\u5b9a\u683c\u4e3a4\uff1a1\uff0c\u968f\u540eAlphaGo\u7684\u56f4\u68cb\u6392\u540d\u4e16\u754c\u6765\u5230\u7b2c\u4e8c\u3002<\/p>\n

\u8bba\u6587\u539f\u6587<\/a><\/p>\n

\u7f16\u8005\u6309\uff1a2014\u5e745\u6708\uff0c\u4eba\u4eec\u8ba4\u4e3a\u81f3\u5c11\u9700\u8981\u5341\u5e74\u7535\u8111\u624d\u80fd\u51fb\u8d25\u804c\u4e1a\u9009\u624b<\/a>\u3002<\/p>\n

\n

<\/a>\u8bd1\u6587<\/h1>\n

\u6807\u9898<\/strong>\uff1a\u7528\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u548c\u6811\u641c\u7d22\u5f81\u670d\u56f4\u68cb<\/p>\n

\u4f5c\u8005<\/strong>\uff1aDavid Silver 1 , Aja Huang 1 , Chris J. Maddison 1 , Arthur Guez 1 , Laurent Sifre 1 , George van den Driessche 1 , Julian Schrittwieser 1 , Ioannis Antonoglou 1 , Veda Panneershelvam 1 , Marc Lanctot 1 , Sander Dieleman 1 , Dominik Grewe 1 , John Nham 2 , Nal Kalchbrenner 1 , Ilya Sutskever 2 , Timothy Lillicrap 1 , Madeleine Leach 1 , Koray Kavukcuoglu 1 , Thore Graepel 1 , Demis Hassabis 1
\n\u4ed6\u4eec\u6765\u81ea Google DeepMind \u82f1\u56fd\u56e2\u961f\uff08\u75281\u8868\u793a\uff09, Google \u603b\u90e8\uff08\u75282\u8868\u793a\uff09
\nDavid Silver , Aja Huang\u662f\u5e76\u5217\u7b2c\u4e00\u4f5c\u8005<\/p>\n

\u6458\u8981<\/strong>\uff1a\u4eba\u4eec\u957f\u4e45\u4ee5\u6765\u8ba4\u4e3a\uff1a\u56f4\u68cb\u5bf9\u4e8e\u4eba\u5de5\u667a\u80fd\u6765\u8bf4\u662f\u6700\u5177\u6709\u6311\u6218\u6027\u7684\u7ecf\u5178\u535a\u5f08\u6e38\u620f\uff0c\u56e0\u4e3a\u5b83\u7684\u5de8\u5927\u7684\u641c\u7d22\u7a7a\u95f4\uff0c\u8bc4\u4f30\u68cb\u5c40\u548c\u8bc4\u4f30\u843d\u5b50\u5730\u70b9\u7684\u96be\u5ea6\u3002\u6211\u4eec\u7ed9\u7535\u8111\u56f4\u68cb\u7a0b\u5e8f\u5f15\u5165\u4e00\u79cd\u65b0\u7684\u65b9\u6cd5\uff0c\u8fd9\u4e2a\u65b9\u6cd5\u4f7f\u7528\u4f30\u503c\u7f51\u7edc\u6765\u8bc4\u4f30\u68cb\u5c40\uff0c\u4ee5\u53ca\u4f7f\u7528\u7b56\u7565\u7f51\u7edc\u6765\u9009\u62e9\u5982\u4f55\u843d\u5b50\u3002\u8fd9\u4e9b\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u88ab\u4e00\u79cd\u65b0\u7684\u7ec4\u5408\u6765\u8bad\u7ec3\uff1a\u4f7f\u7528\u4e86\u4eba\u7c7b\u4e13\u4e1a\u6bd4\u8d5b\u6570\u636e\u7684\u76d1\u7763\u5b66\u4e60<\/a>\uff0c\u4ee5\u53ca\u81ea\u6211\u5bf9\u5f08\u7684\u5f3a\u5316\u5b66\u4e60<\/a>\u3002\u6ca1\u6709\u4f7f\u7528\u4efb\u4f55\u9884\u6d4b\u641c\u7d22<\/a>\u7684\u65b9\u6cd5\uff0c\u795e\u7ecf\u7f51\u7edc\u4e0b\u56f4\u68cb\u8fbe\u5230\u4e86\u6700\u5148\u8fdb\u7684\u8499\u7279\u5361\u6d1b\u6811\u641c\u7d22<\/a>\u7a0b\u5e8f\u7684\u6c34\u51c6\uff0c\u8fd9\u7a0b\u5e8f\u6a21\u62df\u4e86\u6570\u4ee5\u5343\u8ba1\u7684\u81ea\u6211\u5bf9\u5f08\u7684\u968f\u673a\u535a\u5f08\u3002\u6211\u4eec\u540c\u65f6\u4e5f\u5f15\u5165\u4e86\u4e00\u79cd\u65b0\u7684\u641c\u7d22\u7b97\u6cd5\uff0c\u8fd9\u7b97\u6cd5\u628a\u8499\u7279\u5361\u6d1b\u6a21\u62df\u548c\u4f30\u503c\u3001\u7b56\u7565\u7f51\u7edc\u7ed3\u5408\u5728\u4e00\u8d77\u3002\u8fd0\u7528\u4e86\u8fd9\u4e2a\u641c\u7d22\u7b97\u6cd5\uff0c\u6211\u4eec\u7684\u7a0b\u5e8fAlphaGo\u5728\u548c\u5176\u5b83\u56f4\u68cb\u7a0b\u5e8f\u7684\u5bf9\u5f08\u4e2d\u8fbe\u5230\u4e8699.8%\u7684\u80dc\u7387\uff0c\u5e76\u4e14\u4ee55\uff1a0\u7684\u6bd4\u5206\u51fb\u8d25\u4e86\u6b27\u6d32\u51a0\u519b\uff0c\u8fd9\u662f\u53f2\u4e0a\u7b2c\u4e00\u6b21\u8ba1\u7b97\u673a\u7a0b\u5e8f\u5728\u5168\u5c3a\u5bf8\u56f4\u68cb\u4e2d\u51fb\u8d25\u4e00\u4e2a\u4eba\u7c7b\u804c\u4e1a\u68cb\u624b\u3002\u5728\u6b64\u4e4b\u524d\uff0c\u4eba\u4eec\u8ba4\u4e3a\u9700\u8981\u81f3\u5c11\u5341\u5e74\u624d\u4f1a\u8fbe\u6210\u8fd9\u4e2a\u58ee\u4e3e\u3002<\/p>\n

\n

<\/a>\u5f15\u8a00<\/h2>\n

\u6240\u6709\u5b8c\u5168\u4fe1\u606f\u535a\u5f08<\/a>\u90fd\u6709\u4e00\u4e2a\u6700\u4f18\u4f30\u503c\u51fd\u6570v<\/span>\u2217<\/span><\/span><\/span><\/span>(<\/span>s<\/span>)<\/span><\/span><\/span><\/span>\uff0c\u5b83\u5728\u5224\u65ad\u4e86\u6bcf\u4e2a\u68cb\u5c40\u6216\u72b6\u6001 s \u4e4b\u540e\u7684\u535a\u5f08\u7ed3\u679c\u7684\u4f18\u52a3\uff08\u5728\u6240\u6709\u5bf9\u624b\u5b8c\u7f8e\u53d1\u6325\u7684\u60c5\u51b5\u4e0b\uff09\u3002\u89e3\u51b3\u8fd9\u4e9b\u535a\u5f08\u53ef\u4ee5\u901a\u8fc7\u5728\u641c\u7d22\u6811\u4e2d\u9012\u5f52\u8c03\u7528\u6700\u4f18\u4f30\u503c\u51fd\u6570\uff0c\u8fd9\u4e2a\u641c\u7d22\u6811\u5305\u542b\u5927\u7ea6b<\/span>d<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span><\/span>\u79cd\u53ef\u80fd\u7684\u4e0b\u68cb\u5e8f\u5217\uff0c\u5176\u4e2d b \u662f\u535a\u5f08\u7684\u5e7f\u5ea6\uff08\u6bcf\u4e00\u6b21\u4e0b\u68cb\u65f6\u5019\u7684\u5408\u6cd5\u843d\u5b50\u4e2a\u6570\uff09\uff0cd \u662f\u7684\u6df1\u5ea6\uff08\u535a\u5f08\u7684\u6b65\u6570\u957f\u5ea6\uff09\u3002\u5728\u5927\u578b\u535a\u5f08\u4e2d\uff0c\u6bd4\u5982\u56fd\u9645\u8c61\u68cb\uff08b<\/span>\u2248<\/span>35<\/span>,<\/span>d<\/span>\u2248<\/span>80<\/span><\/span><\/span><\/span>\uff09\uff0c\u548c\u7279\u522b\u662f\u56f4\u68cb\uff08b<\/span>\u2248<\/span>250<\/span>,<\/span>d<\/span>\u2248<\/span>150<\/span><\/span><\/span><\/span>\uff09\uff0c\u7a77\u4e3e\u641c\u7d22\u662f\u4e0d\u53ef\u884c\u7684\u7684\uff0c\u4f46\u662f\u6709\u6548\u7684\u641c\u7d22\u7a7a\u95f4\u53ef\u4ee5\u901a\u8fc7\u4e24\u79cd\u901a\u7528\u7684\u539f\u5219\u51cf\u5c11\u3002\u7b2c\u4e00\uff0c\u641c\u7d22\u7684\u6df1\u5ea6\u53ef\u4ee5\u901a\u8fc7\u68cb\u5c40\u8bc4\u4f30\u964d\u4f4e\uff1a\u5728\u72b6\u6001 s \u65f6\u5bf9\u641c\u7d22\u6811\u8fdb\u884c\u526a\u679d\uff0c\u7136\u540e\u7528\u4e00\u4e2a\u8fd1\u4f3c\u4f30\u503c\u51fd\u6570v<\/span>(<\/span>s<\/span>)<\/span>\u2248<\/span>v<\/span>\u2217<\/span><\/span><\/span><\/span>(<\/span>s<\/span>)<\/span>\u00a0<\/span><\/span><\/span>\u00a0<\/span><\/span>\u53d6\u4ee3\u72b6\u6001 s \u4e0b\u9762\u7684\u5b50\u6811\uff0c\u8fd9\u4e2a\u8fd1\u4f3c\u4f30\u503c\u51fd\u6570\u9884\u6d4b\u72b6\u6001 s \u4e4b\u540e\u7684\u5bf9\u5f08\u7ed3\u679c\u3002\u8fd9\u79cd\u65b9\u6cd5\u5df2\u7ecf\u5728\u56fd\u9645\u8c61\u68cb\uff0c\u56fd\u9645\u8df3\u68cb\uff0c\u9ed1\u767d\u68cb\u4e2d\u5f97\u5230\u4e86\u8d85\u8d8a\u4eba\u7c7b\u7684\u4e0b\u68cb\u80fd\u529b\uff0c\u4f46\u662f\u4eba\u4eec\u8ba4\u4e3a\u8fd9\u79cd\u65b9\u6cd5\u5728\u56f4\u68cb\u4e2d\u662f\u96be\u4ee5\u5904\u7406\u7684\uff0c\u56e0\u4e3a\u56f4\u68cb\u7684\u5de8\u5927\u7684\u590d\u6742\u5ea6\u3002\u7b2c\u4e8c\uff0c\u641c\u7d22\u7684\u5e7f\u5ea6\u53ef\u4ee5\u901a\u8fc7\u6765\u81ea\u7b56\u7565\u00a0p<\/span>(<\/span>a<\/span>\u2223<\/span>s<\/span>)<\/span>\u00a0<\/span><\/span><\/span><\/span>\u7684\u91c7\u6837\u52a8\u4f5c\u6765\u964d\u4f4e\uff0c\u8fd9\u4e2a\u7b56\u7565\u662f\u4e00\u4e2a\u5728\u4f4d\u7f6e s \u7684\u53ef\u80fd\u4e0b\u68cb\u8d70\u5b50a \u6982\u7387\u5206\u5e03\u3002\u6bd4\u5982\u8499\u7279\u5361\u6d1b\u8d70\u5b50\u65b9\u6cd5\u641c\u7d22\u5230\u6700\u5927\u6df1\u5ea6\u65f6\u5019\u6839\u672c\u4e0d\u4f7f\u7528\u5206\u6b67\u754c\u5b9a\u6cd5<\/a>\uff0c\u5b83\u4ece\u4e00\u4e2a\u7b56\u7565 p \u4e2d\u91c7\u96c6\u53cc\u65b9\u68cb\u624b\u7684\u4e00\u7cfb\u5217\u4e0b\u68cb\u8d70\u6cd5\u3002\u8ba1\u7b97\u8fd9\u4e9b\u8d70\u5b50\u7684\u5e73\u5747\u6570\u53ef\u4ee5\u4ea7\u751f\u4e00\u4e2a\u6709\u6548\u7684\u68cb\u5c40\u8bc4\u4f30\uff0c\u5728\u897f\u6d0b\u53cc\u9646\u68cb\u620f\u548c\u62fc\u5b57\u6e38\u620f\u4e2d\u83b7\u5f97\u4e86\u8d85\u51fa\u4eba\u7c7b\u7684\u6027\u80fd\u8868\u73b0\uff0c\u5e76\u4e14\u5728\u56f4\u68cb\u4e2d\u8fbe\u5230\u4e86\u4e1a\u4f59\u4f4e\u6bb5\u6c34\u5e73\u3002<\/p>\n

