{"id":32716,"date":"2022-02-27T22:12:25","date_gmt":"2022-02-27T14:12:25","guid":{"rendered":"https:\/\/swarma.org\/?p=32716"},"modified":"2022-02-27T22:12:25","modified_gmt":"2022-02-27T14:12:25","slug":"nat-mach-intell-%e7%bb%bc%e8%bf%b0%ef%bc%9a%e6%99%ba%e8%83%bd%e9%97%ae%e9%a2%98%e8%a7%a3%e5%86%b3-%e6%95%b4%e5%90%88%e7%9a%84%e5%b1%82%e7%ba%a7%e5%8c%96%e5%bc%ba%e5%8c%96%e5%ad%a6","status":"publish","type":"post","link":"https:\/\/swarma.org\/?p=32716","title":{"rendered":"Nat. Mach. Intell. \u7efc\u8ff0\uff1a\u667a\u80fd\u95ee\u9898\u89e3\u51b3\u2014\u2014\u6574\u5408\u7684\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60"},"content":{"rendered":"
\n
\n
\n
\n
\n

<\/p>\n<\/section>\n

\n
\n
\n
\n
\n
\n

<\/section>\n<\/section>\n
\n
\n

\u5bfc\u8bed<\/strong><\/p>\n<\/section>\n<\/section>\n

\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\n
\n

<\/strong>\u5c3d\u7ba1\u4eba\u5de5\u667a\u80fd\u6280\u672f\u7a81\u98de\u731b\u8fdb\uff0c\u4f46\u622a\u6b62\u76ee\u524d\uff0c\u4eba\u7c7b\u548c\u52a8\u7269\u5728\u89e3\u51b3\u8bb8\u591a\u95ee\u9898\uff08\u5c24\u5176\u662f\u590d\u6742\u7cfb\u7edf\u95ee\u9898\uff09\u7684\u80fd\u529b\u660e\u663e\u7531\u4e8e\u4eba\u5de5\u7cfb\u7edf\u3002\u7406\u89e3\u4eba\u7c7b\u7b49\u9ad8\u7b49\u52a8\u7269\u89e3\u51b3\u95ee\u9898\u7684\u80fd\u529b\uff0c\u5e76\u5728\u4eba\u5de5\u667a\u80fd\u7cfb\u7edf\u4e2d\u6a21\u62df\u5b9e\u73b0\uff0c\u662f\u91cd\u8981\u7684\u79d1\u5b66\u547d\u9898\u3002\u8fd1\u65e5\u53d1\u8868\u5728Nature Machine Intelligence<\/em>\u6742\u5fd7\u7684\u957f\u6587\u7efc\u8ff0\uff0c\u6982\u8ff0\u4e86\u5c42\u6b21\u5316\u95ee\u9898\u89e3\u51b3\u7684\u8ba4\u77e5\u57fa\u7840\uff0c\u4ee5\u53ca\u5982\u4f55\u5229\u7528\u5f3a\u5316\u5b66\u4e60\u6280\u672f\u5b9e\u73b0\u62bd\u8c61\u548c\u9884\u6d4b\u80fd\u529b\uff0c\u5e76\u63d0\u51fa\u5c0f\u6837\u672c\u5b66\u4e60\u662f\u5173\u952e\u3002\u8be5\u7efc\u8ff0\u80fd\u591f\u6307\u5bfc\u542f\u53d1\u672a\u6765\u5e94\u5bf9\u590d\u6742\u7cfb\u7edf\u95ee\u9898\u7684\u673a\u5668\u5b66\u4e60\u67b6\u6784\u8bbe\u8ba1\u3002\u672c\u6587\u662f\u5bf9\u8be5\u7efc\u8ff0\u7684\u5168\u6587\u7ffb\u8bd1\u3002<\/span><\/strong><\/p>\n


<\/p>\n

\u7814\u7a76\u9886\u57df\uff1a\u5f3a\u5316\u5b66\u4e60\uff0c\u591a\u4e3b\u4f53\u5efa\u6a21\uff0c\u8ba4\u77e5\u590d\u6742\u6027\uff0c\u5c0f\u6837\u672c\u5b66\u4e60\uff0c\u8ba4\u77e5\u795e\u7ecf\u79d1\u5b66<\/strong><\/strong><\/span><\/strong><\/span><\/strong><\/span><\/strong><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

\n
\n
\n
<\/svg><\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
<\/strong><\/section>\n<\/section>\n
\n

Manfred Eppe\u7b49<\/span> <\/span>| \u4f5c\u8005<\/strong><\/span><\/strong><\/p>\n

\u90ed\u745e\u4e1c<\/span> <\/span>| \u8bd1\u8005<\/strong><\/span><\/strong><\/span><\/p>\n

\u9648\u65af\u4fe1<\/span> <\/span>| \u5ba1\u6821<\/strong><\/span><\/strong><\/span><\/p>\n

\u9093\u4e00\u96ea<\/span> | \u7f16\u8f91<\/strong><\/span>
<\/p>\n


<\/p>\n


<\/p>\n<\/section>\n

<\/p>\n

\n
\n
\n

\u8bba\u6587\u9898\u76ee\uff1a<\/span><\/p>\n

Intelligent problem-solving as integrated hierarchical reinforcement learning<\/span><\/p>\n

\u8bba\u6587\u5730\u5740\uff1a<\/span><\/p>\n

https:\/\/www.nature.com\/articles\/s42256-021-00433-9<\/span><\/p>\n<\/section>\n<\/section>\n<\/blockquote>\n


<\/p>\n

\n
\n
\n
\n
\n

<\/section>\n

<\/section>\n
\n
<\/svg><\/section>\n<\/section>\n<\/section>\n
\n

1. \u6458\u8981<\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n


<\/p>\n

\u8ba4\u77e5\u5fc3\u7406\u5b66\u548c\u76f8\u5173\u5b66\u79d1\u6307\u51fa\uff0c\u751f\u7269\u667a\u80fd\u4f53\u89e3\u51b3\u590d\u6742\u95ee\u9898\u7684\u80fd\u529b\u7684\u53d1\u5c55\uff0c\u4f9d\u8d56\u4e8e\u5c42\u7ea7\u5316\u7684\u8ba4\u77e5\u673a\u5236\u3002\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60<\/span>\uff08hierarchical reinforcement learning\uff09<\/span>\u662f\u4e00\u79cd\u5f88\u6709\u524d\u666f\u7684\u8ba1\u7b97\u65b9\u6cd5\uff0c\u6700\u7ec8\u53ef\u80fd\u4f1a\u5728\u4eba\u5de5\u667a\u80fd\u548c\u673a\u5668\u4eba\u8eab\u4e0a\u4ea7\u751f\u7c7b\u4f3c\u7684\u89e3\u51b3\u95ee\u9898\u7684\u80fd\u529b\u3002\u7136\u800c\uff0c\u81f3\u4eca\u4e3a\u6b62\uff0c\u8bb8\u591a\u4eba\u7c7b\u548c\u975e\u4eba\u7c7b\u52a8\u7269\u7684\u89e3\u51b3\u95ee\u9898\u7684\u80fd\u529b\u663e\u7136\u4f18\u4e8e\u4eba\u5de5\u7cfb\u7edf\u3002\u5728\u672c\u6587\u4e2d\uff0c\u6211\u4eec\u63d0\u51fa\u4e86\u5c06\u53d7\u751f\u7269\u542f\u53d1\u7684\u5c42\u7ea7\u5316\u8ba4\u77e5\u673a\u5236\u6574\u5408\u8d77\u6765\u7684\u51e0\u4e2a\u6b65\u9aa4\uff0c\u6765\u8d4b\u4e88\u4eba\u5de5\u667a\u80fd\u4f53\u5148\u8fdb\u7684\u89e3\u51b3\u95ee\u9898\u6280\u80fd\u3002\u6211\u4eec\u9996\u5148\u56de\u987e\u4e86\u8ba4\u77e5\u5fc3\u7406\u5b66\u7684\u6587\u732e\uff0c\u5f3a\u8c03\u7ec4\u5408\u5f0f\u62bd\u8c61<\/span>\uff08 compositional abstraction\uff09<\/span>\u548c\u9884\u6d4b\u5904\u7406<\/span>\uff08predictive processing\uff09<\/span>\u7684\u91cd\u8981\u6027\u3002\u7136\u540e\uff0c\u6211\u4eec\u5c06\u6240\u83b7\u5f97\u7684\u89c1\u89e3\u4e0e\u5f53\u4ee3\u7684\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\u65b9\u6cd5\u8054\u7cfb\u8d77\u6765\u3002\u6709\u8da3\u7684\u662f\uff0c\u6211\u4eec\u7684\u7814\u7a76\u7ed3\u679c\u8868\u660e\uff0c\u6240\u8bc6\u522b\u7684\u6240\u6709\u8ba4\u77e5\u673a\u5236\uff0c\u90fd\u5df2\u7ecf\u5728\u4e0d\u540c\u7684\u8ba1\u7b97\u67b6\u6784\u4e2d\u5206\u522b\u5b9e\u65bd\u8fc7\u4e86\u3002\u8fd9\u4fbf\u5f15\u51fa\u4e86\u4e00\u4e2a\u95ee\u9898\uff1a\u4e3a\u4ec0\u4e48\u6ca1\u6709\u4e00\u4e2a\u7edf\u4e00\u7684\u67b6\u6784\u53bb\u96c6\u6210\u5b83\u4eec\uff1f\u5bf9\u8be5\u95ee\u9898\u7684\u56de\u7b54\u662f\u672c\u6587\u6700\u540e\u4e00\u4e2a\u8d21\u732e\u3002\u6211\u4eec\u63d0\u4f9b\u4e86\u4e00\u4e2a\u65b0\u7684\u7efc\u5408\u7684\u89c6\u89d2\uff0c\u6765\u8bf4\u660e\u5f62\u6210\u8fd9\u6837\u4e00\u4e2a\u7edf\u4e00\u7684\u7ed3\u6784\uff0c\u9700\u8981\u89e3\u51b3\u7684\u8ba1\u7b97\u6027\u6311\u6218\u3002\u6211\u4eec\u5e0c\u671b\u8be5\u7efc\u8ff0\u80fd\u591f\u5f15\u5bfc\u53d7\u66f4\u590d\u6742\u7684\u8ba4\u77e5\u7ed3\u6784\u542f\u53d1\u7684\u5c42\u7ea7\u5316\u673a\u5668\u5b66\u4e60\u67b6\u6784\u3002<\/span><\/p>\n