\u8499\u7279\u5361\u6d1b\u6811\u641c\u7d22\u4f7f\u7528\u8499\u7279\u5361\u6d1b<\/a>\u8d70\u5b50\u65b9\u6cd5\uff0c\u8bc4\u4f30\u641c\u7d22\u6811\u4e2d\u6bcf\u4e00\u4e2a\u72b6\u6001\u7684\u4f30\u503c\u3002\u968f\u7740\u6267\u884c\u8d8a\u6765\u8d8a\u591a\u7684\u6a21\u62df\uff0c\u8fd9\u4e2a\u641c\u7d22\u6811\u6210\u957f\u8d8a\u6765\u8d8a\u5927\uff0c\u800c\u4e14\u76f8\u5173\u4f30\u503c\u6108\u53d1\u7cbe\u786e\u3002\u7528\u6765\u9009\u62e9\u4e0b\u68cb\u52a8\u4f5c\u7684\u7b56\u7565\u5728\u641c\u7d22\u7684\u8fc7\u7a0b\u4e2d\u4e5f\u4f1a\u968f\u7740\u65f6\u95f4\u7684\u63a8\u79fb\u800c\u6539\u8fdb\uff0c\u901a\u8fc7\u9009\u62e9\u62e5\u6709\u66f4\u9ad8\u4f30\u503c\u7684\u5b50\u6811\u3002\u6e10\u8fd1\u7684\uff0c\u8fd9\u4e2a\u7b56\u7565\u6536\u655b\u5230\u4e00\u4e2a\u6700\u4f18\u4e0b\u6cd5\uff0c\u7136\u540e\u8bc4\u4f30\u6536\u655b\u5230\u6700\u4f18\u4f30\u503c\u51fd\u6570\u3002\u76ee\u524d\u6700\u5f3a\u7684\u56f4\u68cb\u7a0b\u5e8f\u662f\u57fa\u4e8e\u8499\u7279\u5361\u6d1b\u6811\u641c\u7d22\u7684\uff0c\u5e76\u4e14\u53d7\u5230\u4e86\u7b56\u7565\u7684\u589e\u5f3a\uff0c\u8fd9\u4e2a\u7b56\u7565\u88ab\u4eba\u8bad\u7ec3\u7528\u6765\u9884\u6d4b\u4e13\u5bb6\u68cb\u624b\u7684\u4e0b\u6cd5\u3002\u8fd9\u4e9b\u7b56\u7565\u7528\u6765\u7f29\u7a84\u641c\u7d22\u7a7a\u95f4\u5230\u4e00\u675f\u9ad8\u53ef\u80fd\u6027\u4e0b\u68cb\u52a8\u4f5c\uff0c\u548c\u7528\u6765\u5728\u8d70\u5b50\u4e2d\u91c7\u96c6\u4e0b\u6cd5\u52a8\u4f5c\u3002\u8fd9\u4e2a\u65b9\u6cd5\u5df2\u7ecf\u8fbe\u5230\u4e86\u4e1a\u4f59\u9ad8\u624b\u7684\u7ea7\u522b\u3002\u7136\u800c\uff0c\u5148\u524d\u7684\u5de5\u4f5c\u5df2\u7ecf\u53d7\u5230\u4e86\u80a4\u6d45\u7b56\u7565\u7684\u9650\u5236\u6216\u57fa\u4e8e\u8f93\u5165\u7684\u7ebf\u6027\u7ec4\u5408\u7684\u4f30\u503c\u51fd\u6570\u7684\u9650\u5236\u3002<\/p>\n

\u6700\u8fd1\uff0c\u6df1\u5ea6\u5377\u79ef\u795e\u7ecf\u7f51\u7edc<\/a>\u5df2\u7ecf\u5728\u8ba1\u7b97\u673a\u89c6\u89c9\u4e2d\u8fbe\u5230\u4e86\u7a7a\u524d\u7684\u6027\u80fd\uff1a\u6bd4\u5982\u56fe\u50cf\u5206\u7c7b\uff0c\u4eba\u8138\u8bc6\u522b\uff0c\u548c\u73a9\u96c5\u8fbe\u5229<\/a>\u7684\u6e38\u620f\u3002\u5b83\u4eec\u4f7f\u7528\u5f88\u591a\u5c42\u7684\u795e\u7ecf\u7f51\u7edc\uff0c\u5c42\u4e0e\u5c42\u4e4b\u95f4\u50cf\u74e6\u7247\u91cd\u53e0\u6392\u5217\u5728\u4e00\u8d77\uff0c\u7528\u6765\u6784\u5efa\u56fe\u7247\u7684\u6108\u53d1\u62bd\u8c61\u7684\u5c40\u90e8\u4ee3\u8868\u3002\u6211\u4eec\u4e3a\u56f4\u68cb\u7a0b\u5e8f\u90e8\u7f72\u4e86\u7c7b\u4f3c\u7684\u4f53\u7cfb\u67b6\u6784\u3002\u6211\u4eec\u7ed9\u7a0b\u5e8f\u4f20\u5165\u4e86\u4e00\u4e2a19*19\u5927\u5c0f\u68cb\u5c40\u7684\u56fe\u7247\uff0c\u7136\u540e\u4f7f\u7528\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u6765\u6784\u5efa\u4e00\u4e2a\u4f4d\u7f6e\u7684\u4ee3\u8868\u3002\u6211\u4eec\u4f7f\u7528\u8fd9\u4e9b\u795e\u7ecf\u7f51\u7edc\u6765\u964d\u4f4e\u641c\u7d22\u6811\u7684\u6709\u6548\u7684\u6df1\u5ea6\u548c\u5e7f\u5ea6\uff1a\u901a\u8fc7\u4f30\u503c\u7f51\u7edc\u6765\u8bc4\u4f30\u68cb\u5c40\uff0c\u548c\u4f7f\u7528\u7b56\u7565\u7f51\u7edc\u6765\u535a\u5f08\u53d6\u6837\u3002<\/p>\n

\u6211\u4eec\u4f7f\u7528\u4e00\u4e2a\u5305\u542b\u591a\u4e2a\u4e0d\u540c\u9636\u6bb5\u7684\u673a\u5668\u5b66\u4e60\u65b9\u6cd5\u7684\u7ba1\u9053<\/a>\u6765\u8bad\u7ec3\u795e\u7ecf\u7f51\u7edc\u3002\u6211\u4eec\u5f00\u59cb\u4f7f\u7528\u4e00\u4e2a\u76d1\u7763\u5b66\u4e60\uff08SL\uff09\u7b56\u7565\u7f51\u7edc\u00a0p<\/span>\u03b4<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span><\/span>\uff0c\u5b83\u76f4\u63a5\u6765\u81ea\u4eba\u7c7b\u4e13\u5bb6\u7684\u4e0b\u68cb\u3002\u8fd9\u63d0\u4f9b\u4e86\u5feb\u901f\u9ad8\u6548\u7684\u5b66\u4e60\u66f4\u65b0\uff0c\u62e5\u6709\u5feb\u901f\u7684\u53cd\u9988\u548c\u9ad8\u8d28\u91cf\u7684\u68af\u5ea6\u3002\u548c\u5411\u524d\u7684\u5de5\u4f5c\u7c7b\u4f3c\uff0c\u6211\u4eec\u540c\u65f6\u4e5f\u8bad\u7ec3\u4e86\u4e00\u4e2a\u53ef\u4ee5\u8fc5\u901f\u4ece\u8d70\u5b50\u4e2d\u53d6\u6837\u7684\u5feb\u901f\u7b56\u7565\u00a0p<\/span>\u03c0<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span>\u00a0<\/span><\/span>\u3002\u5176\u6b21\uff0c\u6211\u4eec\u8bad\u7ec3\u4e86\u4e00\u4e2a\u5f3a\u5316\u5b66\u4e60\uff08RL\uff09\u7b56\u7565\u7f51\u7edc\uff0cp<\/span>p<\/span><\/span><\/span><\/span><\/span><\/span><\/span>\uff0c\u5b83\u901a\u8fc7\u4f18\u5316\u81ea\u6211\u5bf9\u5f08\u7684\u6700\u7ec8\u7ed3\u5c40\u6765\u63d0\u5347 SL\u7b56\u7565\u7f51\u7edc\u3002\u8fd9\u8c03\u6574\u7b56\u7565\u7f51\u7edc\u671d\u5411\u8d62\u68cb\u7684\u6b63\u786e\u76ee\u6807\u53d1\u5c55\uff0c\u800c\u4e0d\u662f\u6700\u5927\u5316\u63d0\u9ad8\u9884\u6d4b\u7cbe\u5ea6\u3002\u6700\u540e\uff0c\u6211\u4eec\u8bad\u7ec3\u4e86\u4e00\u4e2a\u4f30\u503c\u7f51\u7edcv<\/span>\u03b8<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span>\u00a0<\/span><\/span>\uff0c\u5b83\u9884\u6d4b\u535a\u5f08\u7684\u8d62\u8005\uff0c\u901a\u8fc7\u548cRL\u7b56\u7565\u7f51\u7edc\u548c\u81ea\u5df1\u5bf9\u5f08\u3002\u6211\u4eec\u7684AlphaGo\u7a0b\u5e8f\u6709\u6548\u7684\u628a\u7b56\u7565\u7f51\u7edc\u3001\u4f30\u503c\u7f51\u7edc\uff0c\u548c\u8499\u7279\u5361\u6d1b\u641c\u7d22\u6811\u7ed3\u5408\u5728\u4e00\u8d77\u3002<\/p>\n

\n

<\/a>1 \u7b56\u7565\u7f51\u7edc\u7684\u76d1\u7763\u5b66\u4e60<\/h2>\n

\u5728\u8bad\u7ec3\u7ba1\u9053\u7684\u7b2c\u4e00\u9636\u6bb5\uff0c\u6211\u4eec\u5728\u5148\u524d\u5de5\u4f5c\u7684\u57fa\u7840\u4e0a\uff0c\u4f7f\u7528\u4e86\u76d1\u7763\u5b66\u4e60\u6765\u9884\u6d4b\u4eba\u7c7b\u4e13\u5bb6\u4e0b\u56f4\u68cb\u3002\u76d1\u7763\u5b66\u4e60\uff08SL\uff09\u7b56\u7565\u7f51\u7edcp<\/span>\u03b4<\/span><\/span><\/span><\/span>(<\/span>a<\/span>\u2223<\/span>s<\/span>)<\/span>\u00a0<\/span><\/span><\/span>\u00a0<\/span><\/span>\u5728\u91cd\u91cf\u03b4<\/span><\/span><\/span><\/span>\u7684\u5377\u79ef\u5c42\u548c\u975e\u7ebf\u6027\u7684\u6574\u6d41\u5668\u4e2d\u66ff\u6362\u3002\u7b56\u7565\u7f51\u7edc\u7684\u8f93\u5165 s \u662f\u4e00\u4e2a\u68cb\u5c40\u72b6\u6001\u7684\u7b80\u5355\u4ee3\u8868\uff08\u5982\u6269\u5c55\u6570\u636e\u88682\uff09\u3002\u7b56\u7565\u7f51\u7edc\u4f7f\u7528\u4e86\u968f\u673a\u53d6\u6837\u72b6\u6001-\u52a8\u4f5c\u5bf9\uff08s\uff0ca\uff09\uff0c\u4f7f\u7528\u4e86\u968f\u673a\u68af\u5ea6\u9012\u589e\u6765\u6700\u5927\u5316\u4eba\u7c7b\u5728\u72b6\u6001 s \u9009\u62e9\u4e0b\u68cb\u8d70\u5b50 a \u7684\u53ef\u80fd\u6027\u3002
\n\"\"<\/a>
\n\u6211\u4eec\u7528KGS<\/a>\u56f4\u68cb\u670d\u52a1\u5668\u76843\u5343\u4e07\u4e2a\u68cb\u5c40\uff0c\u8bad\u7ec3\u4e8613\u5c42\u7684\u7b56\u7565\u7f51\u7edc\uff08\u6211\u4eec\u79f0\u4e4b\u4e3aSL \u7b56\u7565\u7f51\u7edc\uff09\u3002\u5728\u8f93\u5165\u7559\u5b58\u6d4b\u8bd5\u6570\u636e\u7684\u6240\u53d7\u7279\u5f81\u7684\u65f6\u5019\uff0c\u8fd9\u4e2a\u7f51\u7edc\u9884\u6d4b\u4eba\u7c7b\u4e13\u5bb6\u4e0b\u68cb\u7684\u7cbe\u51c6\u7684\u8fbe\u5230\u4e8657%\uff0c\u800c\u4e14\u5728\u4ec5\u4ec5\u4f7f\u7528\u539f\u59cb\u68cb\u5c40\u548c\u4e0b\u68cb\u8bb0\u5f55\u7684\u65f6\u5019\uff0c\u7cbe\u5ea6\u8fbe\u5230\u4e8655.7%\u3002\u4e0e\u4e4b\u76f8\u6bd4\uff0c\u622a\u81f3\u5230\u672c\u7bc7\u6587\u8bba\u63d0\u4ea4\uff082015\u5e74\uff09\uff0c\u5176\u4ed6\u7814\u7a76\u56e2\u961f\u7684\u6700\u5148\u8fdb\u7684\u7cbe\u5ea6\u662f44.4%\uff08\u5168\u90e8\u7ed3\u679c\u5728\u6269\u5c55\u6570\u636e\u88683\uff09\u3002\u5728\u7cbe\u786e\u5ea6\u65b9\u9762\u7684\u5c0f\u63d0\u5347\u4f1a\u5f15\u8d77\u4e0b\u68cb\u80fd\u529b\u7684\u5f88\u5927\u63d0\u5347\uff08\u56fe\u72472\uff0ca\uff09\uff1b\u66f4\u5927\u7684\u795e\u7ecf\u7f51\u7edc\u62e5\u6709\u66f4\u9ad8\u7684\u7cbe\u786e\u5ea6\uff0c\u4f46\u662f\u5728\u641c\u7d22\u8fc7\u7a0b\u4e2d\u8bc4\u4f30\u901f\u5ea6\u66f4\u6162\u3002\u6211\u4eec\u4e5f\u8bad\u7ec3\u4e86\u4e00\u4e2a\u66f4\u5feb\u7684\u4f46\u662f\u7cbe\u786e\u5ea6\u66f4\u4f4e\u7684\u8d70\u5b50\u7b56\u7565p<\/span>\u03c0<\/span><\/span><\/span><\/span>(<\/span>a<\/span>\u2223<\/span>s<\/span>)<\/span>\u00a0<\/span><\/span><\/span><\/span>\uff0c\u5b83\u4f7f\u7528\u4e86\u4e00\u4e2a\u6743\u91cd\u4e3a\u03c0<\/span>\u00a0<\/span><\/span><\/span>\u03c0\u00a0<\/span><\/span>\u7684\u5c0f\u578b\u6a21\u5f0f\u7279\u5f81\u7684\u7ebf\u6027softmax\u3002\u5b83\u8fbe\u5230\u4e8624.2%\u7684\u7cbe\u786e\u5ea6\uff0c\u6bcf\u9009\u62e9\u4e0b\u4e00\u6b65\u68cb\u53ea\u75282\u5fae\u79d2\uff0c\u4e0e\u4e4b\u76f8\u6bd4\uff0c\u7b56\u7565\u7f51\u7edc\u9700\u89813\u6beb\u79d2\u3002<\/p>\n

\"\"<\/a>
\n\u56fe1\uff1a\u795e\u7ecf\u7f51\u7edc\u8bad\u7ec3\u7ba1\u9053\u548c\u4f53\u7cfb\u7ed3\u6784<\/strong>\u3002a<\/strong>\uff1a\u5728\u4e00\u4e2a\u68cb\u5c40\u6570\u636e\u96c6\u5408\u4e2d\uff0c\u8bad\u7ec3\u4e00\u4e2a\u5feb\u901f\u8d70\u5b50\u7b56\u7565p<\/span>\u03c0<\/span><\/span><\/span><\/span><\/span><\/span><\/span>\u548c\u76d1\u7763\u5b66\u4e60\uff08SL\uff09\u7b56\u7565\u7f51\u7edcp<\/span>\u03b4<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span><\/span>\u7528\u6765\u9884\u6d4b\u4eba\u7c7b\u4e13\u5bb6\u4e0b\u68cb\u3002\u4e00\u4e2a\u5f3a\u5316\u5b66\u4e60\uff08RL\uff09\u7b56\u7565\u7f51\u7edcp<\/span>\u03c1<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span><\/span>\u7531SL\u7b56\u7565\u7f51\u7edc\u521d\u59cb\u5316\uff0c\u7136\u540e\u7531\u7b56\u7565\u68af\u5ea6\u5b66\u4e60\u8fdb\u884c\u63d0\u9ad8\u3002\u548c\u5148\u524d\u7248\u672c\u7684\u7b56\u7565\u7f51\u7edc\u76f8\u6bd4\uff0c\u6700\u5927\u5316\u7ed3\u5c40\uff08\u6bd4\u5982\u8d62\u66f4\u591a\u7684\u535a\u5f08\uff09\u3002\u4e00\u4e2a\u65b0\u7684\u6570\u636e\u96c6\u5408\u4ea7\u751f\u4e86\uff0c\u901a\u8fc7\u81ea\u6211\u5bf9\u5f08\u7ed3\u5408RL\u7b56\u7565\u7f51\u7edc\u3002\u6700\u7ec8\u901a\u8fc7\u56de\u5f52\u8bad\u7ec3\uff0c\u4ea7\u751f\u4e00\u4e2a\u4f30\u503c\u7f51\u7edcv<\/span>\u03b8<\/span><\/span><\/span><\/span><\/span><\/span>\u00a0<\/span><\/span>\uff0c\u7528\u6765\u5728\u81ea\u6211\u5bf9\u5f08\u7684\u6570\u636e\u96c6\u5408\u4e2d\u9884\u6d4b\u671f\u5f85\u7684\u7ed3\u5c40\uff08\u6bd4\u5982\u5f53\u524d\u68cb\u624b\u662f\u5426\u80fd\u8d62\uff09\u3002b<\/strong>\uff1aAlphaGo\u4f7f\u7528\u7684\u795e\u7ecf\u7f51\u7edc\u4f53\u7cfb\u67b6\u6784\u7684\u539f\u7406\u56fe\u4ee3\u8868\u3002\u7b56\u7565\u7f51\u7edc\u628a\u68cb\u5c40\u72b6\u6001 s \u5f53\u4f5c\u8f93\u5165\u7684\u4ee3\u8868\uff0c\u7b56\u7565\u7f51\u7edc\u628a s \u4f20\u8f93\u901a\u8fc7\u5f88\u591a\u5377\u79ef\u5c42\uff08\u8fd9\u4e9b\u5377\u79ef\u5c42\u662f\u53c2\u6570\u4e3a\u03b4<\/span>\u00a0<\/span><\/span><\/span>\u00a0<\/span><\/span>\u7684SL\u7b56\u7565\u7f51\u7edc\u6216\u8005\u53c2\u6570\u4e3a\u03c1<\/span>\u00a0<\/span><\/span><\/span>\u03c1\u00a0<\/span><\/span>\u7684RL\u7b56\u7565\u7f51\u7edc\uff09\uff0c\u7136\u540e\u8f93\u51fa\u4e00\u4e2a\u5173\u4e8e\u4e0b\u68cb\u52a8\u4f5c a \u7684\u6982\u7387\u5206\u5e03\u00a0p<\/span>\u03b4<\/span><\/span><\/span><\/span>(<\/span>a<\/span>\u2223<\/span>s<\/span>)<\/span>\u00a0<\/span>o<\/span>r<\/span>\u00a0<\/span>p<\/span>\u03c1<\/span><\/span><\/span><\/span>(<\/span>a<\/span>\u2223<\/span>s<\/span>)<\/span>\u00a0<\/span><\/span><\/span><\/span>\uff0c\u7528\u4e00\u4e2a\u68cb\u76d8\u7684\u6982\u7387\u5730\u56fe\u6765\u8868\u793a\u3002\u4f30\u503c\u7f51\u7edc\u7c7b\u4f3c\u7684\u4f7f\u7528\u4e86\u5f88\u591a\u53c2\u6570\u03b8<\/span>\u00a0<\/span><\/span><\/span><\/span>\u7684\u5377\u79ef\u5c42\uff0c\u4f46\u662f\u8f93\u51fa\u4e00\u4e2a\u6807\u91cf\u503cv<\/span>\u03b8<\/span><\/span><\/span><\/span>(<\/span>s<\/span><\/span>\u2032<\/span><\/span><\/span><\/span><\/span>)<\/span>\u00a0<\/span><\/span><\/span><\/span>\u7528\u6765\u9884\u6d4b\u68cb\u5c40\u72b6\u6001\u00a0s<\/span><\/span>\u2032<\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span>\u540e\u7684\u7ed3\u5c40\u3002<\/p>\n

\"\"<\/a>
\n\u56fe2\uff1a\u7b56\u7565\u7f51\u7edc\u548c\u4f30\u503c\u7f51\u7edc\u7684\u80fd\u529b\u548c\u7cbe\u786e\u5ea6\u3002a<\/strong>\u56fe\u663e\u793a\u4e86\u7b56\u7565\u7f51\u7edc\u7684\u4e0b\u68cb\u80fd\u529b\u968f\u7740\u5b83\u4eec\u7684\u8bad\u7ec3\u7cbe\u786e\u5ea6\u7684\u51fd\u6570\u3002\u62e5\u6709128\uff0c192\uff0c256\uff0c384\u5377\u79ef\u8fc7\u6ee4\u6bcf\u5c42\u7684\u7b56\u7565\u7f51\u7edc\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u5f97\u5230\u5468\u671f\u6027\u7684\u8bc4\u4f30\uff1b\u8fd9\u4e2a\u56fe\u663e\u793a\u4e86AlphaGo\u4f7f\u7528\u4e0d\u540c\u7b56\u7565\u7f51\u7edc\u7684\u8d62\u68cb\u6982\u7387\u968f\u7740\u7684\u4e0d\u540c\u7cbe\u786e\u5ea6\u7248\u672c\u7684AlphaGo\u7684\u53d8\u5316\u3002b<\/strong>\uff1a\u4f30\u503c\u7f51\u7edc\u548c\u4e0d\u540c\u7b56\u7565\u7f51\u7edc\u7684\u8bc4\u4f30\u5bf9\u6bd4\u3002\u68cb\u5c40\u548c\u7ed3\u5c40\u662f\u4ece\u4eba\u7c7b\u4e13\u5bb6\u535a\u5f08\u5bf9\u5c40\u4e2d\u91c7\u6837\u7684\u3002\u6bcf\u4e00\u4e2a\u68cb\u5c40\u90fd\u662f\u7531\u4e00\u4e2a\u5355\u72ec\u7684\u5411\u524d\u4f20\u9012\u7684\u4f30\u503c\u7f51\u7edcv<\/span>\u03b8<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span>\u00a0<\/span><\/span>\u8bc4\u4f30\u7684\uff0c\u6216\u8005100\u4e2a\u8d70\u5b50\u7684\u5e73\u5747\u503c\uff0c\u8fd9\u4e9b\u8d70\u5b50\u662f\u7531\u7edf\u4e00\u968f\u673a\u8d70\u5b50\uff0c\u6216\u5feb\u901f\u8d70\u5b50\u7b56\u7565p<\/span>\u03c0<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span>\u00a0<\/span><\/span>\uff0c\u6216 SL \u7b56\u7565\u7f51\u7edcp<\/span>\u03b4<\/span><\/span><\/span><\/span><\/span><\/span>\u00a0<\/span><\/span>\uff0c\u6216 RL \u7b56\u7565\u7f51\u7edcp<\/span>\u03c1<\/span><\/span><\/span><\/span><\/span><\/span>\u00a0<\/span><\/span>\u3002\u56fe\u4e2d\uff0c\u9884\u6d4b\u4f30\u503c\u548c\u535a\u5f08\u5b9e\u9645\u7ed3\u5c40\u4e4b\u95f4\u7684\u5e73\u5747\u65b9\u5dee\u968f\u7740\u535a\u5f08\u7684\u8fdb\u884c\u9636\u6bb5\uff08\u535a\u5f08\u603b\u5171\u4e0b\u4e86\u591a\u5c11\u6b65\uff09\u7684\u53d8\u5316\u800c\u53d8\u5316\u3002<\/p>\n