<\/p>\n

\u4eba\u7c7b\u548c\u5176\u5b83\u667a\u6167\u52a8\u7269\uff0c\u80fd\u591f\u5c06\u590d\u6742\u95ee\u9898\u62c6\u89e3\u4e3a\u7b80\u5355\u7684\u3001\u4e4b\u524d\u5b66\u8fc7\u7684\u5b50\u95ee\u9898\u3002\u8fd9\u79cd\u5c42\u7ea7\u5316\u7684\u65b9\u5f0f\uff0c\u4f7f\u4ed6\u4eec\u80fd\u591f\u4e00\u6b21\u89e3\u51b3\u4e4b\u524d\u6ca1\u6709\u9047\u5230\u7684\u95ee\u9898\uff0c\u5373\uff0c\u4e0d\u9700\u8981\u4efb\u4f55\u8bd5\u9519\u3002\u4f8b\u5982\uff0c\u56fe1\u63cf\u8ff0\u4e86\u4e4c\u9e26\u662f\u5982\u4f55\u91c7\u75283\u4e2a\u6709\u56e0\u679c\u8fde\u63a5\u7684\u6b65\u9aa4\uff0c\u89e3\u51b3\u4e86\u4e00\u4e2a\u590d\u6742\u7684\u98df\u7269\u83b7\u53d6\u96be\u9898\u7684\uff1a\u5b83\u9996\u5148\u6361\u8d77\u4e00\u4e2a\u68cd\u5b50\uff0c\u7528\u68cd\u5b50\u83b7\u5f97\u77f3\u5934\uff0c\u4e4b\u540e\u4f7f\u7528\u77f3\u5934\u89e6\u53d1\u673a\u5173\u91ca\u653e\u98df\u7269[1]\u3002\u8bb8\u591a\u7c7b\u4f3c\u7684\u5b9e\u9a8c\uff0c\u8bc1\u660e\u4e86\u7075\u957f\u7c7b\u52a8\u7269\u3001\u7ae0\u9c7c\uff0c\u5f53\u7136\u8fd8\u6709\u4eba\u7c7b\u7684\u7c7b\u4f3c\u80fd\u529b[2,3]\u3002\u4eba\u7c7b\u548c\u52a8\u7269\u8ba4\u77e5\u7814\u7a76\u6307\u51fa\uff0c\u5c42\u7ea7\u5316\u7684\u5b66\u4e60\u548c\u89e3\u51b3\u95ee\u9898\u7684\u80fd\u529b\uff0c\u5bf9\u5efa\u7acb\u89e3\u51b3\u590d\u6742\u95ee\u9898\u7684\u80fd\u529b\u81f3\u5173\u91cd\u8981[2,4,5]\u3002\u8fd9\u5e26\u6765\u4e86\u4e00\u4e2a\u95ee\u9898\uff1a\u6211\u4eec\u5982\u4f55\u80fd\u591f\u8ba9\u667a\u80fd\u4f53\u548c\u673a\u5668\u4eba\uff0c\u4e5f\u5177\u6709\u7c7b\u4f3c\u7684\u5c42\u7ea7\u5316\u5b66\u4e60\u548c\u96f6\u6837\u672c\u95ee\u9898\u89e3\u51b3<\/span>\uff08zero-shot problem-solving\uff09<\/span>\u7684\u80fd\u529b\u5462\uff1f<\/span><\/p>\n

        <\/span><\/p>\n

<\/span><\/section>\n
\u56fe1. \u4e00\u53ea\u65b0\u5580\u91cc\u591a\u5c3c\u4e9a\u4e4c\u9e26\u89e3\u51b3\u4e86\u4e00\u4e2a\u83b7\u53d6\u98df\u7269\u7684\u95ee\u9898\u3002a. \u9996\u5148\uff0c\u8fd9\u53ea\u4e4c\u9e26\u6361\u8d77\u4e00\u6839\u68cd\u5b50\u3002b.c. \u7136\u540e\uff0c\u5b83\u901a\u8fc7\u68cd\u5b50\u4ece\u4e00\u4e2a\u7ba1\u9053\u91cc\u62d6\u51fa\u4e00\u5757\u77f3\u5934\u3002d.e. \u6700\u540e\uff0c\u5b83\u7528\u77f3\u5934\u6765\u89e6\u53d1\u673a\u5173\uff0c\u83b7\u53d6\u98df\u7269\u3002\u5c3d\u7ba1\u4e4c\u9e26\u4ee5\u524d\u4ece\u672a\u89e3\u51b3\u8fc7\u8fd9\u79cd\u95ee\u9898\uff0c\u4f46\u5728\u77ed\u6682\u7684\u8003\u5bdf\u540e\uff0c\u5b83\u7acb\u5373\u80fd\u591f\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\u3002\u8fd9\u6837\u7684\u96f6\u6837\u672c\u95ee\u9898\u89e3\u51b3\uff0c\u662f\u5c42\u7ea7\u5316\u7684\u3001\u6709\u76ee\u6807\u7684\u8ba1\u5212\u7684\u4e00\u4e2a\u5178\u578b\u4f8b\u5b50\uff0c\u5b83\u6d89\u53ca\u5230\u5fc3\u7406\u6a21\u62df\uff0c\u5e76\u4f9d\u8d56\u4e8e\u9002\u5408\u4efb\u52a1\u7684\u3001\u7ec4\u5408\u5f0f\u7684\u62bd\u8c61\u3002\u4f5c\u8005\u7684\u63a8\u6d4b\u662f\uff0c\u56e0\u4e3a\u4e4c\u9e26\u4e00\u6b21\u53ea\u80fd\u770b\u5230\u7acb\u65b9\u4f53\u8bbe\u7f6e\u7684\u4e00\u4e2a\u4fa7\u9762\uff0c\u5b83\u5fc5\u987b\u63d0\u524d\u8ba1\u5212[1]\u3002\u56e0\u6b64\uff0c\u4e4c\u9e26\u5fc5\u987b\u5728\u5fc3\u667a\u4e0a\u6a21\u62df\u672a\u6765\u7684\u6b65\u9aa4\u3002\u5176\u4e2d\u8fd8\u6d89\u53ca\u8fc1\u79fb\u5b66\u4e60\uff1a\u8fc1\u79fb\u5e76\u7ed3\u5408\u4ee5\u524d\u5b66\u5230\u7684\u7c7b\u4f3c\u5b50\u95ee\u9898\u7684\u89e3\u51b3\u65b9\u6848\uff0c\u6765\u89e3\u51b3\u65b0\u95ee\u9898\u3002\u6210\u529f\u7684\u8ba1\u5212\u548c\u8f6c\u79fb\u5b66\u4e60\u53ea\u6709\u5728\u62bd\u8c61\u7684\u7ec4\u5408\u5f0f\u7684\u5fc3\u7406\u8868\u5f81\u53d1\u5c55\u8d77\u6765\u540e\u624d\u53ef\u80fd\u53d1\u751f\uff0c\u4f8b\u5982\uff0c\u8fd9\u4e9b\u8868\u5f81\u53ef\u4ee5\u7f16\u7801\u5de5\u5177\u64cd\u4f5c\u9009\u9879\u3002\u7136\u800c\uff0c\u4e3a\u4e86\u5b66\u4e60\u4e0d\u540c\u7684\u5de5\u5177\u64cd\u4f5c\uff0c\u4e4c\u9e26\u4f9d\u9760\u5176\u5185\u5728\u52a8\u673a\uff0c\u901a\u8fc7\u73a9\u800d\u548c\u8bd5\u9a8c\u5de5\u5177\u6765\u83b7\u5f97\u77e5\u8bc6\u3002\u56fe\u7247\u7531R. Gruber\u548cA. Taylor\u63d0\u4f9b[1]\u3002<\/span><\/section>\n


<\/p>\n

\u6211\u4eec\u8bd5\u56fe\u5728\u5f3a\u5316\u5b66\u4e60\u7684\u6846\u67b6\u4e0b\uff0c\u89e3\u51b3\u8be5\u95ee\u9898[6-8]\u3002\u591a\u9879\u7814\u7a76\u8868\u660e\uff0c\u5f3a\u5316\u5b66\u4e60\u5728\u751f\u7269\u4e0a\u548c\u8ba4\u77e5\u4e0a\u662f\u53ef\u884c\u7684\u3002\u5f88\u591a\u73b0\u6709\u7684\u57fa\u4e8e\u5f3a\u5316\u5b66\u4e60\u7684\u65b9\u6cd5\uff0c\u8bd5\u56fe\u8fdb\u884c\u96f6\u6837\u672c\u95ee\u9898\u89e3\u51b3\u548c\u8fc1\u79fb\u5b66\u4e60[8,9]\uff0c\u4f46\u76ee\u524d\uff0c\u8fd9\u53ea\u9002\u7528\u4e8e\u4e0e\u4e4b\u524d\u7684\u4efb\u52a1\u76f8\u540c\u6216\u8005\u76f8\u8fd1\u7684\u4efb\u52a1[10]\uff0c\u6216\u662f\u5728\u7b80\u5355\u751f\u6210\u57df\u4e2d\u7684\u4efb\u52a1[11,12]\uff0c\u4f8b\u5982\u5728\u4e8c\u7ef4\u7684\u7f51\u683c\u72b6\u4e16\u754c\u4e2d\u3002\u7c7b\u4f3c\u56fe1\u4e2d\u4e4c\u9e26\u7684\uff0c\u5728\u8fde\u7eed\u7a7a\u95f4\u4e2d\u7684\u89e3\u51b3\u95ee\u9898\u7684\u884c\u4e3a\uff0c\u8fd8\u6ca1\u6709\u5728\u4efb\u4f55\u4eba\u5de5\u7cfb\u7edf\u4e2d\u5b9e\u73b0\u3002\u9274\u4e8e\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\u6240\u5360\u6bd4\u4f8b\u8f83\u5c0f\uff0c\u6211\u4eec\u8ba4\u4e3a\uff0c\u5bf9\u5c42\u7ea7\u5316\u5b66\u4e60\u548c\u95ee\u9898\u89e3\u51b3\u7684\u5173\u6ce8\u662f\u4e0d\u591f\u7684\u3002<\/span><\/p>\n