\n

<\/a>2 \u7b56\u7565\u7f51\u7edc\u7684\u5f3a\u5316\u5b66\u4e60<\/h2>\n

\u8bad\u7ec3\u7ba1\u9053\u7b2c\u4e8c\u9636\u6bb5\u7684\u76ee\u6807\u662f\u901a\u8fc7\u7b56\u7565\u68af\u5ea6\u5f3a\u5316\u5b66\u4e60\uff08RL\uff09\u6765\u63d0\u9ad8\u7b56\u7565\u7f51\u7edc\u3002\u5f3a\u5316\u5b66\u4e60\u7b56\u7565\u7f51\u7edcp<\/span>\u03c1<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span><\/span>\u5728\u7ed3\u6784\u4e0a\u548c SL\u7b56\u7565\u7f51\u7edc\u662f\u4e00\u6837\u7684\uff0c\u6743\u91cd\u03c1<\/span><\/span><\/span><\/span>\u521d\u59cb\u503c\u4e5f\u662f\u4e00\u6837\u7684\uff0c\u03c1<\/span>=<\/span>\u03b4<\/span>\u00a0<\/span><\/span><\/span><\/span>\u3002\u6211\u4eec\u5728\u5f53\u524d\u7684\u7b56\u7565\u7f51\u7edc\u548c\u968f\u673a\u9009\u62e9\u67d0\u5148\u524d\u4e00\u6b21\u8fed\u4ee3\u7684\u7b56\u7565\u7f51\u7edc\u4e4b\u95f4\u535a\u5f08\u3002\u4ece\u4e00\u4e2a\u5bf9\u624b\u7684\u5019\u9009\u6c60\u4e2d\u968f\u673a\u9009\u62e9\uff0c\u53ef\u4ee5\u7a33\u5b9a\u8bad\u7ec3\u8fc7\u7a0b\uff0c\u9632\u6b62\u8fc7\u5ea6\u62df\u5408\u4e8e\u5f53\u524d\u7684\u7b56\u7565\u3002\u6211\u4eec\u4f7f\u7528\u4e00\u4e2a\u5956\u52b1\u51fd\u6570 r(s)\uff0c\u5bf9\u4e8e\u6240\u6709\u975e\u7ec8\u7aef\u7684\u6b65\u9aa4 t < T\uff0c\u5b83\u7684\u503c\u7b49\u4e8e\u96f6\u3002\u4ece\u5f53\u524d\u68cb\u624b\u5728\u6b65\u9aa4 t \u7684\u89d2\u5ea6\u6765\u8bb2\uff0c\u7ed3\u679c\u00a0z<\/span>t<\/span><\/span><\/span><\/span>=<\/span>\u00b1<\/span>r<\/span>(<\/span>s<\/span>T<\/span><\/span><\/span><\/span>)<\/span>\u00a0<\/span><\/span><\/span><\/span>\u662f\u5728\u535a\u5f08\u7ed3\u675f\u65f6\u5019\u7684\u7ec8\u7aef\u5956\u52b1\uff0c\u5982\u679c\u8d62\u68cb\uff0c\u7ed3\u679c\u7b49\u4e8e +1\uff0c\u5982\u679c\u8f93\u68cb\uff0c\u7ed3\u679c\u7b49\u4e8e -1\u3002\u7136\u540e\u6743\u91cd\u5728\u6bcf\u4e00\u4e2a\u6b65\u9aa4 t \u66f4\u65b0\uff1a\u671d\u5411\u6700\u5927\u5316\u9884\u671f\u7ed3\u679c\u7684\u65b9\u5411\u968f\u673a\u68af\u5ea6\u9012\u589e
\n\"\"<\/a>
\n\u6211\u4eec\u5728\u535a\u5f08\u8fc7\u7a0b\u4e2d\u8bc4\u4f30 RL\u7b56\u7565\u7f51\u7edc\u7684\u6027\u80fd\u8868\u73b0\uff0c\u4ece\u8f93\u51fa\u7684\u4e0b\u68cb\u52a8\u4f5c\u7684\u6982\u7387\u5206\u5e03\uff0c\u5bf9\u6bcf\u4e00\u4e0b\u68cb\u52a8\u4f5ca<\/span>t<\/span><\/span><\/span><\/span>\u223c<\/span>p<\/span>p<\/span><\/span><\/span><\/span>(<\/span>\u22c5<\/span>\u2223<\/span>s<\/span>t<\/span><\/span><\/span><\/span>)<\/span>\u00a0<\/span><\/span><\/span><\/span>\u8fdb\u884c\u53d6\u6837\u3002\u6211\u4eec\u81ea\u5df1\u9762\u5bf9\u9762\u535a\u5f08\uff0cRL\u7b56\u7565\u7f51\u7edc\u5bf9 SL\u7b56\u7565\u7f51\u7edc\u7684\u80dc\u7387\u9ad8\u4e8e80%\u3002\u6211\u4eec\u4e5f\u6d4b\u8bd5\u4e86\u548c\u6700\u5f3a\u7684\u5f00\u6e90\u56f4\u68cb\u8f6f\u4ef6 Pachi \u5bf9\u5f08\uff0c\u5b83\u662f\u4e00\u4e2a\u968f\u673a\u7684\u8499\u7279\u5361\u6d1b\u641c\u7d22\u7a0b\u5e8f\uff0c\u5728KGS\u4e2d\u8fbe\u5230\u4e1a\u4f592\u6bb5\u3002\u5728\u6ca1\u6709\u4f7f\u7528\u4efb\u4f55\u641c\u7d22\u7684\u60c5\u51b5\u4e0b\uff0cRL\u7b56\u7565\u7f51\u7edc\u5bf9 Pachi\u7684\u80dc\u7387\u8fbe\u5230\u4e8685%\u3002\u4e0e\u4e4b\u76f8\u6bd4\uff0c\u4e4b\u524d\u7684\u6700\u5148\u8fdb\u7684\u4ec5\u4ec5\u57fa\u4e8e\u76d1\u7763\u5b66\u4e60\u7684\u5377\u79ef\u7f51\u7edc\uff0c\u5bf9 Pachi\u7684\u80dc\u7387\u4ec5\u53ea\u670911%\uff0c\u5bf9\u7a0d\u5f31\u7684\u7a0b\u5e8f Fuego\u7684\u80dc\u7387\u662f12%\u3002<\/p>\n

\n

<\/a>3 \u4f30\u503c\u7f51\u7edc\u7684\u5f3a\u5316\u5b66\u4e60<\/h2>\n

\u8bad\u7ec3\u7ba1\u9053\u7684\u6700\u540e\u4e00\u4e2a\u9636\u6bb5\u5173\u6ce8\u4e8e\u68cb\u5c40\u8bc4\u4f30\uff0c\u8bc4\u4f30\u4e00\u4e2a\u4f30\u503c\u51fd\u6570\u00a0v<\/span>p<\/span><\/span><\/span><\/span>(<\/span>s<\/span>)<\/span>\u00a0<\/span><\/span><\/span><\/span>\uff0c\u5b83\u9884\u6d4b\u4ece\u68cb\u5c40\u72b6\u6001 s \u5f00\u59cb\uff0c\u535a\u5f08\u53cc\u65b9\u90fd\u6309\u7167\u7b56\u7565\u7f51\u7edc p \u4e0b\u68cb\u7684\u7ed3\u5c40\uff0c
\n\"\"<\/a>
\n\u7406\u60f3\u60c5\u51b5\u4e0b\uff0c\u6211\u4eec\u671f\u671b\u77e5\u9053\u5728\u5b8c\u7f8e\u4e0b\u6cd5v<\/span>\u2217<\/span><\/span><\/span><\/span>(<\/span>s<\/span>)<\/span><\/span><\/span><\/span>\u60c5\u51b5\u4e0b\u7684\u6700\u4f18\u503c\uff1b\u7136\u800c\u5728\u73b0\u5b9e\u4e2d\uff0c\u6211\u4eec\u4f7f\u7528 RL\u7b56\u7565\u7f51\u7edc\uff0c\u6765\u8bc4\u4f30\u4f30\u503c\u51fd\u6570v<\/span>P<\/span>p<\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span>\uff0c\u4f5c\u4e3a\u6211\u4eec\u7684\u6700\u4f73\u7b56\u7565\u3002\u6211\u4eec\u4f7f\u7528\u6743\u91cd\u662f\u03b8<\/span>\u00a0<\/span><\/span><\/span><\/span>\u7684\u4f30\u503c\u7f51\u7edc\u00a0v<\/span>\u03b8<\/span><\/span><\/span><\/span>(<\/span>s<\/span>)<\/span><\/span><\/span><\/span>\u6765\u903c\u8fd1\u4f30\u503c\u51fd\u6570\uff0cv<\/span>\u03b8<\/span><\/span><\/span><\/span>(<\/span>s<\/span>)<\/span>\u2248<\/span>v<\/span>P<\/span>p<\/span><\/span><\/span><\/span><\/span><\/span><\/span>\u2248<\/span>v<\/span>\u2217<\/span><\/span><\/span><\/span>(<\/span>s<\/span>)<\/span><\/span><\/span><\/span>\u3002\u8fd9\u4e2a\u795e\u7ecf\u7f51\u7edc\u548c\u7b56\u7565\u7f51\u7edc\u62e5\u6709\u8fd1\u4f3c\u7684\u4f53\u7cfb\u7ed3\u6784\uff0c\u4f46\u662f\u8f93\u51fa\u4e00\u4e2a\u5355\u4e00\u7684\u9884\u6d4b\uff0c\u800c\u4e0d\u662f\u4e00\u4e2a\u6982\u7387\u5206\u5e03\u3002\u6211\u4eec\u901a\u8fc7\u56de\u5f52\u5230\u72b6\u6001-\u7ed3\u679c\u5bf9\uff08s, z\uff09\u6765\u8bad\u7ec3\u4f30\u503c\u7f51\u7edc\u7684\u6743\u91cd\uff0c\u4f7f\u7528\u4e86\u968f\u673a\u68af\u5ea6\u9012\u51cf\uff0c\u6700\u5c0f\u5316\u9884\u6d4b\u4f30\u503cv<\/span>\u03b8<\/span><\/span><\/span><\/span>(<\/span>s<\/span>)<\/span><\/span><\/span><\/span>\u548c\u76f8\u5e94\u7684\u7ed3\u5c40 z \u4e4b\u95f4\u7684\u5e73\u5747\u65b9\u5dee\uff08MSE\uff09\u3002
\n\"\"<\/a>
\n\u8fd9\u4e2a\u5929\u771f\u7684\u4ece\u62e5\u6709\u5b8c\u6574\u5bf9\u5f08\u7684\u6570\u636e\u6765\u9884\u6d4b\u535a\u5f08\u7ed3\u5c40\u7684\u65b9\u6cd5\u5bfc\u81f4\u8fc7\u5ea6\u62df\u5408\u3002\u95ee\u9898\u5728\u4e8e\uff0c\u8fde\u7eed\u7684\u68cb\u5c40\u4e4b\u95f4\u7684\u8054\u7cfb\u5341\u5206\u5f3a\u5927\uff0c\u548c\u4ec5\u5355\u72ec\u4e0b\u4e00\u6b65\u68cb\u6709\u5dee\u8ddd\uff0c\u4f46\u662f\u56de\u5f52\u76ee\u6807\u548c\u6574\u4e2a\u535a\u5f08\u53c8\u662f\u76f8\u901a\u7684\u3002\u5f53\u901a\u8fc7\u8fd9\u79cd\u65b9\u5f0f\u5728KGS\u6570\u636e\u96c6\u5408\u4e0a\u8bad\u7ec3\u662f\uff0c\u4f30\u503c\u7f51\u7edc\u8bb0\u4f4f\u4e86\u535a\u5f08\u7684\u7ed3\u5c40\u800c\u4e0d\u662f\u63a8\u5e7f\u51fa\u65b0\u7684\u68cb\u5c40\uff0c\u5728\u6d4b\u8bd5\u6570\u636e\u4e0a\u9762MSE\u6700\u5c0f\u8fbe\u5230\u4e860.37\uff0c\u4e0e\u4e4b\u76f8\u6bd4\u5728\u8bad\u7ec3\u6570\u636e\u96c6\u5408\u4e0a\u9762MSE\u662f0.19\u3002\u4e3a\u4e86\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\uff0c\u6211\u4eec\u60f3\u51fa\u4e86\u65b0\u7684\u81ea\u6211\u5bf9\u5f08\u7684\u6570\u636e\u96c6\u5408\uff0c\u5305\u542b\u4e86\u4e09\u5343\u4e07\u4e2a\u4e0d\u540c\u7684\u68cb\u5c40\uff0c\u6bcf\u4e00\u4e2a\u90fd\u662f\u4ece\u4e0d\u540c\u76d8\u535a\u5f08\u4e2d\u91c7\u6837\u3002\u6bcf\u4e00\u76d8\u535a\u5f08\u90fd\u662f\u5728 RL\u7b56\u7565\u7f51\u7edc\u548c\u81ea\u5df1\u4e4b\u95f4\u5bf9\u5f08\uff0c\u76f4\u5230\u535a\u5f08\u672c\u8eab\u7ed3\u675f\u3002\u5728\u8fd9\u4e2a\u6570\u636e\u96c6\u5408\u4e0a\u8bad\u7ec3\u5bfc\u81f4\u4e86MSE\u4e3a0.226\uff0c\u548c\u8bad\u7ec3\u548c\u6d4b\u8bd5\u6570\u636e\u96c6\u5408\u7684MSE\u4e3a0.234\uff0c\u8fd9\u9884\u793a\u7740\u5f88\u5c0f\u7684\u8fc7\u5ea6\u62df\u5408\u3002\u56fe2\uff0cb\u5c55\u793a\u4e86\u4f30\u503c\u7f51\u7edc\u5bf9\u68cb\u5c40\u8bc4\u4f30\u7684\u7cbe\u786e\u5ea6\uff1a\u5bf9\u6bd4\u4f7f\u7528\u4e86\u5feb\u901f\u8d70\u5b50\u7b56\u7565\u7f51\u7edcp<\/span>\u03c0<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span>p\u03c0\u00a0<\/span><\/span>\u7684\u8499\u7279\u5361\u6d1b\u8d70\u5b50\u7684\u7cbe\u786e\u5ea6\uff0c\u4f30\u503c\u51fd\u6570\u4e00\u76f4\u66f4\u52a0\u7cbe\u786e\u3002\u4e00\u4e2a\u5355\u4e00\u7684\u8bc4\u4f30v<\/span>\u03b8<\/span><\/span><\/span><\/span>(<\/span>s<\/span>)<\/span>\u00a0<\/span><\/span><\/span><\/span>\u7684\u7cbe\u786e\u5ea6\u4e5f\u903c\u8fd1\u4e86\u4f7f\u7528\u4e86 RL\u7b56\u7565\u7f51\u7edcv<\/span>\u03b8<\/span><\/span><\/span><\/span>(<\/span>s<\/span>)<\/span><\/span><\/span><\/span>\u7684\u8499\u7279\u5361\u6d1b\u8d70\u5b50\u7684\u7cbe\u786e\u5ea6\uff0c\u4e0d\u8fc7\u8ba1\u7b97\u91cf\u662f\u539f\u6765\u768415000\u5206\u4e4b\u4e00\u3002<\/p>\n