<\/p>\n

\u5728\u8be5\u7efc\u8ff0\u4e2d\uff0c\u6211\u4eec\u901a\u8fc7\u4e09\u6b65\u6765\u5f25\u8865\u4e0a\u8ff0\u7684\u5dee\u8ddd\u3002\u9996\u5148\uff0c\u6211\u4eec\u8bc4\u4f30\u4e86\u5c42\u7ea7\u5316\u51b3\u7b56\u7684\u795e\u7ecf\u8ba4\u77e5\u57fa\u7840\uff0c\u5e76\u8bc6\u522b\u51fa\u4f7f\u9ad8\u7ea7\u95ee\u9898\u89e3\u51b3\u80fd\u529b\u5f97\u4ee5\u5b9e\u73b0\u7684\u91cd\u8981\u673a\u5236\u3002\u4e4b\u540e\uff0c\u6211\u4eec\u63ed\u793a\u4e86\u5f53\u524d\u7684\u5c42\u7ea7\u5f3a\u5316\u5b66\u4e60\u65b9\u6cd5\u7684\u6574\u5408\u95ee\u9898\uff0c\u6307\u51fa\u5927\u90e8\u5206\u5df2\u88ab\u5b9e\u73b0\u7684\u751f\u7269\u5b66\u673a\u5236\uff0c\u90fd\u662f\u4ee5\u5b64\u7acb\u800c\u975e\u6574\u5408\u7684\u65b9\u5f0f\u5b9e\u73b0\u7684\u3002\u7b2c\u4e09\uff0c\u6211\u4eec\u6307\u51fa\u6574\u5408\u5173\u952e\u65b9\u6cd5\u548c\u673a\u5236\u7684\u6b65\u9aa4\uff0c\u4ee5\u514b\u670d\u4e0a\u8ff0\u6574\u5408\u95ee\u9898\uff0c\u53d1\u5c55\u4e00\u4e2a\u7edf\u4e00\u7684\u8ba4\u77e5\u6846\u67b6\u3002<\/span><\/p>\n


<\/p>\n


<\/p>\n

\n
\n
\n
\n
\n

<\/section>\n

<\/section>\n
\n
<\/svg><\/section>\n<\/section>\n<\/section>\n
\n

2. \u795e\u7ecf\u8ba4\u77e5\u57fa\u7840<\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n


<\/span><\/p>\n

\u4e0b\u6587\u6307\u51fa\uff1a\u751f\u7269\u4f53\u7684\u590d\u6742\u7684\u95ee\u9898\u89e3\u51b3\u6280\u80fd\uff0c\u53ef\u4ee5\u7528\u51e0\u79cd\u7279\u5b9a\u7684\u8ba4\u77e5\u80fd\u529b\u6e05\u6670\u63cf\u8ff0\uff0c\u8fd9\u4e9b\u80fd\u529b\u4f9d\u8d56\u4e8e\u51e0\u9879\u6838\u5fc3\u8ba4\u77e5\u673a\u5236\uff0c\u5305\u62ec\u62bd\u8c61<\/span>\uff08abstraction\uff09<\/span>\uff0c\u5185\u5728\u52a8\u673a<\/span>\uff08intrinsic motivation\uff09<\/span>\u548c\u5fc3\u667a\u6a21\u62df<\/span>\uff08mental simulation\uff09<\/span>\u3002\u56fe2\u5c55\u793a\u4e86\u8fd9\u4e9b\u673a\u5236\u5982\u4f55\u5728\u524d\u5411\u6216\u9006\u5411\u7684\u6a21\u578b<\/span>\uff08\u4f5c\u4e3a\u795e\u7ecf\u6d3b\u52a8\u7684\u5148\u51b3\u6761\u4ef6\uff09<\/span>\u4e2d\u53d1\u5c55\u51fa\u6765\u3002<\/span><\/p>\n

        <\/span>
<\/p>\n

<\/span><\/p>\n

\u56fe2. \u751f\u7269\u4f53\u89e3\u51b3\u95ee\u9898\u7684\u5148\u51b3\u6761\u4ef6\u3001\u673a\u5236\u548c\u7279\u5f81\u3002\u524d\u5411\u548c\u53cd\u5411\u6a21\u578b\u662f\u9ad8\u9636\u8ba4\u77e5\u673a\u5236\u548c\u80fd\u529b\u7684\u5148\u51b3\u6761\u4ef6\u3002\u7ec4\u6210\u5f0f\u62bd\u8c61\uff0c\u5141\u8bb8\u667a\u80fd\u4f53\u5c06\u95ee\u9898\u5206\u89e3\u4e3a\u53ef\u590d\u7528\u7684\u5b50\u7ed3\u6784\uff0c\u4f8b\u5982\uff0c\u5b66\u4e60\u5230\u7269\u4f53\u7684\u54ea\u4e00\u90e8\u5206\u662f\u6613\u4e8e\u6293\u53d6\u7684\uff0c\u6bd4\u5982\u73bb\u7483\u676f\u7684\u676f\u9888\u3002\u5728\u524d\u5411\u6a21\u578b\u7684\u5e2e\u52a9\u4e0b\uff0c\u5185\u5728\u52a8\u673a\u53ef\u4ee5\u9a71\u52a8\u667a\u80fd\u4f53\uff0c\u4ee5\u4e00\u79cd\u641c\u7d22\u4fe1\u606f\u7684\u3001\u8ba4\u77e5\u7684\u65b9\u5f0f\u4e0e\u73af\u5883\u4e92\u52a8\u3002\u4f8b\u5982\uff0c\u5185\u5728\u52a8\u673a\u5f15\u5bfc\u667a\u80fd\u4f53\u4e0e\u676f\u5b50\u4e92\u52a8\uff0c\u6765\u4e86\u89e3\u676f\u5b50\u7279\u5f81\u3002\u5fc3\u667a\u6a21\u62df\u4f7f\u5f97\u63a2\u7d22\u53ef\u80fd\u7684\u72b6\u6001-\u884c\u4e3a\u987a\u5e8f\u6210\u4e3a\u53ef\u80fd\uff0c\u4f8b\u5982\uff0c\u60f3\u8c61\u4f38\u5411\u73bb\u7483\u676f\u7684\u624b\u7684\u4f4d\u7f6e\u5c06\u5982\u4f55\u6f14\u53d8\u3002\u5728\u4e0a\u8ff0\u673a\u5236\u7684\u5e2e\u52a9\u4e0b\uff0c\u667a\u80fd\u4f53\u80fd\u591f\u7ecf\u7531\u5c42\u7ea7\u5316\u7684\u3001\u7ec4\u5408\u5f0f\u7684\u76ee\u6807\u9a71\u52a8\u89c4\u5212\uff0c\u7075\u6d3b\u5730\u8fbe\u5230\u60f3\u8981\u7684\u72b6\u6001\uff0c\u4f8b\u5982\uff0c\u559d\u6c34\u7684\u89c4\u5212\u662f\u5148\u6b63\u786e\u5730\u6293\u53d6\u676f\u5b50\uff0c\u518d\u59a5\u5f53\u5730\u79fb\u52a8\u676f\u5b50\u3002\u7ec4\u5408\u5f0f\u62bd\u8c61\u4f7f\u5f97\u667a\u80fd\u4f53\u80fd\u591f\u8fdb\u884c\u8fc1\u79fb\u5b66\u4e60\uff0c\u4f8b\u5982\uff0c\u50cf\u6293\u53d6\u5176\u5b83\u6709\u67c4\u7684\u7269\u4f53\uff08\u5982\u8336\u58f6\uff09\u4e00\u6837\uff0c\u53bb\u9002\u5f53\u5730\u6293\u53d6\u676f\u676f\u9888\uff0c\u4ece\u800c\u4f7f\u5c0f\u6837\u672c\u95ee\u9898\u89e3\u51b3\u6210\u4e3a\u53ef\u80fd\u3002<\/span><\/p>\n


<\/p>\n


<\/p>\n

\n
\n
\n
\n
\n

<\/section>\n

<\/section>\n
\n
<\/svg><\/section>\n<\/section>\n<\/section>\n
\n

3. \u8ba4\u77e5\u80fd\u529b<\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n


<\/p>\n

\u6211\u4eec\u5173\u6ce8\u8ba4\u77e5\u80fd\u529b\u7684\u4ee5\u4e0b\u7279\u5f81\u548c\u6027\u8d28\uff1a<\/span><\/p>\n


<\/span><\/p>\n

\u5c0f\u6837\u672c\u95ee\u9898\u89e3\u51b3\uff08Few-shot problem-solving\uff09<\/span><\/strong>\uff1a\u5c0f\u6837\u672c\u95ee\u9898\u89e3\u51b3\uff0c\u6307\u5728\u5c11\u4e8e5\u6b21\u7684\u5c1d\u8bd5\u4e0b\uff0c\u89e3\u51b3\u672a\u77e5\u95ee\u9898\u7684\u80fd\u529b\u3002\u800c\u96f6\u6837\u672c\u95ee\u9898\u89e3\u51b3\uff0c\u662f\u5c0f\u6837\u672c\u95ee\u9898\u89e3\u51b3\u7684\u4e00\u79cd\u7279\u6b8a\u60c5\u51b5\uff0c\u6307\u4e0d\u9700\u8981\u989d\u5916\u5c1d\u8bd5\u5373\u53ef\u89e3\u51b3\u65b0\u95ee\u9898\u3002\u4f8b\u5982\uff0c\u5728\u4e4b\u524d\u89e3\u51b3\u8fc7\u76f8\u5173\u95ee\u9898\uff0c\u4f46\u6ca1\u6709\u7ecf\u8fc7\u8fdb\u4e00\u6b65\u8bad\u7ec3\u7684\u60c5\u51b5\u4e0b\uff0c\u4e4c\u9e26\u53ef\u4ee5\u4f7f\u7528\u6728\u68cd\u4f5c\u4e3a\u5de5\u5177\uff0c\u89e3\u51b3\u5168\u65b0\u7684\u98df\u7269\u83b7\u53d6\u95ee\u9898<\/span>\uff08\u56fe1\uff09<\/span>\u3002\u6211\u4eec\u6307\u51fa\uff0c\u8981\u8fd9\u6837\u89e3\u51b3\u95ee\u9898\uff0c\u6709\u4e24\u79cd\u80fd\u529b\u81f3\u5173\u91cd\u8981\uff1a\u8fc1\u79fb\u5b66\u4e60<\/span>\uff08transfer learning\uff09<\/span>\u548c\u51b3\u7b56\u65f6\u89c4\u5212<\/span>\uff08decision-time planning\uff09<\/span>\u3002<\/span><\/p>\n