\n

<\/a>4 \u8fd0\u7528\u7b56\u7565\u7f51\u7edc\u548c\u4f30\u503c\u7f51\u7edc\u641c\u7d22<\/h2>\n

AlphaGo\u5728\u628a\u7b56\u7565\u7f51\u7edc\u3001\u4f30\u503c\u7f51\u7edc\u548cMCTS\u7b97\u6cd5\u7ed3\u5408\uff0cMCTS\u901a\u8fc7\u9884\u6d4b\u641c\u7d22\u9009\u62e9\u4e0b\u68cb\u52a8\u4f5c\u3002\u6bcf\u4e00\u4e2a\u641c\u7d22\u6811\u7684\u8fb9\uff08s\uff0ca\uff09\u5b58\u50a8\u7740\u4e00\u4e2a\u52a8\u4f5c\u4f30\u503c Q(s, a)\uff0c\u8bbf\u95ee\u8ba1\u6570 N(s, a)\uff0c\u548c\u5148\u9a8c\u6982\u7387 P(s, a)\u3002\u8fd9\u68f5\u6811\u4ece\u6839\u8282\u70b9\u5f00\u59cb\uff0c\u901a\u8fc7\u6a21\u62df\u6765\u904d\u5386\uff08\u6bd4\u5982\u5728\u5b8c\u6574\u7684\u535a\u5f08\u4e2d\u6cbf\u7740\u6811\u65e0\u6ca1\u6709\u5907\u4efd\u5730\u5411\u4e0b\u641c\u7d22\uff09\u3002\u5728\u6bcf\u4e00\u6b21\u6a21\u62df\u7684\u65f6\u95f4\u6b65\u9aa4 t\uff0c\u5728\u72b6\u6001 s\u7684\u65f6\u5019\u9009\u62e9\u4e00\u4e2a\u4e0b\u68cb\u52a8\u4f5ca<\/span>t<\/span><\/span><\/span><\/span><\/span><\/span>\u00a0<\/span><\/span>\uff0c
\n\"\"<\/a>
\n\u7528\u6765\u6700\u5927\u5316\u52a8\u4f5c\u4f30\u503c\u52a0\u4e0a\u4e00\u4e2a\u989d\u5916\u5956\u52b1u<\/span>(<\/span>s<\/span>,<\/span>a<\/span>)<\/span>\u223c<\/span>P<\/span>(<\/span>s<\/span>,<\/span>a<\/span>)<\/span><\/span>1<\/span>+<\/span>N<\/span>(<\/span>s<\/span>,<\/span>a<\/span>)<\/span><\/span><\/span><\/span><\/span>\u00a0<\/span><\/span>\uff0c\u5b83\u548c\u5148\u9a8c\u6982\u7387\u6210\u6b63\u5411\u5173\u7cfb\uff0c\u4f46\u662f\u548c\u91cd\u590d\u8bbf\u95ee\u6b21\u6570\u6210\u53cd\u5411\u5173\u7cfb\uff0c\u8fd9\u6837\u662f\u4e3a\u4e86\u9f13\u52b1\u66f4\u591a\u7684\u63a2\u7d22\u3002\u5f53\u5728\u6b65\u9aa4 L \u904d\u5386\u5230\u8fbe\u4e00\u4e2a\u53f6\u8282\u70b9s<\/span>L<\/span><\/span><\/span><\/span><\/span><\/span>\u00a0<\/span><\/span>\u65f6\uff0c\u8be5\u53f6\u8282\u70b9\u53ef\u80fd\u4e0d\u4f1a\u88ab\u6269\u5c55\u3002\u53f6\u8282\u70b9\u68cb\u5c40s<\/span>L<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span>sL\u00a0<\/span><\/span>\u4ec5\u88ab SL\u7b56\u7565\u7f51\u7edcp<\/span>\u03b4<\/span><\/span><\/span><\/span><\/span><\/span><\/span>\u6267\u884c\u4e00\u6b21\u3002\u8f93\u51fa\u7684\u6982\u7387\u5b58\u50a8\u4e0b\u6765\u4f5c\u4e3a\u6bcf\u4e00\u5408\u6cd5\u4e0b\u6cd5\u52a8\u4f5c a \u7684\u5148\u9a8c\u6982\u7387 P\uff0cP<\/span>(<\/span>s<\/span>,<\/span>a<\/span>)<\/span>=<\/span>p<\/span>\u03b4<\/span><\/span><\/span><\/span>(<\/span>a<\/span>\u2223<\/span>s<\/span>)<\/span>\u00a0<\/span><\/span><\/span><\/span>\u3002\u53f6\u8282\u70b9\u901a\u8fc7\u4e24\u79cd\u65b9\u5f0f\u7684\u5f97\u5230\u8bc4\u4f30\uff1a\u7b2c\u4e00\uff0c\u901a\u8fc7\u4ef7\u503c\u7f51\u7edcv<\/span>\u03b8<\/span><\/span><\/span><\/span>(<\/span>s<\/span>L<\/span><\/span><\/span><\/span>)<\/span><\/span><\/span><\/span>\u8bc4\u4f30\uff1b\u7b2c\u4e8c\uff0c\u7528\u5feb\u901f\u8d70\u5b50\u7b56\u7565p<\/span>\u03c0<\/span><\/span><\/span><\/span><\/span><\/span><\/span>\u968f\u673a\u8d70\u5b50\uff0c\u76f4\u5230\u7ec8\u70b9\u6b65\u9aa4 T\uff0c\u4ea7\u751f\u7684\u7ed3\u679cz<\/span>L<\/span><\/span><\/span><\/span><\/span><\/span><\/span>\u4f5c\u4e3a\u8bc4\u4f30\u65b9\u6cd5\u3002\u8fd9\u4e9b\u8bc4\u4f30\u65b9\u6cd5\u7ed3\u5408\u5728\u4e00\u8d77\uff0c\u5728\u53f6\u8282\u70b9\u7684\u8bc4\u4f30\u51fd\u6570V<\/span>(<\/span>S<\/span>L<\/span><\/span><\/span><\/span>)<\/span>\u00a0<\/span><\/span><\/span><\/span>\u4e2d\u4f7f\u7528\u4e00\u4e2a\u6df7\u5408\u53c2\u6570\u03bb<\/span>\u00a0<\/span><\/span><\/span><\/span>\uff0c<\/p>\n

\"\"<\/a><\/p>\n

\u5728\u6a21\u62df\u7684\u7ed3\u5c3e n\uff0c\u66f4\u65b0\u6240\u6709\u88ab\u904d\u5386\u8fc7\u7684\u8fb9\u7684\u4e0b\u68cb\u52a8\u4f5c\u4f30\u503c\u548c\u8bbf\u95ee\u6b21\u6570\u3002\u6bcf\u4e00\u6761\u8fb9\u7d2f\u52a0\u8bbf\u95ee\u6b21\u6570\uff0c\u548c\u6c42\u51fa\u6240\u6709\u7ecf\u8fc7\u8be5\u8fb9\u7684\u6a21\u62df\u4f30\u503c\u7684\u5e73\u5747\u503c\u3002<\/p>\n

\"\"<\/a>
\n\u5176\u4e2ds<\/span>i<\/span><\/span><\/span>L<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span>sLi\u00a0<\/span><\/span>\u662f\u7b2c i \u6b21\u6a21\u62df\u7684\u53f6\u8282\u70b9\uff0c1(s, a, i) \u4ee3\u8868\u4e00\u6761\u8fb9 (s, a) \u5728\u7b2c i \u6b21\u6a21\u62df\u65f6\u662f\u5426\u88ab\u904d\u5386\u8fc7\u3002\u4e00\u65e6\u641c\u7d22\u5b8c\u6210\uff0c\u7b97\u6cd5\u9009\u62e9\u4ece\u6839\u8282\u70b9\u5f00\u59cb\uff0c\u88ab\u8bbf\u95ee\u6b21\u6570\u6700\u591a\u7684\u8282\u70b9\u3002<\/p>\n

\u5728AlphaGo\u4e2d\uff0cSL\u7b56\u7565\u7f51\u7edcp<\/span>\u03b4<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span><\/span>\u7684\u8868\u73b0\u4f18\u4e8e RL\u7b56\u7565\u7f51\u7edcp<\/span>\u03c1<\/span><\/span><\/span><\/span><\/span><\/span><\/span>\uff0c\u63a8\u6d4b\u53ef\u80fd\u662f\u56e0\u4e3a\u4eba\u7c7b\u4ece\u4e00\u675f\u4e0d\u540c\u7684\u524d\u666f\u5f88\u597d\u7684\u4e0b\u68cb\u8d70\u6cd5\u4e2d\u9009\u62e9\uff0c\u7136\u800c RL\u4f18\u5316\u5355\u4e00\u6700\u4f18\u4e0b\u68cb\u8d70\u6cd5\u3002\u7136\u800c\uff0c\u4ece\u66f4\u5f3a\u7684 RL\u7b56\u7565\u7f51\u7edc\u8bad\u7ec3\u51fa\u6765\u7684\u4f30\u503c\u51fd\u6570v<\/span>\u03b8<\/span><\/span><\/span><\/span>\u2248<\/span>v<\/span>P<\/span>\u03c1<\/span><\/span><\/span><\/span><\/span><\/span><\/span>(<\/span>s<\/span>)<\/span>\u00a0<\/span><\/span><\/span><\/span>\u4f18\u4e8e\u4ece SL\u7b56\u7565\u7f51\u7edc\u8bad\u7ec3\u51fa\u6765\u7684\u4f30\u503c\u51fd\u6570v<\/span>\u03b8<\/span><\/span><\/span><\/span>\u2248<\/span>v<\/span>P<\/span>\u03b4<\/span><\/span><\/span><\/span><\/span><\/span><\/span>(<\/span>s<\/span>)<\/span>\u00a0<\/span><\/span><\/span><\/span><\/p>\n

\u8bc4\u4f30\u7b56\u7565\u7f51\u7edc\u548c\u4f30\u503c\u7f51\u7edc\u548c\u4f20\u7edf\u7684\u542f\u53d1\u5f0f\u641c\u7d22\u76f8\u6bd4\uff0c\u9700\u8981\u591a\u51e0\u4e2a\u6570\u91cf\u7ea7\u7684\u8ba1\u7b97\u91cf\u3002\u4e3a\u4e86\u9ad8\u6548\u7684\u628a MCTS \u548c\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u7ed3\u5408\u5728\u4e00\u8d77\uff0cAlphaGo \u5728\u5f88\u591aCPU\u4e0a\u4f7f\u7528\u5f02\u6b65\u591a\u7ebf\u7a0b\u641c\u7d22\u6280\u672f\u8fdb\u884c\u4e86\u6a21\u62df\uff0c\u5728\u5f88\u591aGPU\u4e0a\u8ba1\u7b97\u7b56\u7565\u7f51\u7edc\u548c\u4f30\u503c\u7f51\u7edc\u3002\u6700\u7ec8\u7248\u672c\u7684 AlphaGo\u4f7f\u7528\u4e8640\u4e2a\u641c\u7d22\u7ebf\u7a0b\uff0c48\u4e2aCPU\uff0c\u548c8\u4e2aGPU\u3002\u6211\u4eec\u4e5f\u5b9e\u73b0\u4e86\u4e00\u4e2a\u5206\u5e03\u5f0f\u7684AlphaGo\u7248\u672c\uff0c\u5b83\u5229\u7528\u4e86\u591a\u53f0\u7535\u8111\uff0c40\u4e2a\u641c\u7d22\u7ebf\u7a0b\uff0c1202\u4e2aCPU\uff0c176\u4e2aGPU\u3002\u5728\u65b9\u6cd5\u90e8\u5206\u63d0\u4f9b\u4e86\u5173\u4e8e\u5f02\u6b65\u548c\u5206\u5e03MCTS\u7684\u5168\u90e8\u7684\u7ec6\u8282\u3002
\n\"\"<\/a>
\n\u56fe3\uff1aAlphaGo\u4e2d\u7684\u8499\u7279\u5361\u6d1b\u6811\u641c\u7d22\u3002a<\/strong>\u00a0\u6bcf\u4e00\u6b21\u6a21\u62df\u904d\u5386\u641c\u7d22\u6811\uff0c\u901a\u8fc7\u9009\u62e9\u62e5\u6709\u6700\u5927\u4e0b\u68cb\u52a8\u4f5c\u4f30\u503c Q\u7684\u8fb9\uff0c\u52a0\u4e0a\u4e00\u4e2a\u989d\u5916\u5956\u52b1 u(P)\uff08\u4f9d\u8d56\u4e8e\u5b58\u50a8\u7684\u8be5\u8fb9\u7684\u5148\u9a8c\u6982\u7387 P\uff09 \u3002b<\/strong>\u53f6\u8282\u70b9\u53ef\u80fd\u88ab\u5c55\u5f00\uff0c\u65b0\u7684\u7ed3\u70b9\u88ab\u7b56\u7565\u7f51\u7edcp<\/span>\u03b8<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span><\/span>\u6267\u884c\u4e00\u6b21\uff0c\u7136\u540e\u7ed3\u679c\u6982\u7387\u5b58\u50a8\u4e0b\u6765\u4f5c\u4e3a\u6bcf\u4e00\u4e2a\u4e0b\u68cb\u52a8\u4f5c\u7684\u5148\u9a8c\u6982\u7387\u3002c<\/strong>\u5728\u4e00\u6b21\u6a21\u62df\u7684\u7ed3\u5c3e\uff0c\u53f6\u8282\u70b9\u7531\u4e24\u79cd\u65b9\u5f0f\u8bc4\u4f30\uff1a\u4f7f\u7528\u4f30\u503c\u7f51\u7edc\u00a0v<\/span>\u03b8<\/span><\/span><\/span><\/span><\/span><\/span><\/span>\u548c\u8fd0\u884c\u4e00\u4e2a\u8d70\u5b50\u7b56\u7565\u76f4\u5230\u535a\u5f08\u7ed3\u5c3e\uff08\u4f7f\u7528\u4e86\u5feb\u901f\u8d70\u5b50\u7b56\u7565p<\/span>\u03c0<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span><\/span>\uff09\uff0c\u7136\u540e\u5bf9\u4e8e\u8d62\u8005\u8fd0\u7b97\u51fd\u6570 r\u3002d<\/strong>\u00a0\u4e0b\u68cb\u52a8\u4f5c\u4f30\u503c Q \u5f97\u5230\u66f4\u65b0\uff0c\u7528\u6765\u8ddf\u8e2a\u6240\u6709\u8bc4\u4f30 r(.) \u7684\u5e73\u5747\u503c\u548c\u8be5\u4e0b\u68cb\u52a8\u4f5c\u7684\u5b50\u6811\u7684v<\/span>\u03b8<\/span><\/span><\/span><\/span>(<\/span>\u22c5<\/span>)<\/span><\/span><\/span><\/span><\/p>\n

\n

<\/a>5 \u8bc4\u4f30AlphaGo\u7684\u4e0b\u68cb\u80fd\u529b<\/h2>\n

\u4e3a\u4e86\u8bc4\u4f30 AlphaGo \u7684\u6c34\u5e73\uff0c\u6211\u4eec\u4e3e\u529e\u4e86\u5185\u90e8\u6bd4\u8d5b\uff0c\u6210\u5458\u5305\u62ec\u4e0d\u540c\u7248\u672c\u7684 AlphaGo \u548c\u51e0\u4e2a\u5176\u5b83\u7684\u56f4\u68cb\u7a0b\u5e8f\uff0c\u5305\u62ec\u6700\u5f3a\u7684\u5546\u4e1a\u7a0b\u5e8f CrazyStone\u548cZen\uff0c\u548c\u6700\u5f3a\u7684\u5f00\u6e90\u7a0b\u5e8fPachi\u548cFuego\u3002\u6240\u6709\u8fd9\u4e9b\u7a0b\u5e8f\u90fd\u662f\u57fa\u4e8e\u9ad8\u6027\u80fd\u8499\u7279\u5361\u6d1b\u6811\u641c\u7d22\u7b97\u6cd5\u7684\u3002\u6b64\u5916\uff0c\u6211\u4eec\u7684\u5185\u90e8\u6bd4\u8d5b\u8fd8\u5305\u62ec\u4e86\u5f00\u6e90\u7a0b\u5e8fGnuGo\uff0c\u5b83\u4f7f\u7528\u4e86\u6700\u5148\u8fdb\u641c\u7d22\u65b9\u6cd5\u7684\u8499\u7279\u5361\u6d1b\u6811\u641c\u7d22\u3002\u6240\u6709\u7a0b\u5e8f\u6bcf\u6b21\u6267\u884c\u4e0b\u4e00\u6b65\u68cb\u5141\u8bb85\u79d2\u949f\u3002<\/p>\n

\u6bd4\u8d5b\u7684\u7ed3\u679c\u5982\u56fe4,a\uff0c\u5b83\u9884\u793a\u7740\u5355\u51fb\u7248\u672c\u7684AlphaGo\u6bd4\u5148\u524d\u4efb\u4f55\u4e00\u4e2a\u56f4\u68cb\u7a0b\u5e8f\u5f3a\u4e0a\u5f88\u591a\u6bb5\uff0c\u5728495\u573a\u6bd4\u8d5b\u4e2d\uff0cAlphaGo\u8d62\u4e86\u5176\u4e2d\u7684494\u573a\u6bd4\u8d5b\u3002\u6211\u4eec\u4e5f\u5728\u8ba9\u5bf9\u624b4\u76ee\u68cb\u7684\u60c5\u51b5\u4e0b\u8fdb\u884c\u4e86\u6bd4\u8d5b\uff1aAlphaGo\u548cCrazyStone\uff0cZen\uff0cPachi\u7684\u80dc\u7387\u5206\u522b\u662f77%\uff0c86%\uff0c99%\u3002\u5206\u5e03\u5f0f\u7248\u672c\u7684AlphaGo\u5f3a\u5927\u5f88\u591a\uff1a\u548c\u5355\u673a\u7248\u672c\u7684AlphaGo\u5bf9\u5f08\u7684\u80dc\u7387\u662f77%\uff0c\u548c\u5176\u4ed6\u7684\u56f4\u68cb\u7a0b\u5e8f\u5bf9\u5f08\u7684\u80dc\u7387\u662f100%\u3002<\/p>\n

\u6211\u4eec\u4e5f\u8bc4\u4f30\u4e86\u4e0d\u540c\u7248\u672c\u7684AlphaGo\uff0c\u4e0d\u540c\u7248\u672c\u4ec5\u4ec5\u4f7f\u7528\u4f30\u503c\u7f51\u7edc\uff08\u03bb<\/span>=<\/span>0<\/span><\/span><\/span><\/span>\uff09\u6216\u8005\u4ec5\u4ec5\u4f7f\u7528\u5feb\u901f\u8d70\u5b50\uff08\u03bb<\/span>=<\/span>1<\/span><\/span><\/span><\/span>\uff09\uff08\u5982\u56fe4,b\uff09\u3002\u5373\u4f7f\u4e0d\u4f7f\u7528\u5feb\u901f\u8d70\u5b50\uff0cAlphaGo\u7684\u6027\u80fd\u8d85\u51fa\u4e86\u6240\u6709\u5176\u4ed6\u56f4\u68cb\u7a0b\u5e8f\uff0c\u663e\u793a\u4e86\u4f30\u503c\u7f51\u7edc\u63d0\u4f9b\u4e86\u4e00\u4e2a\u53ef\u884c\u7684\u4ee3\u66ff\u8499\u7279\u5361\u6d1b\u8bc4\u4f30\u7684\u53ef\u80fd\u3002\u4e0d\u8fc7\uff0c\u5176\u4f30\u503c\u7f51\u7edc\u548c\u5feb\u901f\u8d70\u5b50\u7684\u6df7\u5408\u7248\u672c\u8868\u73b0\u6700\u597d\uff08\u03bb<\/span>=<\/span>0.5<\/span><\/span><\/span><\/span>\uff09\uff0c\u5728\u5bf9\u5176\u4ed6\u7248\u672c\u7684AlphaGo\u7684\u65f6\u5019\u80dc\u7387\u8fbe\u5230\u4e8695%\u4ee5\u4e0a\u3002\u8fd9\u9884\u793a\u7740\u8fd9\u4e24\u79cd\u68cb\u5c40\u8bc4\u4f30\u7cfb\u7edf\u662f\u4e92\u8865\u7684\uff1a\u4f30\u503c\u7f51\u7edc\u901a\u8fc7\u80fd\u529b\u5f88\u5f3a\uff0c\u4f46\u662f\u4e0d\u5207\u5b9e\u9645\u7684\u6162\u7684p<\/span>\u03c1<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span><\/span>\u6765\u903c\u8fd1\u535a\u5f08\u7684\u7ed3\u5c40\uff1b\u5feb\u901f\u8d70\u5b50\u53ef\u4ee5\u901a\u8fc7\u80fd\u529b\u66f4\u5f31\u4f46\u662f\u66f4\u5feb\u7684\u5feb\u901f\u8d70\u5b50\u7b56\u7565p<\/span>\u03c1<\/span><\/span><\/span><\/span>\u00a0<\/span><\/span><\/span><\/span>\u6765\u7cbe\u786e\u7684\u8bc4\u4f30\u535a\u5f08\u7684\u7ed3\u5c40\u3002\u56fe5\u5c06AlphaGo\u5728\u771f\u5b9e\u535a\u5f08\u68cb\u5c40\u4e2d\u8bc4\u4f30\u80fd\u529b\u53ef\u89c6\u5316\u4e86\u3002<\/p>\n

\u6700\u7ec8\uff0c\u6211\u4eec\u628a\u5206\u5e03\u5f0f\u7248\u672c\u7684AlphaGo\u548c\u6a0a\u9ebe\u8fdb\u884c\u4e86\u8bc4\u4f30\uff0c\u4ed6\u4f5c\u4e3a\u4e00\u4e2a\u804c\u4e1a2\u6bb5\u68cb\u624b\uff0c\u662f2013\uff0c2014\uff0c2015\u5e74\u7684\u6b27\u6d32\u56f4\u68cb\u51a0\u519b\u3002\u57282015\u5e7410\u67085-9\u65e5\uff0cAlphaGo\u548c\u6a0a\u9ebe\u5728\u771f\u5b9e\u6bd4\u8d5b\u4e2d\u4e0b\u4e865\u76d8\u68cb\u3002AlphaGo\u4ee55\uff1a0\u7684\u6bd4\u5206\u8d62\u4e86\u6bd4\u8d5b\uff08\u56fe6\u548c\u6269\u5c55\u6570\u636e\u88681\uff09\u3002\u8fd9\u662f\u53f2\u4e0a\u7b2c\u4e00\u6b21\uff0c\u5728\u4eba\u7c7b\u4e0d\u8ba9\u5b50\u548c\u5b8c\u6574\u68cb\u76d8\u7684\u60c5\u51b5\u4e0b\uff0c\u4e00\u4e2a\u56f4\u68cb\u7a0b\u5e8f\u5728\u8d62\u4e86\u4e00\u4e2a\u4eba\u7c7b\u804c\u4e1a\u68cb\u624b\u3002\u8fd9\u4e2a\u58ee\u4e3e\u4e4b\u524d\u8ba4\u4e3a\u9700\u8981\u81f3\u5c11\u5341\u5e74\u624d\u80fd\u8fbe\u5230\u3002<\/p>\n