<\/span><\/p>\n

\u8fc1\u79fb\u5b66\u4e60\uff1a<\/span><\/strong>\u8fc1\u79fb\u5b66\u4e60\u5141\u8bb8\u751f\u7269\u4f53\u5c06\u4e4b\u524d\u4efb\u52a1\u7684\u89e3\u6cd5\uff0c\u8fc1\u79fb\u5230\u4e00\u4e2a\u65b0\u7684\u4f46\u76f8\u4f3c\u7684\u4efb\u52a1\u4e0a\uff0c\u4ece\u800c\u4f7f\u5c0f\u6837\u672c\u7684\u95ee\u9898\u89e3\u51b3\u6210\u4e3a\u53ef\u80fd\u3002\u8fd9\u663e\u8457\u5730\u51cf\u5c11\u4e86\uff0c\u6709\u65f6\u751a\u81f3\u6d88\u9664\u4e86\u89e3\u51b3\u4e00\u4e2a\u65b0\u95ee\u9898\u6240\u9700\u8981\u5c1d\u8bd5\u7684\u6b21\u6570\u3002<\/span><\/p>\n


<\/p>\n

\u8fd9\u6837\u7684\u7c7b\u6bd4\u63a8\u7406\uff0c\u901a\u5e38\u88ab\u8ba4\u4e3a\u662f\u4eba\u7c7b\u8ba4\u77e5\u7684\u5173\u952e\u6027\u57fa\u77f3\u3002\u4f8b\u5982\uff0c\u8ba4\u77e5\u5b66\u7406\u8bba\u6307\u51fa\uff0c\u4eba\u7c7b\u662f\u901a\u8fc7\u7c7b\u6bd4\u5176\u5b83\u9886\u57df\u7684\u77e5\u8bc6\uff0c\u6765\u7406\u89e3\u673a\u68b0\u7cfb\u7edf\u7684[13]\u3002\u9664\u6b64\u4e4b\u5916\uff0c\u6559\u80b2\u7406\u8bba\u6307\u51fa\uff0c\u4eba\u7c7b\u5728\u660e\u786e\u5730\u63a5\u53d7\u8bc6\u522b\u4e0d\u540c\u95ee\u9898\u95f4\u76f8\u4f3c\u6027\u7684\u8bad\u7ec3\u540e\uff0c\u8fc1\u79fb\u5b66\u4e60\u80fd\u529b\u4f1a\u63d0\u9ad8[14]\u3002\u7136\u800c\uff0c\u5c3d\u7ba1\u4eba\u7c7b\u80fd\u591f\u719f\u7ec3\u5730\u7528\u7c7b\u6bd4\u6765\u89e3\u51b3\u95ee\u9898\uff0c\u8981\u53d1\u73b0\u65b0\u7684\u7c7b\u6bd4\uff0c\u5374\u597d\u50cf\u5f88\u56f0\u96be\u3002\u8fd9\u4fbf\u662f\u529f\u80fd\u56fa\u5316<\/span>\uff08functional fixedness\uff09<\/span>\u7684\u73b0\u8c61\uff1a\u4eba\u7c7b\u5728\u89e3\u51b3\u95ee\u9898\u65f6\uff0c\u503e\u5411\u4e8e\u4ee5\u4e60\u4ee5\u4e3a\u5e38\u7684\u65b9\u6cd5\u53bb\u4f7f\u7528\u7269\u4f53[15]\u3002<\/span><\/p>\n


<\/p>\n

\u76ee\u6807\u5bfc\u5411\u7684\u89c4\u5212\uff08goal-directed planning\uff09<\/span><\/strong>\u3002\u884c\u4e3a\u4f20\u7edf\u4e0a\u5206\u4e3a\u4e24\u7c7b\uff0c\u523a\u6fc0\u9a71\u52a8\u7684\u4e60\u60ef\u6027\u884c\u4e3a\uff0c\u4ee5\u53ca\u76ee\u6807\u5bfc\u5411\u7684\uff0c\u89c4\u5212\u597d\u7684\u884c\u4e3a[16-18]\u3002\u4e60\u60ef\u6027\u884c\u4e3a\u6307\u9ad8\u5ea6\u81ea\u52a8\u5316\u7684\uff0c\u8ba1\u7b97\u9ad8\u6548\u7684\uff0c\u4e3b\u8981\u4ece\u4e4b\u524d\u7684\u5f3a\u5316\u8fc7\u7a0b\u4e2d\u5b66\u5230\u7684\u5e94\u6fc0\u6027\u63a7\u5236[17]\u3002\u5bf9\u8be5\u65b9\u5f0f\u7684\u4e00\u79cd\u6279\u5224\u6765\u81eaTolman[19,20]\uff0c\u4ed6\u6307\u51fa\uff0c\u5927\u9f20\u5728\u8ff7\u5bab\u4e2d\u63a2\u7d22\u4e00\u6b21\u540e\uff0c\u80fd\u591f\u66f4\u5feb\u5730\u627e\u5230\u98df\u7269\uff0c\u5373\u4f7f\u6700\u521d\u7684\u63a2\u7d22\u884c\u4e3a\u6ca1\u6709\u5e26\u6765\u5956\u52b1\u3002\u8fd9\u4e00\u7ed3\u679c\u6307\u51fa\uff0c\u5927\u9f20\u5bf9\u8ff7\u5bab\u6784\u5efa\u4e86\u4e00\u4e2a\u8868\u5f81<\/span>\uff08representation\uff09<\/span>\u6216\u8ba4\u77e5\u5730\u56fe<\/span>\uff08cognitive map\uff09<\/span>\uff0c\u4f7f\u5f97\u5b83\u4eec\u80fd\u591f\u5728\u5956\u52b1<\/span>\uff08reward\uff09<\/span>\u51fa\u73b0\u65f6\uff0c\u89c4\u5212\u5b83\u4eec\u7684\u884c\u4e3a\u3002\u5728\u7406\u89e3\u76ee\u6807\u5bfc\u5411\u7684\u89c4\u5212\u65b9\u9762\uff0c\u4eba\u5de5\u667a\u80fd\u548c\u8ba4\u77e5\u79d1\u5b66\u7684\u8fdb\u5c55\u76f8\u4e92\u5f71\u54cd\u3002\u5bf9\u4e8e\u52a8\u4f5c\u63a7\u5236\uff0c\u4e00\u4e2a\u6709\u5f71\u54cd\u529b\u7684\u6a21\u578b\u662fTOTE<\/span>\uff08test\u2013operate\u2013test\u2013exit\uff0c\u6d4b\u8bd5-\u52a8\u4f5c-\u6d4b\u8bd5-\u79bb\u5f00\uff09<\/span>[21]\uff0c\u5176\u4e2d\uff0c\u52a8\u4f5c\u662f\u57fa\u4e8e\u4e00\u4e2a\u8fed\u4ee3\u7684\u3001\u5c42\u7ea7\u5316\u7684\u53cd\u9988\u5faa\u73af\u8fdb\u884c\u9009\u62e9\u7684\u3002\u5728\u8fd9\u91cc\uff0c\u60f3\u8981\u7684\u7684\u9884\u671f\u72b6\u6001\u4e0e\u5f53\u524d\u7684\u611f\u89c9\u72b6\u6001\u8fdb\u884c\u6bd4\u8f83\uff0c\u6765\u9009\u62e9\u4e0b\u4e00\u4e2a\u52a8\u4f5c\u3002\u7c7b\u4f3cTOTE\uff0c\u5f88\u591a\u6700\u65b0\u7684\u7406\u8bba\uff0c\u90fd\u5c06\u89c4\u5212\u770b\u4f5c\u4e00\u79cd\u5c42\u7ea7\u5316\u7684\u8fc7\u7a0b\uff0c\u6d89\u53ca\u4e0d\u540c\u5c42\u7ea7\u7684\u62bd\u8c61[5,22,23]\u3002<\/span><\/p>\n


<\/p>\n

\u4f46\u89c4\u5212\u8fc7\u7a0b\u8003\u8651\u4e86\u4ec0\u4e48\u6837\u7684\u4fe1\u606f\u5462\uff1f\u52a8\u4f5c\u751f\u6210\u7684\u89c2\u5ff5\u8fd0\u52a8<\/span>\uff08Ideomotor\uff09<\/span>\u7406\u8bba\u6307\u51fa\uff0c\u9884\u671f\u52a8\u4f5c\u7684\u6548\u679c\u51b3\u5b9a\u4e86\u52a8\u4f5c\u7684\u9009\u62e9\u3002\u4e0e\u6b64\u76f8\u5bf9\u5e94\uff0c\u4e8b\u4ef6\u7f16\u7801<\/span>\uff08event encoding\uff09<\/span>\u7406\u8bba[25]\u6307\u51fa\uff0c\u4e00\u4e2a\u52a8\u4f5c\u7684\u8868\u5f81\u548c\u5b83\u4eec\u7684\u6548\u679c\u5b58\u50a8\u5728\u5171\u4eab\u7684\u7f16\u7801\u4e2d\uff0c\u5bf9\u52a8\u4f5c-\u6548\u679c\u7684\u9884\u671f\uff0c\u53ef\u4ee5\u76f4\u63a5\u6fc0\u6d3b\u76f8\u5173\u52a8\u4f5c\u7684\u7f16\u7801\u3002\u4e0a\u8ff0\u52a8\u4f5c\u4ea7\u751f\u548c\u6548\u679c\u9884\u6d4b\u4e4b\u95f4\u7684\u8054\u7cfb\uff0c\u8fd8\u53d7\u5230\u4e86\u4e00\u7cfb\u5217\u5b9e\u9a8c\u8bc1\u636e\u7684\u652f\u6301[25-27]\u3002<\/span><\/p>\n