\"\"<\/a>
\n\u56fe4\uff1aAlphaGo\u7684\u6bd4\u8d5b\u8bc4\u4f30<\/strong>\u3002a<\/strong>\u00a0\u548c\u4e0d\u540c\u56f4\u68cb\u7a0b\u5e8f\u6bd4\u8d5b\u7684\u7ed3\u679c\uff08\u89c1\u6269\u5c55\u6570\u636e\u88686-11\uff09\u3002\u6bcf\u4e2a\u7a0b\u5e8f\u4f7f\u7528\u63a5\u8fd15\u79d2\u6bcf\u8d70\u4e00\u6b65\u68cb\u7684\u901f\u5ea6\u3002\u4e3a\u4e86\u7ed9AlphaGo\u66f4\u9ad8\u7684\u6311\u6218\u96be\u5ea6\uff0c\u4e00\u4e9b\u7a0b\u5e8f\u5f97\u5230\u4e86\u6240\u6709\u5bf9\u624b\u8ba94\u6b65\u5b50\u7684\u4f18\u52bf\u3002\u7a0b\u5e8f\u7684\u8bc4\u4f30\u57fa\u4e8eELO\u4f53\u7cfb\uff1a230\u5206\u7684\u5dee\u8ddd\uff0c\u8fd9\u76f8\u5f53\u4e8e79%\u7684\u80dc\u7387\u5dee\u8ddd\uff0c\u8fd9\u5927\u81f4\u76f8\u5f53\u4e8e\u5728KGS\u4e2d\u9ad8\u4e00\u4e2a\u4e1a\u4f59\u7b49\u7ea7\u3002\u4e00\u4e2a\u548c\u4eba\u7c7b\u63a5\u8fd1\u7684\u76f8\u5f53\u4e5f\u663e\u793a\u4e86\uff0c\u6c34\u5e73\u7684\u7ebf\u663e\u793a\u4e86\u7a0b\u5e8f\u5728\u5728\u7ebf\u6bd4\u8d5b\u4e2d\u8fbe\u5230\u7684KSG\u7b49\u7ea7\u3002\u548c\u6b27\u6d32\u51a0\u519b\u6a0a\u9ebe\u7684\u6bd4\u8d5b\u4e5f\u5305\u62ec\u5728\u5185\u3002\u8fd9\u4e9b\u6bd4\u8d5b\u4f7f\u7528\u66f4\u957f\u7684\u65f6\u95f4\u63a7\u5236\u3002\u56fe\u4e2d\u663e\u793a\u4e8695%\u7684\u7f6e\u4fe1\u533a\u95f4\u3002b<\/strong>\u00a0\u5355\u673a\u7248\u672c\u7684AlphaGo\u5728\u7ec4\u6210\u90e8\u5206\u7684\u4e0d\u540c\u7ec4\u5408\u4e0b\u7684\u6027\u80fd\u8868\u73b0\u3002\u5176\u4e2d\u4ec5\u4ec5\u4f7f\u7528\u4e86\u7b56\u7565\u7f51\u7edc\u7684\u7248\u672c\u6ca1\u6709\u4f7f\u7528\u4efb\u4f55\u641c\u7d22\u7b97\u6cd5\u3002c<\/strong>\u00a0\u8499\u7279\u5361\u6d1b\u641c\u7d22\u6811\u7b97\u6cd5\u5173\u4e8e\u641c\u7d22\u7ebf\u7a0b\u548cGPU\u7684\u53ef\u6269\u5c55\u6027\u7814\u7a76\uff0c\u5176\u4e2d\u4f7f\u7528\u4e86\u5f02\u6b65\u641c\u7d22\uff08\u6d45\u84dd\u8272\uff09\u548c\u5206\u5e03\u5f0f\u641c\u7d22\uff08\u6df1\u84dd\u8272\uff09\uff0c\u6bcf\u4e0b\u4e00\u6b65\u65f6\u95f4\u4e24\u79d2\u3002<\/p>\n

\"\"<\/a>
\n\u56fe5\uff1aAlphaGo\uff08\u6267\u9ed1\uff09\u662f\u5728\u4e00\u4e2a\u548c\u6a0a\u9ebe\u7684\u975e\u6b63\u5f0f\u7684\u6bd4\u8d5b\u4e2d\u9009\u62e9\u4e0b\u68cb\u8d70\u5b50\u7684\u3002<\/strong>\u63a5\u4e0b\u6765\u7684\u6bcf\u4e00\u4e2a\u7edf\u8ba1\u4e2d\uff0c\u4f30\u503c\u6700\u5927\u7684\u843d\u5b50\u5730\u70b9\u7528\u6a58\u9ec4\u8272\u6807\u8bb0\u3002a<\/strong>\u6839\u8282\u70b9 s \u7684\u6240\u6709\u540e\u7ee7\u7ed3\u70b9 s\u2019 \u7684\u4f30\u503c\uff0c\u4f7f\u7528\u4f30\u503c\u7f51\u7edc\u00a0v<\/span>\u03b8<\/span><\/span><\/span><\/span>(<\/span>s<\/span>\u2032<\/span><\/span>)<\/span>\u00a0<\/span><\/span><\/span><\/span>\uff0c\u8bc4\u4f30\u5f88\u9760\u524d\u7684\u4f1a\u8d62\u7684\u767e\u5206\u6570\u663e\u793a\u51fa\u6765\u4e86\u3002b<\/strong>\u00a0\u4ece\u6839\u8282\u70b9\u5f00\u59cb\u7684\u6bcf\u4e00\u6761\u8fb9(s, a)\u7684\u8d70\u5b50\u52a8\u4f5c\u4f30\u503c Q(s, a)\uff1b\u4ec5\u4ec5\u4f7f\u7528\u4f30\u503c\u7f51\u7edc\uff08\u03bb<\/span>=<\/span>0<\/span><\/span><\/span>\u00a0<\/span><\/span>\uff09\u65b9\u6cd5\u7684\u5747\u503c\u3002c<\/strong>\u00a0\u4e0b\u68cb\u52a8\u4f5c Q(s, a)\uff0c\u4ec5\u4ec5\u4f7f\u7528\u5feb\u901f\u8d70\u5b50\uff08\u03bb<\/span>=<\/span>1<\/span><\/span><\/span>\u00a0<\/span><\/span>\uff09\u65b9\u6cd5\u7684\u5747\u503c\u3002d<\/strong>\u00a0\u76f4\u63a5\u4f7f\u7528 SL\u7b56\u7565\u7f51\u7edc\u7684\u4e0b\u68cb\u8d70\u5b50\u6982\u7387\uff0cp<\/span>\u03b4<\/span><\/span><\/span><\/span>(<\/span>a<\/span>|<\/span><\/span><\/span>s<\/span>)<\/span><\/span><\/span><\/span>\uff1b\u5982\u679c\u5927\u4e8e0.1%\u7684\u8bdd\uff0c\u4ee5\u767e\u5206\u6bd4\u7684\u5f62\u5f0f\u62a5\u544a\u51fa\u6765\u3002e<\/strong>\u00a0\u4ece\u6839\u8282\u70b9\u5f00\u59cb\u7684\u6a21\u62df\u8fc7\u7a0b\u4e2d\u4e0b\u68cb\u8d70\u5b50\u5730\u70b9\u9009\u62e9\u7684\u9891\u7387\u767e\u5206\u6bd4\u3002f<\/strong>\u00a0AlphaGo\u7684\u6811\u641c\u7d22\u7684\u7406\u8bba\u4e0a\u7684\u8d70\u5b50\u9009\u62e9\u5e8f\u5217\uff08\u4e00\u4e2a\u641c\u7d22\u8fc7\u7a0b\u4e2d\u8bbf\u95ee\u6b21\u6570\u6700\u591a\u7684\u8def\u5f84\uff09\u3002\u4e0b\u68cb\u8d70\u5b50\u7528\u4e00\u4e2a\u6570\u5b57\u5e8f\u5217\u8868\u793a\u3002AlphaGo\u9009\u62e9\u4e0b\u68cb\u7684\u843d\u5b50\u5730\u70b9\u7528\u7ea2\u8272\u5706\u5708\u6807\u8bb0\u51fa\u6765\uff1b\u6a0a\u9ebe\u4e0b\u5728\u767d\u8272\u65b9\u5f62\u7684\u5730\u65b9\u4f5c\u4e3a\u56de\u5e94\uff1b\u5728\u4ed6\u7684\u590d\u76d8\u8fc7\u7a0b\u4e2d\uff0c\u4ed6\u8bc4\u8bba\u9053\uff1a\u4e0b\u5728\u5730\u70b9 1 \u5e94\u8be5\u662f\u66f4\u597d\u7684\u9009\u62e9\uff0c\u800c\u8fd9\u4e2a\u843d\u5b50\u5730\u70b9\u6b63\u597d\u662fAlphaGo\u9884\u6d4b\u7684\u767d\u68cb\u7684\u843d\u5b50\u5730\u70b9\u3002<\/p>\n

\n

<\/a>6 \u8ba8\u8bba<\/h2>\n

\u5728\u8fd9\u4e2a\u5de5\u4f5c\u4e2d\uff0c\u6211\u4eec\u57fa\u4e8e\u4e00\u4e2a\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u548c\u6811\u641c\u7d22\u7684\u7ed3\u5408\u5f00\u53d1\u4e86\u4e00\u4e2a\u56f4\u68cb\u7a0b\u5e8f\uff0c\u5b83\u7684\u4e0b\u68cb\u6c34\u5e73\u8fbe\u5230\u4e86\u4eba\u7c7b\u6700\u5f3a\u7684\u6c34\u5e73\uff0c\u56e0\u6b64\u6210\u529f\u6218\u80dc\u4e86\u4e00\u9879\u4eba\u5de5\u667a\u80fd\u9886\u57df\u7684\u4f1f\u5927\u6311\u6218\u3002\u6211\u4eec\u9996\u6b21\uff0c\u5bf9\u56f4\u68cb\u5f00\u53d1\u4e86\u4e00\u4e2a\u6709\u6548\u7684\u4e0b\u68cb\u8d70\u5b50\u9009\u62e9\u5668\u548c\u68cb\u5c40\u8bc4\u4f30\u51fd\u6570\uff0c\u5b83\u662f\u57fa\u4e8e\u88ab\u4e00\u4e2a\u521b\u65b0\u578b\u7684\u76d1\u7763\u5b66\u4e60\u548c\u5f3a\u5316\u5b66\u4e60\u7684\u7ec4\u5408\u8bad\u7ec3\u7684\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u3002\u6211\u4eec\u5f15\u5165\u4e86\u65b0\u7684\u641c\u7d22\u7b97\u6cd5\uff0c\u5b83\u6210\u529f\u7684\u628a\u795e\u7ecf\u7f51\u7edc\u8bc4\u4f30\u548c\u8499\u7279\u5361\u6d1b\u8d70\u5b50\u7ed3\u5408\u5728\u4e00\u8d77\u3002\u6211\u4eec\u7684\u7a0b\u5e8fAlphaGo\u628a\u8fd9\u4e9b\u7ec4\u6210\u90e8\u5206\u6309\u7167\u6bd4\u4f8b\u96c6\u6210\u5728\u4e00\u8d77\uff0c\u6210\u4e3a\u4e86\u4e00\u4e2a\u9ad8\u6027\u80fd\u7684\u6811\u641c\u7d22\u5f15\u64ce\u3002<\/p>\n

\u5728\u548c\u6a0a\u9ebe\u7684\u6bd4\u8d5b\u4e2d\uff0cAlphaGo\u5bf9\u68cb\u5c40\u8bc4\u4f30\u7684\u6b21\u6570\u548c\u6df1\u84dd\u5bf9\u5361\u65af\u5e15\u7f57\u592b\u4e0b\u56fd\u9645\u8c61\u68cb\u7684\u65f6\u5019\u7684\u6b21\u6570\u76f8\u6bd4\uff0c\u662f\u5176\u5343\u5206\u4e4b\u4e00\u3002\u4f5c\u4e3a\u8865\u507f\u7684\uff0c\u662f\u66f4\u52a0\u667a\u80fd\u7684\u68cb\u5c40\u9009\u62e9\u80fd\u529b\uff0c\u4f7f\u7528\u4e86\u66f4\u52a0\u7cbe\u786e\u7684\u8bc4\u4f30\u68cb\u5c40\u7684\u80fd\u529b\uff0c\u4f7f\u7528\u4e86\u4f30\u503c\u7f51\u7edc\uff08\u4e00\u4e2a\u4e5f\u8bb8\u662f\u66f4\u52a0\u63a5\u8fd1\u4e8e\u4eba\u7c7b\u4e0b\u68cb\u65b9\u5f0f\u7684\u65b9\u6cd5\uff09\u3002\u6b64\u5916\uff0c\u6df1\u84dd\u4f7f\u7528\u7684\u662f\u4eba\u7c7b\u624b\u5de5\u8c03\u53c2\u6570\u7684\u4f30\u503c\u51fd\u6570\uff0c\u7136\u800cAlphaGo \u7684\u795e\u7ecf\u7f51\u7edc\u662f\u76f4\u63a5\u4ece\u6bd4\u8d5b\u5bf9\u5f08\u6570\u636e\u4e2d\u8bad\u7ec3\u51fa\u6765\u7684\uff0c\u7eaf\u901a\u8fc7\u4e00\u4e2a\u901a\u7528\u76ee\u7684\u7684\u76d1\u7763\u5b66\u4e60\u548c\u5f3a\u5316\u5b66\u4e60\u65b9\u6cd5\u3002<\/p>\n