<\/p>\n


<\/p>\n

\n
\n
\n
\n
\n

<\/section>\n

<\/section>\n
\n
<\/svg><\/section>\n<\/section>\n<\/section>\n
\n

4. \u8fc1\u79fb\u5b66\u4e60\u548c\u89c4\u5212\u7684\u8ba4\u77e5\u673a\u5236<\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n


<\/p>\n

\u524d\u6587\u8ba8\u8bba\u4e86\u8fc1\u79fb\u5b66\u4e60\u548c\u89c4\u5212\u80fd\u529b\u662f\u5982\u4f55\u8ba9\u667a\u80fd\u4f53\u5177\u6709\u5c0f\u6837\u672c\u95ee\u9898\u89e3\u51b3\u80fd\u529b\u7684\u3002\u4f46\u662f\u4ec0\u4e48\u4f7f\u8fc1\u79fb\u5b66\u4e60\u548c\u89c4\u5212\u80fd\u529b\u6210\u4e3a\u53ef\u80fd\u7684\uff1f\u6211\u4eec\u7684\u8c03\u7814\u6307\u51fa\uff0c\u8fc1\u79fb\u5b66\u4e60\u548c\u89c4\u5212\u6240\u9700\u7684\u4e09\u79cd\u8ba4\u77e5\u673a\u5236\u662f\u611f\u89c9\u8fd0\u52a8\u62bd\u8c61<\/span>\uff08sensorimotor abstraction\uff09<\/span>\uff0c\u5185\u5728\u52a8\u673a<\/span>\uff08intrinsic motivation\uff09<\/span>\u548c\u5fc3\u667a\u6a21\u62df<\/span>\uff08mental simulation\uff09<\/span>\u3002<\/span><\/p>\n


<\/p>\n

\u611f\u77e5\u8fd0\u52a8\u7684\u62bd\u8c61\u3002\u6839\u636e\u5177\u8eab\u8ba4\u77e5<\/span>\uff08embodied cognition\uff09<\/span>\u7406\u8bba\uff0c\u62bd\u8c61\u7684\u5fc3\u667a\u6982\u5ff5\uff0c\u53ef\u4ee5\u4ece\u57fa\u4e8e\u8eab\u4f53\u7684\uff0c\u611f\u89c9\u8fd0\u52a8\u7684\uff0c\u4e0e\u73af\u5883\u7684\u4e92\u52a8\u7ecf\u9a8c\u4e2d\u63a8\u51fa[2,28-30]\u3002\u8fd9\u91cc\uff0c\u6211\u4eec\u533a\u5206\u52a8\u4f5c\u62bd\u8c61<\/span>\uff08action abstraction\uff09<\/span>\u548c\u72b6\u6001\u62bd\u8c61<\/span>\uff08state abstraction\uff09<\/span>\u3002\u52a8\u4f5c\u62bd\u8c61\u6307\u5bf9\u4e00\u4e32\u5728\u65f6\u95f4\u4e0a\u5ef6\u5c55\u7684\u539f\u59cb\u52a8\u4f5c\u3001\u9009\u9879\u6216\u8fd0\u52a8\u57fa\u5143\u7684\u62bd\u8c61[32,33]\u3002\u4f8b\u5982\uff0c\u8fd0\u8f93\u4e00\u4e2a\u7269\u4f53\uff0c\u5c31\u662f\u52a8\u4f5c\u62bd\u8c61\uff0c\u56e0\u4e3a\u5b83\u5305\u542b\u4e86\u4e00\u8fde\u4e32\u66f4\u57fa\u672c\u7684\u52a8\u4f5c\uff0c\u4f8b\u5982\u4f38\u624b\u53bb\u591f\u7269\u4f53\u548c\u62ac\u8d77\u7269\u4f53<\/span>\uff08\u56fe3\uff09<\/span>\u3002<\/span><\/p>\n


<\/p>\n

\u72b6\u6001\u62bd\u8c61\u6307\u5bf9<\/span>\uff08\u90e8\u5206\uff09<\/span>\u73af\u5883\u7684\u7cbe\u7b80\u7f16\u7801\uff0c\u5176\u4e2d\u4e0d\u76f8\u5173\u7684\u7ec6\u8282\u88ab\u53bb\u9664\u3002\u7531\u6b64\uff0c\u72b6\u6001\u62bd\u8c61\u4f1a\u5c06\u4e00\u4e2a\u573a\u666f\u5206\u51fa\u548c\u4efb\u52a1\u76f8\u5173\u7684\u90e8\u5206\uff0c\u4f8b\u5982\u53ef\u6293\u6027\uff0c\u5bb9\u5668\u6027\u6216\u7a7a\u5fc3\u5ea6\u3002\u4f8b\u5982\uff0c\u9152\u676f\u548c\u8336\u676f\u90fd\u662f\u5bb9\u5668\uff0c\u4e5f\u90fd\u662f\u53ef\u4ee5\u6293\u53d6\u7684<\/span>\uff08\u56fe3\uff09<\/span>\u3002\u540c\u65f6\uff0c\u5bb9\u5668\u610f\u5473\u7740\u5b83\u662f\u4e2d\u7a7a\u7684\uff0c\u8fd9\u8bf4\u660e\u8fd9\u79cd\u62bd\u8c61\u6709\u5c42\u7ea7\u5316\u548c\u7ec4\u5408\u5f0f<\/span>\uff08compositional\uff09<\/span>\u7684\u7279\u5f81\u3002\u8ba4\u77e5\u8bed\u8a00\u5b66\u548c\u76f8\u5173\u7814\u7a76\u8868\u660e\uff0c\u62bd\u8c61\u80fd\u529b\u53d7\u76ca\u4e8e\u7ec4\u5408\u6027<\/span>\uff08compositionality\uff09<\/span>[34]\u3002\u6b63\u5f0f\u5730\u8bf4\uff0c\u5982\u679c\u4e00\u4e2a\u8868\u8ff0\uff0c\u662f\u7531\u5176\u5b50\u8868\u8ff0\uff0c\u548c\u51b3\u5b9a\u5b50\u8868\u8ff0\u7ec4\u5408\u8bed\u4e49\u7684\u89c4\u5219\u6240\u7ec4\u6210\uff0c\u6211\u4eec\u5c31\u79f0\u8be5\u8868\u8ff0\u5177\u6709\u7ec4\u5408\u6027\u3002\u601d\u7ef4\u8bed\u8a00\u5047\u8bf4<\/span>\uff08the language of thought theory\uff09<\/span>[35]\u5c06\u6765\u81ea\u8bed\u8a00\u7684\u7ec4\u5408\u6027\u539f\u7406\uff0c\u7528\u5230\u62bd\u8c61\u7684\u5fc3\u667a\u8868\u5f81\u4e0a\uff0c\u58f0\u79f0\u601d\u60f3\u4e5f\u5177\u6709\u7ec4\u5408\u6027\u3002\u4f8b\u5982\uff0c\u4eba\u7c7b\u53ef\u4ee5\u4ece\u8bed\u8a00\u63cf\u8ff0\u4e2d\uff0c\u5c06\u4ed6\u4eec\u5df2\u77e5\u7684\u6210\u5206\u8fdb\u884c\u7ec4\u5408\uff0c\u8f7b\u6613\u5730\u60f3\u8c61\u4e4b\u524d\u6ca1\u6709\u89c1\u8fc7\u7684\u5f71\u50cf[36]\uff0c\u5373\u4f7f\u662f\u901a\u8fc7\u5f88\u5947\u602a\u7684\u63cf\u8ff0\u2014\u2014\u60f3\u60f3\u4e00\u6761\u9a91\u7740\u6ed1\u677f\u7684\u7c89\u7ea2\u8272\u7684\u9f99\u4f1a\u662f\u4ec0\u4e48\u6837\u5b50\u3002<\/span><\/p>\n


<\/p>\n

       <\/span>        <\/span><\/p>\n

\u56fe3. \u7ec4\u5408\u5f0f\u7684\u52a8\u4f5c\u62bd\u8c61\u4e0e\u72b6\u6001\u62bd\u8c61\u3002a. \u52a8\u4f5c\u62bd\u8c61\u63cf\u8ff0\u4e86\u4e00\u8fde\u4e32\u539f\u59cb\u52a8\u4f5c\u3002b. \u72b6\u6001\u62bd\u8c61\u7f16\u7801\u4e86\u72b6\u6001\u7a7a\u95f4\u7684\u7279\u5b9a\u6027\u8d28\u3002c. \u62bd\u8c61\u7684\u7ec4\u5408\u6027\uff0c\u4f7f\u5f97\u5b83\u4eec\u80fd\u591f\u901a\u8fc7\u5c06\u62bd\u8c61\u7684\u5b9a\u4e49\uff08\u4f38\u624b\u53ef\u6293\u53d6\u7684\u76ee\u6807\uff09\u5b9e\u4f8b\u5316\uff08instantiate\uff09\u4e3a\u5177\u4f53\u7684\u7269\u4f53\uff08\u8336\u58f6\uff09\uff0c\u6765\u88ab\u901a\u7528\u5730\u5e94\u7528\u3002<\/span><\/section>\n


<\/p>\n

\u5bf9\u62bd\u8c61\u5fc3\u667a\u8868\u5f81\u8fdb\u884c\u7f16\u7801\uff0c\u662f\u7531\u5c40\u90e8\u7b26\u53f7\uff0c\u8fd8\u662f\u5206\u5e03\u5f0f\u5730\uff0c\u8fd8\u5b58\u5728\u4e89\u8bae[37]\uff0c\u4f46\u5728\u8fd9\u4e24\u79cd\u60c5\u51b5\u4e0b\uff0c\u90fd\u6709\u8bc1\u636e\u8868\u660e\u5b58\u5728\u7ec4\u5408\u53ef\u6cdb\u5316\u80fd\u529b\u3002\u4f8b\u5982\uff0cHaynes[38]\u6307\u51fa\uff0c\u4f4d\u4e8e\u8179\u5916\u4fa7\u524d\u989d\u53f6\u76ae\u5c42\u7684\u795e\u7ecf\u7f16\u7801\uff0c\u8bb0\u5f55\u7740\u590d\u5408\u884c\u4e3a\u7684\u89c4\u5219\uff0c\u53ef\u4ee5\u88ab\u5206\u89e3\u4e3a\u7f16\u7801\u5404\u81ea\u7ec4\u6210\u90e8\u5206\u7684\u795e\u7ecf\u7f16\u7801\u3002<\/span><\/p>\n


<\/p>\n

\u5bf9\u4e8e\u611f\u89c9\u52a8\u4f5c\u62bd\u8c61<\/span>\uff08sensorimotor abstractions\uff09<\/span>\uff0c\u7ec4\u5408\u89c4\u5219<\/span>\uff08rules of composition\uff09<\/span>\u9700\u8981\u7531\u6211\u4eec\u7684\u5fc3\u667a\u7075\u6d3b\u63a8\u65ad\uff0c\u4ece\u800c\u6839\u636e\u6211\u4eec\u7684\u4e16\u754c\u77e5\u8bc6\uff0c\u5f62\u6210\u5408\u7406\u7684\u73af\u5883\u914d\u7f6e[29,39,40]\u3002\u8fd9\u79cd\u5e38\u8bc6\u7ec4\u5408\u6027\uff0c\u4f7f\u62bd\u8c61\u80fd\u591f\u4ee5\u6709\u610f\u4e49\u7684\u65b9\u5f0f\u5e94\u7528\u3002\u5bf9\u4e00\u4e2a\u62bd\u8c61\u6210\u5206<\/span>\uff08\u4f8b\u5982\u6293\u53d6\u7684\u76ee\u6807\uff09<\/span>\u7684\u586b\u5165\u88ab\u9650\u5236\u5728\u4f7f\u7528\u7684\u5bf9\u8c61\u4e0a<\/span>\uff08\u4f8b\u5982\uff0c\u53ef\u6293\u53d6\u7684\u7269\u4f53\uff0c\u5982\u8336\u58f6\uff09<\/span>\uff0c\u5982\u56fe3\u3002\u4ece\u7b97\u6cd5\u89c6\u89d2\u6765\u770b\u3002\u5e38\u8bc6\u7ec4\u5408\u6027\u6709\u5de8\u5927\u7684\u8ba1\u7b97\u4f18\u52bf\uff1a\u62bd\u8c61\uff0c\u4f8b\u5982\u5bf9\u6293\u53d6\u884c\u4e3a\u7684\u7f16\u7801\uff0c\u5982\u679c\u80fd\u591f\u4f7f\u7528\u4e8e\u591a\u79cd\u573a\u666f<\/span>\uff08\u4f8b\u5982\u4e0d\u540c\u7684\u6293\u53d6\u76ee\u6807\uff09<\/span>\uff0c\u4f1a\u66f4\u52a0\u7ecf\u6d4e\u3002\u5e38\u8bc6\u7ec4\u5408\u6027\uff0c\u901a\u8fc7\u5728\u8868\u5f81\u95f4\u5efa\u7acb\u6620\u5c04\uff0c\u672c\u8d28\u4e0a\u7b80\u5316\u4e86\u5bf9\u7c7b\u6bd4\u7684\u8bc6\u522b\uff0c\u8fd9\u4e00\u4f18\u52bf\u5728\u5173\u4e8e\u521b\u9020\u529b\u7684\u6982\u5ff5\u6574\u5408<\/span>\uff08concept blending\uff09<\/span>\u7684\u8ba4\u77e5\u7406\u8bba\u4e2d\uff0c\u4e5f\u5e7f\u4e3a\u4eba\u77e5[41,42]\u3002<\/span><\/p>\n


<\/p>\n

\u5185\u5728\u52a8\u673a\u3002\u901a\u8fc7\u8bbe\u7acb\u5185\u5728\u52a8\u673a\u7684\u76ee\u6807\uff0c\u5185\u5728\u52a8\u673a\u5f71\u54cd\u7740\u76ee\u6807\u5bfc\u5411\u7684\u5f52\u56e0\u3002\u4ece\u8ba4\u77e5\u53d1\u5c55\u7684\u89d2\u5ea6\u6765\u770b\uff0c\u5185\u5728\u52a8\u673a\u8fd9\u4e2a\u672f\u8bed\uff0c\u662f\u7528\u6765\u63cf\u8ff0\u2018\u641c\u7d22\u65b0\u4e8b\u7269\u2026\u2026\u63a2\u7d22\uff0c\u5b66\u4e60\u7684\u5185\u5728\u503e\u5411\u2019\u7684[43]\u3002\u8fd9\u91cc\u7684\u5185\u5728\uff0c\u548c\u5916\u5728\u4ef7\u503c\u76f8\u5bf9\uff0c\u6307\u7684\u4e0d\u662f\u6ee1\u8db3\u9700\u6c42\u6216\u8005\u89c4\u907f\u4e0d\u60f3\u89c1\u5230\u7684\u7ed3\u679c\u3002\u5185\u5728\u52a8\u673a\u5f15\u51fa\u4e86\u4e00\u822c\u7684\u63a2\u7d22\u884c\u4e3a\uff0c\u597d\u5947\u5fc3\u548c\u5b09\u620f\u3002\u597d\u5947\u5fc3\u6307\u4ee5\u6536\u96c6\u4fe1\u606f\u4e3a\u76ee\u6807\u7684\u63a2\u7d22\u884c\u4e3a[45-47]\u3002\u597d\u5947\u5fc3\u548c\u5b09\u620f\u5bc6\u5207\u76f8\u5173\uff0c\u540e\u8005\u662f\u4ee5\u63d0\u9ad8\u6280\u80fd\u4e3a\u5bfc\u5411\u7684\u65b0\u5947\u4e92\u52a8\u3002\u5728\u591a\u79cd\u667a\u6167\u52a8\u7269\u4e2d\uff0c\u90fd\u53ef\u4ee5\u89c2\u5bdf\u5230\u5b09\u620f\u884c\u4e3a\uff0c\u4f8b\u5982\u72d7\u548c\u4e4c\u9e26[48]\u3002<\/span><\/p>\n


<\/p>\n

\u5fc3\u667a\u6a21\u62df\u3002\u5fc3\u667a\u6a21\u62df\uff0c\u6307\u7684\u662f\u667a\u80fd\u4f53\u80fd\u591f\u5728\u591a\u79cd\u8868\u5f81\u5c42\u9762\uff0c\u9884\u6d4b\u73af\u5883\u52a8\u6001\u7684\u80fd\u529b\u3002\u4f8b\u5982\uff0c\u4eba\u4eec\u8ba4\u4e3a\uff0c\u52a8\u4f5c\u60f3\u8c61\u4f7f\u5f97\u8fd0\u52a8\u5458\u80fd\u591f\u901a\u8fc7\u5728\u5185\u5fc3\u6a21\u62df\u8fd9\u4e2a\u8fd0\u52a8\u52a8\u4f5c\uff0c\u6765\u6539\u8fdb\u5bf9\u9ad8\u96be\u5ea6\u8eab\u4f53\u52a8\u4f5c<\/span>\uff08\u4f8b\u5982\u5012\u7ffb\u8f6c\uff09<\/span>\u7684\u6267\u884c[49]\uff0c\u5de5\u7a0b\u5e08\u5728\u5f00\u53d1\u673a\u68b0\u7cfb\u7edf\u65f6\uff0c\u4e5f\u4f9d\u8d56\u5fc3\u667a\u6a21\u62df[13]\u3002\u591a\u9879\u7814\u7a76\u6307\u51fa\uff0c\u5fc3\u667a\u6a21\u62df\u5728\u66f4\u9ad8\u7ea7\u7684\u63a8\u7406\u8fc7\u7a0b\u4e2d\uff0c\u4e5f\u53d1\u6325\u4f5c\u7528\uff0c\u4f8b\u5982\uff0c[50-52]\u62a5\u544a\u4e86\u5fc3\u667a\u6a21\u62df\u662f\u5982\u4f55\u5728\u56e0\u679c\u6982\u5ff5\u5c42\u9762\uff0c\u6539\u5584\u5bf9\u672a\u6765\u884c\u4e3a\u7684\u89c4\u5212\u7684\u3002<\/span><\/p>\n


<\/p>\n

\u4f5c\u4e3a\u529f\u80fd\u524d\u63d0\u7684\u524d\u5411\u548c\u53cd\u5411\u6a21\u578b<\/span><\/strong><\/span><\/p>\n


<\/span><\/p>\n

\u611f\u89c9\u52a8\u4f5c\u62bd\u8c61\uff0c\u5185\u5728\u52a8\u673a\u548c\u5fc3\u667a\u6a21\u62df\uff0c\u90fd\u662f\u9700\u8981\u4e00\u4e2a\u5408\u9002\u795e\u7ecf\u529f\u80fd\u57fa\u7840\u7684\u590d\u6742\u673a\u5236\u3002\u6211\u4eec\u63d0\u51fa\uff0c\u8fd9\u4e00\u5173\u952e\u57fa\u7840\u53ef\u4ee5\u7531\u524d\u5411\u548c\u53cd\u5411\u6a21\u578b\u6784\u6210[24]\u3002\u524d\u5411\u6a21\u578b\u9884\u6d4b\u4e00\u4e2a\u52a8\u4f5c\u5982\u4f55\u5f71\u54cd\u4e16\u754c\uff0c\u800c\u53cd\u5411\u6a21\u578b\u51b3\u5b9a\u8981\u5b9e\u884c\u4e00\u4e2a\u76ee\u6807\u6240\u9700\u8981\u6267\u884c\u7684\u52a8\u4f5c\u3002<\/span><\/p>\n