\u56f4\u68cb\u5728\u5f88\u591a\u65b9\u9762\u662f\u6a2a\u4e98\u5728\u4eba\u5de5\u667a\u80fd\u9762\u524d\u7684\u56f0\u96be\uff1a\u4e00\u4e2a\u6709\u6311\u6218\u6027\u7684\u51b3\u7b56\u4efb\u52a1\uff1b\u4e00\u4e2a\u96be\u4ee5\u5bf9\u4ed8\u7684\u89e3\u7a7a\u95f4\uff1b\u548c\u4e00\u4e2a\u975e\u5e38\u590d\u6742\u7684\u6700\u4f18\u89e3\uff0c\u4ee5\u81f3\u4e8e\u5b83\u770b\u4e0a\u53bb\u4e0d\u53ef\u80fd\u4e16\u754c\u4f7f\u7528\u7b56\u7565\u6216\u8005\u4f30\u503c\u51fd\u6570\u903c\u8fd1\u3002\u4e4b\u524d\u7684\u5173\u4e8e\u56f4\u68cb\u7a0b\u5e8f\u7684\u91cd\u5927\u7a81\u7834\uff0c\u8499\u7279\u5361\u6d1b\u6811\u641c\u7d22\uff0c\u5728\u5176\u5b83\u9886\u57df\u5bfc\u81f4\u4e86\u76f8\u5e94\u7684\u8fdb\u6b65\uff1a\u6bd4\u5982\u901a\u7528\u7684\u535a\u5f08\u6bd4\u8d5b\uff0c\u7ecf\u5178\u7684\u89c4\u5212\u95ee\u9898\uff0c\u5c40\u90e8\u89c2\u5bdf\u89c4\u5212\u95ee\u9898\uff0c\u8c03\u5ea6\u95ee\u9898\uff0c\u548c\u7ea6\u675f\u6ee1\u8db3\u95ee\u9898\u3002\u901a\u8fc7\u628a\u6811\u641c\u7d22\u548c\u7b56\u7565\u7f51\u7edc\u3001\u4f30\u503c\u7f51\u7edc\u7ed3\u5408\u5728\u4e00\u8d77\uff0cAlphaGo \u6700\u7ec8\u8fbe\u5230\u4e86\u56f4\u68cb\u804c\u4e1a\u9009\u624b\u7684\u6c34\u5e73\uff0c\u5e76\u4e14\u63d0\u4f9b\u4e86\u5e0c\u671b\uff1a\u5728\u5176\u5b83\u770b\u4f3c\u96be\u4ee5\u89e3\u51b3\u7684\u4eba\u5de5\u667a\u80fd\u9886\u57df\u91cc\uff0c\u8ba1\u7b97\u673a\u73b0\u5728\u662f\u53ef\u4ee5\u8fbe\u5230\u4eba\u7c7b\u6c34\u5e73\u7684\u3002<\/p>\n

\"\"<\/a>
\n\u56fe6\uff1a\u00a0AlphaGo \u548c\u6b27\u6d32\u51a0\u519b\u6a0a\u9ebe\u7684\u535a\u5f08\u68cb\u5c40<\/strong>\u3002\u4e0b\u68cb\u8d70\u7684\u6bcf\u4e00\u6b65\u6309\u7167\u4e0b\u68cb\u987a\u5e8f\u7531\u6570\u5b57\u5e8f\u5217\u663e\u793a\u51fa\u6765\u3002\u91cd\u590d\u843d\u5b50\u7684\u5730\u65b9\u5728\u68cb\u76d8\u7684\u4e0b\u9762\u6210\u53cc\u6210\u5bf9\u663e\u793a\u51fa\u6765\u3002\u6bcf\u4e00\u5bf9\u6570\u5b57\u4e2d\u7b2c\u4e00\u4e2a\u6570\u5b57\u7684\u843d\u5b50\uff0c\u91cd\u590d\u4e0b\u5230\u4e86\u7b2c\u4e8c\u4e2a\u6570\u5b57\u663e\u793a\u7684\u4ea4\u53c9\u5730\u65b9\u3002<\/p>\n

\n

\u53c2\u8003\u6587\u732e<\/strong>
\n\u7565<\/p>\n

\u5176\u4ed6\u8d21\u732e\u8005<\/strong>
\n\u7565<\/p>\n

\u611f\u8c22<\/strong>
\n\u6211\u4eec\u611f\u8c22\u6a0a\u9ebe\u7b54\u5e94\u548cAlphaGo\u8fdb\u884c\u6bd4\u8d5b\uff1b\u611f\u8c22T.M\u62c5\u5f53\u6bd4\u8d5b\u7684\u88c1\u5224\uff1b\u611f\u8c22R.M\u548cT.S\u7ed9\u4e88\u6709\u5e2e\u52a9\u7684\u8ba8\u8bba\u548c\u5efa\u8bae\uff1b\u611f\u8c22A.C\u548cM.C\u5728\u53ef\u89c6\u5316\u65b9\u9762\u7684\u5de5\u4f5c\uff1b\u611f\u8c22P.D, G.W, D.K, D.P, H.vH, A.G\u548cG.O\u4fee\u8ba2\u4e86\u8fd9\u7bc7\u8bba\u6587\uff1b\u611f\u8c22 DeepMing \u56e2\u961f\u5176\u5b83\u7684\u6210\u5458\u7684\u652f\u6301\uff0c\u60f3\u6cd5\uff0c\u548c\u9f13\u52b1\u3002<\/p>\n

\n

<\/a>\u540e\u8bb0<\/h1>\n

\u7406\u89e3 AlphaGo \u6709\u4e24\u4e2a\u5173\u952e\u90e8\u5206\uff1a<\/strong><\/p>\n

    \n
  1. \u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u7684\u8bad\u7ec3\u8fc7\u7a0b\uff0c\u6587\u7ae0\u628a\u8fd9\u4e2a\u8fc7\u7a0b\u63cf\u8ff0\u6210\u4e3a\u4e00\u4e2a\u7ba1\u9053<\/strong>\u3002\u6240\u8c13\u7ba1\u9053\uff0c\u5f88\u50cfLinux\u7cfb\u7edf\u4e2d\u7684\u7ba1\u9053\u547d\u4ee4\uff0c\u628a\u524d\u8005\u7684\u8f93\u51fa\u4f5c\u4e3a\u540e\u8005\u7684\u8f93\u5165<\/li>\n
  2. \u8499\u7279\u5361\u6d1b\u6811\u641c\u7d22\u7684\u8fc7\u7a0b\uff0c\u770b\u8fd9\u4e2a\u641c\u7d22\u662f\u5982\u4f55\u628a SL\u7b56\u7565\u7f51\u7edc\uff0c\u4f30\u503c\u7f51\u7edc\uff0c\u5feb\u901f\u8d70\u5b50\u7b56\u7565\u7ed3\u5408\u5728\u4e00\u8d77\u7684\u3002<\/li>\n<\/ol>\n

    \u5148\u8c08\u8c08 1\uff1a
    \n\u8f93\u5165\u4eba\u7c7b\u7684\u68cb\u8c31\uff0c\u7ecf\u8fc7\u76d1\u7763\u5b66\u4e60\uff0c\u8f93\u51faSL\u7b56\u7565\u7f51\u7edc<\/strong><\/span>
    \n\u8f93\u5165SL\u7b56\u7565\u7f51\u7edc\uff0c\u7ecf\u8fc7\u5f3a\u5316\u5b66\u4e60\uff0c\u8f93\u51faRL\u7b56\u7565\u7f51\u7edc<\/strong><\/span>
    \n\u8f93\u5165RL\u7b56\u7565\u7f51\u7edc\uff0c\u7ecf\u8fc7\u5f3a\u5316\u5b66\u4e60\uff0c\u8f93\u51fa\u4f30\u503c\u7f51\u7edc<\/strong><\/span><\/p>\n

    \u518d\u8c08\u8c08 2\uff1a
    \n\u8499\u7279\u5361\u6d1b\u6811\u641c\u7d22\u662f\u6a21\u62df\u4e0b\u68cb\uff0c\u5e76\u4e14\u8bc4\u4f30\u7684\u8fc7\u7a0b\u3002\u8499\u7279\u5361\u6d1b\u5c31\u662f\u201c\u968f\u673a\u201d\u7684\u610f\u601d\uff0c\u53ea\u4e0d\u8fc7\u903c\u683c\u66f4\u9ad8\u800c\u5df2\u3002\u5982\u679c\u4f60\u201c\u968f\u673a\u201d\u4e0b\u68cb\uff0c\u80af\u5b9a\u8f93\u5440\uff0c\u600e\u4e48\u529e\uff1f\u4f7f\u7528\u00a0SL\u7b56\u7565\u7f51\u7edc\u6765\u9884\u6d4b\u4eba\u7c7b\u662f\u5982\u4f55\u4e0b\u68cb\u7684<\/strong>\u3002AlphaGo\u6bcf\u6b21\u8981\u4e0b\u68cb\u7684\u65f6\u5019\uff0c\u5148\u8fd0\u884c SL\u7b56\u7565\u7f51\u7edc\u4e00\u904d\uff0c\u5f97\u5230\u4e00\u4e2a\u6982\u7387\u5206\u5e03\uff0c\u5728\u6b64\u57fa\u7840\u4e0a\u8fdb\u884c\u201c\u968f\u673a\u201d\uff1a\u66f4\u6709\u53ef\u80fd\u5728\u6982\u7387\u66f4\u5927\u7684\u5730\u65b9\u843d\u5b50\u3002
    \nAlphaGo\u4e00\u8fb9\u6a21\u62df\u81ea\u5df1\u4e0b\u68cb\uff0c\u4e00\u8fb9\u6a21\u62df\u5bf9\u624b\u4e0b\u68cb\uff0c\u6700\u540e\uff0c\u4e0b\u5b8c\u4e86\u3002\u6240\u8c13\u4e0b\u5b8c\u4e86\uff0c\u5c31\u662f\u5728\u6811\u641c\u7d22\u7684\u65f6\u5019\u8fbe\u5230\u4e86\u53f6\u8282\u70b9\u3002\u4e0b\u5b8c\u4e86\u4e4b\u540e\uff0c\u5bf9\u68cb\u5c40\u8fdb\u884c\u8bc4\u4f30\u3002\u7ed3\u5408\u4f30\u503c\u7f51\u7edc\u548c\u5feb\u901f\u8d70\u5b50\u7b56\u7565\uff0c\u5f97\u5230\u4e00\u4e2a\u4f30\u503c\u51fd\u6570<\/strong>\uff0c\u8be5\u51fd\u6570\u7684\u503c\u8d8a\u9ad8\uff0c\u8d8a\u597d\u3002
    \n\u6a21\u62df\u5f88\u591a\u5f88\u591a\u904d<\/strong>\u3002\u6a21\u62df\u7ed3\u675f\u4e4b\u540e\uff0c\u8fdb\u884c\u201c\u7edf\u8ba1<\/strong>\u201d\u5de5\u4f5c\u3002\u7edf\u8ba1\u6bcf\u4e00\u6761\u8fb9\u8d70\u8fc7\u7684\u6b21\u6570\uff0c\u548c\u4f30\u503c\u51fd\u6570\u7684\u4f30\u503c\u3002\u6700\u540e\uff0cAlphaGo\u505a\u51fa\u9009\u62e9\uff1a\u73b0\u5728\u662f\u68cb\u5c40 s\uff0c\u5982\u679c\u5728 a \u5730\u65b9\uff0c\u7ed3\u5408 a \u5728\u6a21\u62df\u8fc7\u7a0b\u4e2d\u8d70\u8fc7\u7684\u6b21\u6570\uff0c\u4ee5\u53ca a \u4e0b\u9762\u7684\u53f6\u8282\u70b9\u7684\u4f30\u503c\u51fd\u6570\uff0c\u7d2f\u52a0\u8d77\u6765\u6700\u9ad8\uff0c\u90a3\u4e48AlphaGo\u9009\u62e9\u5728 a \u5730\u65b9\u843d\u5b50<\/strong>\u3002<\/p>\n

     <\/p>\n

    \u6458\u81ea\uff1ahttps:\/\/blog.csdn.net\/u013390476\/article\/details\/50925347<\/p>\n","protected":false},"excerpt":{"rendered":"

    AlphaGo\u8bba\u6587\u7684\u8bd1\u6587\uff1a\u7528\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u548c\u6811\u641c\u7d22\u5f81\u670d\u56f4\u68cb Mastering the game of Go with deep neural networks and tree search<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":[],"categories":[723],"tags":[726,728],"_links":{"self":[{"href":"https:\/\/yanjingang.com\/blog\/index.php?rest_route=\/wp\/v2\/posts\/2282"}],"collection":[{"href":"https:\/\/yanjingang.com\/blog\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/yanjingang.com\/blog\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/yanjingang.com\/blog\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/yanjingang.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=2282"}],"version-history":[{"count":0,"href":"https:\/\/yanjingang.com\/blog\/index.php?rest_route=\/wp\/v2\/posts\/2282\/revisions"}],"wp:attachment":[{"href":"https:\/\/yanjingang.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=2282"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/yanjingang.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=2282"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/yanjingang.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=2282"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}