<\/p>\n

\u5fc3\u667a\u6a21\u62df\u7684\u524d\u5411\u548c\u53cd\u5411\u6a21\u578b\u3002\u4e3a\u4e86\u8fdb\u884c\u5fc3\u667a\u6a21\u62df\uff0c\u4e00\u4e2a\u667a\u80fd\u4f53\u9700\u8981\u4e00\u4e2a\u524d\u5411\u6a21\u578b\uff0c\u6765\u6a21\u62df\u73af\u5883\u662f\u5982\u4f55\u6f14\u5316\u7684<\/span>\uff08\u53ef\u80fd\u53d6\u51b3\u4e8e\u667a\u80fd\u4f53\u7684\u52a8\u4f5c\uff09<\/span>\u3002\u7136\u800c\uff0c\u5bf9\u6240\u6709\u53ef\u80fd\u52a8\u4f5c\u7684\u53ef\u80fd\u7ed3\u679c\u8fdb\u884c\u9884\u6d4b\uff0c\u5f88\u5feb\u53d8\u6210\u8ba1\u7b97\u4e0a\u4e0d\u53ef\u884c\u7684\u3002\u56e0\u6b64\uff0c\u52a8\u4f5c\u9009\u62e9\u8981\u6c42\u5efa\u7acb\uff0c\u4ece\u5f53\u524d\u72b6\u6001\u548c\u60f3\u8981\u7684\u5185\u5728\u5fc3\u667a\u72b6\u6001\uff0c\u5230\u53ef\u80fd\u52a8\u4f5c\u7684\uff0c\u66f4\u76f4\u63a5\u7684\u8fde\u63a5\u3002\u8fd9\u53ef\u901a\u8fc7\u884c\u4e3a\u7b56\u7565\u7684\u53cd\u5411\u6a21\u578b\u5b9e\u73b0\uff0c\u8be5\u6a21\u578b\u751f\u6210\u5b8c\u6210\u5f53\u524d\u76ee\u6807\u7684\u53ef\u80fd\u52a8\u4f5c\u3002<\/span><\/p>\n


<\/p>\n

\u5185\u5728\u52a8\u673a\u7684\u524d\u5411\u548c\u53cd\u5411\u6a21\u578b\u3002\u5185\u5728\u52a8\u673a\u4e5f\u9700\u8981\u6b63\u5411\u548c\u53cd\u5411\u6a21\u578b\u3002\u4f8b\u5982\uff0c\u5bf9\u5185\u5728\u7684\u524d\u5411\u5b66\u4e60\u8fc7\u7a0b\u6a21\u578b\uff0c\u88ab\u8bc1\u660e\u662f\u5c06\u597d\u5947\u5fc3\u5efa\u6a21\u4e3a\u4e00\u79cd\u5185\u5728\u5956\u52b1\u7684\u6709\u6548\u65b9\u5f0f[47,53,54]\u3002Friston\u7b49\u4eba\u6307\u51fa\uff0c\u5f53\u4f7f\u7528\u524d\u5411\u6a21\u578b\u8fdb\u884c\u4e3b\u52a8\u63a8\u65ad\u65f6\uff0c\u4f1a\u6d8c\u73b0\u51fa\u5185\u5728\u9a71\u52a8\u7684\u884c\u4e3a[44,45]\uff0c\u8fd9\u610f\u5473\u7740\uff0c\u63a8\u65ad\u52a8\u4f5c\u7684\u8fc7\u7a0b\u6700\u5c0f\u5316\u4e86\u9884\u671f\u81ea\u7531\u80fd\u3002\u4f5c\u8005\u8ba4\u4e3a\uff0c\u4e3b\u52a8\u63a8\u65ad\u4f1a\u5bfc\u81f4\u8ba4\u77e5\u63a2\u7d22\uff0c\u5728\u8fd9\u4e2a\u8fc7\u7a0b\u4e2d\uff0c\u667a\u80fd\u4f53\u91c7\u53d6\u4e0d\u540c\u7684\u884c\u4e3a\uff0c\u4ee5\u51cf\u5c11\u5176\u524d\u5411\u9884\u6d4b\u7684\u4e0d\u786e\u5b9a\u6027[45,46]\u3002\u7136\u800c\uff0c\u4e3b\u52a8\u63a8\u65ad<\/span>\uff08active inference\uff09<\/span>\u548c\u81ea\u7531\u80fd\u539f\u7406<\/span>\uff08free energy principle\uff09<\/span>\uff0c\u90fd\u88ab\u6279\u5224\u4e3a\u8fc7\u4e8e\u5bbd\u6cdb\uff0c\u800c\u4e14\u53ef\u8bc1\u4f2a\u6027\u5b58\u7591[57]\u3002<\/span><\/p>\n


<\/p>\n

\u62bd\u8c61\u548c\u5960\u57fa\u7684\u524d\u5411\u548c\u53cd\u5411\u6a21\u578b\u3002\u591a\u79cd\u8ba4\u77e5\u7406\u8bba\uff0c\u5305\u62ec\u9884\u6d4b\u6027\u7f16\u7801<\/span>\uff08predictive coding\uff09<\/span>[58]\uff0c\u81ea\u7531\u80fd\u539f\u7406[59]\u548c\u8d1d\u53f6\u65af\u5927\u8111\u5047\u8bf4<\/span>\uff08Bayesian brain hypothesis\uff09<\/span>[60]\uff0c\u90fd\u5c06\u5927\u8111\u770b\u6210\u4e00\u79cd\u751f\u6210\u673a\u5236\uff0c\u4f1a\u6301\u7eed\u5730\u8bd5\u56fe\u5339\u914d\u8f93\u5165\u7684\u611f\u5b98\u4fe1\u53f7\u548c\u5bf9\u5b83\u4eec\u7684\u6982\u7387\u9884\u6d4b[61]\u3002\u5728\u8fd9\u4e00\u6846\u67b6\u4e0b\uff0c\u5728\u591a\u4e2a\u5904\u7406\u5c42\u7ea7\u4e0a\uff0c\u9884\u6d4b\u4ea4\u4e92\u5f0f\u5730\u53cc\u5411\u8fdb\u884c\uff1a\u81ea\u4e0a\u800c\u4e0b\u7684\u4fe1\u606f\uff0c\u4e3a\u9884\u6d4b\u611f\u5b98\u4fe1\u53f7\u63d0\u4f9b\u4e86\u989d\u5916\u7684\u4fe1\u606f\uff1b\u800c\u81ea\u4e0b\u800c\u4e0a\u7684\u9519\u8bef\u4fe1\u606f\uff0c\u88ab\u7528\u6765\u4fee\u6b63\u9ad8\u5c42\u7ea7\u7684\u9884\u6d4b[62]\u3002Butz[29]\u63d0\u51fa\uff0c\u8fd9\u6837\u7684\u5c42\u7ea7\u5316\u5904\u7406\u6d8c\u73b0\u51fa\u4e86\u62bd\u8c61\u7684\u3001\u7ec4\u5408\u5f0f\u7684\u3001\u66f4\u9ad8\u5c42\u7ea7\u7684\u7f16\u7801\uff0c\u4f8b\u5982\u201c\u5bb9\u5668\u201d\u7684\u6982\u5ff5\uff0c\u8fd9\u4f7f\u5f97\u5bf9\u4f4e\u5c42\u7279\u5f81\u7684\u9884\u6d4b<\/span>\uff08\u4f8b\u5982\u8be5\u5bb9\u5668\u6240\u5305\u542b\u7684\u7269\u4f53\u7684\u5c01\u95ed\u548c\u8fd0\u8f93\uff09<\/span>\u6210\u4e3a\u53ef\u80fd\u3002\u65f6\u95f4\u5206\u9694\u7406\u8bba<\/span>\uff08Event segmentation theory\uff09<\/span>[63]\uff0c\u4f7f\u5f97\u524d\u5411\u9884\u6d4b\u5728\u5b66\u4e60\u611f\u89c9\u8fd0\u52a8\u62bd\u8c61\u8fc7\u7a0b\u4e2d\u7684\u4f5c\u7528\u66f4\u4e3a\u660e\u663e\u3002\u6839\u636e\u8fd9\u4e00\u7406\u8bba\uff0c\u77ac\u65f6\u6027\u7684\u9884\u6d4b\u8bef\u5dee\uff0c\u88ab\u7528\u6765\u5c06\u8fde\u7eed\u7684\u4f53\u9a8c\u6d41\u5206\u5272\u4e3a\u6982\u5ff5\u4e8b\u4ef6\u7f16\u7801\u3002<\/span><\/p>\n


<\/p>\n


<\/p>\n

\n
\n
\n
\n
\n

<\/section>\n

<\/section>\n
\n
<\/svg><\/section>\n<\/section>\n<\/section>\n
\n

5. \u8ba1\u7b97\u5b9e\u73b0<\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n


<\/span><\/p>\n

\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\u7684\u4e00\u7cfb\u5217\u673a\u5236\u5728\u8ba1\u7b97\u673a\u4e2d\u7684\u5b9e\u73b0<\/span>\uff08\u5982\u56fe4\uff09<\/span>\u5e76\u4e0d\u50cf\u5728\u751f\u7269\u4f53\u4e2d\u6574\u5408\u5f97\u90a3\u4e48\u597d\u3002\u4e0b\u6587\u5c06\u6982\u8ff0\u5f53\u524d\u6700\u5148\u8fdb\u7684\uff0c\u4e0e\u795e\u7ecf\u8ba4\u77e5\u57fa\u7840\u76f8\u5173\u7684\u8ba1\u7b97\u5b9e\u73b0\u65b9\u5f0f\u3002<\/span><\/p>\n

       <\/span>        <\/span><\/p>\n

<\/span><\/section>\n
\u56fe4. \u901a\u7528\u7684\u5c42\u7ea7\u5316\u95ee\u9898\u89e3\u51b3\u67b6\u6784\u3002\u6bcf\u4e00\u4e2a\u5c42\u7ea7\u7684\u62bd\u8c61\uff0c\u90fd\u5206\u522b\u6709\u72ec\u7acb\u7684\u7b56\u7565\u03c0i<\/sup>\u3002\u5927\u90e8\u5206\u73b0\u6709\u7684\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\u5173\u6ce8\u4e24\u4e2a\u5c42\u7ea7\uff0c\u9ad8\u5c42\u7ea7\u7684\u52a8\u4f5c\u662f\u4e00\u4e2a\u52a8\u4f5c<\/span>a1<\/sup><\/span>\u57fa<\/span>\u5143\u6216\u5b50\u76ee\u6807\uff0c\u800c\u4f4e\u5c42\u7ea7\u7684\u52a8\u4f5ca0<\/sup>\u662f\u76f4\u63a5\u5f71\u54cd\u73af\u5883\u7684\u8fd0\u52a8\u6307\u4ee4\u3002<\/span><\/section>\n


<\/p>\n

\u8fc1\u79fb\u5b66\u4e60\u548c\u5c11\u6837\u672c\u89c4\u5212\u80fd\u529b<\/span><\/strong><\/span><\/p>\n


<\/span><\/p>\n

\u6211\u4eec\u5bf9\u795e\u7ecf\u8ba4\u77e5\u57fa\u7840\u7684\u8c03\u7814\u6307\u51fa\uff0c\u8fc1\u79fb\u5b66\u4e60\u548c\u89c4\u5212\uff0c\u662f\u5c0f\u6837\u672c\u95ee\u9898\u89e3\u51b3\u7684\u4e24\u9879\u57fa\u7840\u8ba4\u77e5\u80fd\u529b\u3002\u4f46\u6211\u4eec\u5982\u4f55\u5728\u8ba1\u7b97\u4e0a\u5bf9\u5b83\u4eec\u8fdb\u884c\u5efa\u6a21\u5462\uff1f<\/span><\/p>\n


<\/p>\n

\u8fc1\u79fb\u5b66\u4e60\u3002\u8fc1\u79fb\u5b66\u4e60\u5bf9\u4e8e\u5c42\u7ea7\u5316\u67b6\u6784\u6765\u8bf4\u662f\u5f88\u81ea\u7136\u7684\uff0c\u56e0\u4e3a\u8be5\u67b6\u6784\u53ef\u4ee5\u590d\u7528\u9ad8\u5c42\u7ea7\u7684\u901a\u7528\u6280\u80fd\uff0c\u5c06\u5b83\u4eec\u7528\u5728\u4e0d\u540c\u7684\u4f4e\u5c42\u7ea7\u4efb\u52a1\u4e0a\uff0c\u53cd\u4e4b\u4ea6\u7136\u3002\u56e0\u6b64\uff0c\u5728\u5f53\u524d\u8fc1\u79fb\u5b66\u4e60\u4e2d\uff0c\u6709\u5f88\u5927\u6bd4\u4f8b\u7684\u65b9\u6cd5\u662f\u5efa\u7acb\u5728\u590d\u7528\u4f4e\u5c42\u7ea7\u7b56\u7565\u4e0a\u7684[64-72]\u3002\u7814\u7a76\u8005\u4e0d\u4ec5\u8003\u8651\u4e86\u5728\u4e0d\u540c\u4efb\u52a1\u95f4\u7684\uff0c\u4e5f\u8003\u8651\u4e86\u5728\u4e0d\u540c\u673a\u5668\u4eba\uff0c\u5373\u4e0d\u540c\u5f62\u6001\u7684\u667a\u80fd\u4f53\u4e4b\u95f4\u7684\u8fc1\u79fb\u5b66\u4e60[65,73,74]\u3002<\/span><\/p>\n


<\/p>\n

\u89c4\u5212\u3002\u89c4\u5212\u53ef\u5206\u4e3a\u51b3\u7b56\u65f6\u89c4\u5212\u548c\u540e\u53f0\u89c4\u5212<\/span>\uff08background planning<\/span>\uff09[8,75]\u3002\u51b3\u7b56\u65f6\u89c4\u5212\u6307\u641c\u7d22\u4e00\u7cfb\u5217\u52a8\u4f5c\uff0c\u6765\u51b3\u5b9a\u4e3a\u8fbe\u6210\u7279\u5b9a\u76ee\u6807\uff0c\u63a5\u4e0b\u6765\u8981\u91c7\u53d6\u4ec0\u4e48\u52a8\u4f5c\u3002\u641c\u7d22\u662f\u901a\u8fc7\u7528\u7279\u5b9a\u9886\u57df\u7684\u52a8\u529b\u5b66\u7684\u5185\u90e8\u9884\u6d4b\u6a21\u578b\u6a21\u62df\u52a8\u4f5c\u6765\u5b9e\u73b0\u7684\u3002\u51b3\u7b56\u65f6\u89c4\u5212\uff0c\u4f7f\u96f6\u6837\u672c\u95ee\u9898\u89e3\u51b3\u6210\u4e3a\u53ef\u80fd\uff0c\u56e0\u4e3a\u667a\u80fd\u4f53\u53ea\u5728\u5185\u90e8\u9884\u6d4b\u6a21\u578b\u6307\u51fa\u52a8\u4f5c\u80fd\u8fbe\u6210\u9884\u671f\u76ee\u6807\u65f6\uff0c\u624d\u4f1a\u6267\u884c\u89c4\u5212\u7684\u52a8\u4f5c\u3002\u540e\u53f0\u89c4\u5212\uff0c\u662fSutton[76]\u63d0\u51fa\u7684\u4e00\u79cd\u57fa\u4e8e\u6a21\u578b\u7684\u5f3a\u5316\u5b66\u4e60\u65b9\u6cd5\u3002\u5b83\u4f7f\u7528\u524d\u5411\u6a21\u578b\u6a21\u62df\u52a8\u4f5c\uff0c\u6765\u8bad\u7ec3\u5f97\u5230\u7279\u5b9a\u7b56\u7565\u3002\u8fd9\u53ef\u4ee5\u63d0\u9ad8\u91c7\u6837\u6548\u7387\uff0c\u4f46\u8fd8\u4e0d\u8db3\u4ee5\u4f7f\u5c0f\u6837\u672c\u7684\u95ee\u9898\u89e3\u51b3\u6210\u4e3a\u53ef\u80fd\u3002<\/span><\/p>\n


<\/p>\n

\u5728\u4f20\u7edf\u7684\u4eba\u5de5\u667a\u80fd\u4e2d\uff0c\u5c42\u7ea7\u5316\u7684\u51b3\u7b56\u65f6\u89c4\u5212\u662f\u4e00\u4e2a\u5e7f\u4e3a\u4eba\u77e5\u7684\u8303\u5f0f[77]\uff0c\u4f46\u5c06\u5b83\u548c\u5c42\u7ea7\u5316\u7684\u5f3a\u5316\u5b66\u4e60\u7ed3\u5408\u7684\u65b9\u6cd5\u5374\u5f88\u7f55\u89c1\u3002\u6709\u4e9b\u65b9\u6cd5\u4f1a\u901a\u8fc7\u57fa\u4e8e\u52a8\u4f5c\u89c4\u5212\u7684\u9ad8\u5c42\u7ea7\u7684\u51b3\u7b56\uff0c\u548c\u57fa\u4e8e\u5f3a\u5316\u5b66\u4e60\u7684\u4f4e\u5c42\u7ea7\u7684\u52a8\u4f5c\u63a7\u5236\uff0c\u5c06\u52a8\u4f5c\u89c4\u5212\u548c\u5f3a\u5316\u5b66\u4e60\u7ed3\u5408[10,70,78,79]\u3002\u8fd9\u4e9b\u65b9\u6cd5\u8bc1\u660e\uff0c\u51b3\u7b56\u65f6\u89c4\u5212\u5bf9\u4e8e\u79bb\u6563\u72b6\u6001-\u52a8\u4f5c\u7a7a\u95f4\u4e2d\u7684\u9ad8\u5c42\u7ea7\u63a8\u65ad\u683c\u5916\u6709\u7528\u3002<\/span><\/p>\n


<\/p>\n

\u8fc1\u79fb\u5b66\u4e60\u548c\u89c4\u5212\u80cc\u540e\u7684\u673a\u5236<\/span><\/strong><\/span><\/p>\n


<\/span><\/p>\n

\u6211\u4eec\u5bf9\u8fc1\u79fb\u5b66\u4e60\u548c\u89c4\u5212\u80cc\u540e\u8ba4\u77e5\u539f\u7406\u7684\u603b\u7ed3\uff0c\u63ed\u793a\u4e86\u751f\u7269\u667a\u80fd\u4f53\u7684\u5b66\u4e60\u548c\u95ee\u9898\u89e3\u51b3\u80fd\u529b\u7684\u4e09\u4e2a\u91cd\u8981\u7684\u8ba4\u77e5\u673a\u5236\uff1a\u611f\u89c9\u8fd0\u52a8\u7684\u7ec4\u5408\u5f0f\u62bd\u8c61<\/span>\uff08compositional sensorimotor abstraction\uff09<\/span>\uff0c\u5185\u5728\u52a8\u673a\u548c\u5fc3\u667a\u6a21\u62df\u3002<\/span><\/p>\n


<\/p>\n

\u611f\u89c9\u52a8\u4f5c\u7684\u7ec4\u5408\u62bd\u8c61\u53ca\u5960\u57fa\u3002\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\uff0c\u5177\u6709\u65f6\u5e8f\u548c\u8868\u5f81\u4e24\u4e2a\u7ef4\u5ea6\uff0c\u8fd8\u80fd\u533a\u5206\u52a8\u4f5c\u548c\u72b6\u6001\u7684\u62bd\u8c61\u3002\u5bf9\u52a8\u4f5c\u7684\u65f6\u5e8f\u62bd\u8c61\uff0c\u5c06\u590d\u6742\u7684\u95ee\u9898\u5206\u89e3\u4e3a\u5c42\u7ea7\u5316\u7684\u7b80\u5355\u95ee\u9898\u3002\u5f71\u54cd\u529b\u6700\u5927\u7684\u8868\u5f81\u52a8\u4f5c\u62bd\u8c61\u65b9\u6cd5\uff0c\u662f\u57fa\u4e8e\u884c\u4e3a\u57fa\u5143[65,68,70,80-84]\u6784\u5efa\u7684<\/span>\uff08\u56fe5a\uff09<\/span>\uff0c\u5176\u4e2d\u5305\u62ec\u4e86\u9009\u62e9\u3001\u6b21\u7ea7\u7b56\u7565\u548c\u539f\u5b50\u5316\u7684\u9ad8\u5c42\u7ea7\u6280\u80fd\u3002\u8fd9\u4e9b\u884c\u4e3a\u57fa\u5143\u4f4d\u4e8e\u4e00\u4e2a\u62bd\u8c61\u7684\u8868\u5f81\u7a7a\u95f4\u4e2d\uff0c\u4f8b\u5982\uff0c\u5728\u4e00\u4e2a\u5145\u6ee1\u52a8\u4f5c\u6807\u7b7e\u548c\u7d22\u5f15\u7684\u79bb\u6563\u6709\u9650\u7a7a\u95f4\u4e2d\u3002\u800c\u8f83\u4e3a\u65b0\u8fdb\u7684\u9ad8\u5c42\u7ea7\u52a8\u4f5c\u8868\u5f81\uff0c\u5219\u662f\u7531\u4f4e\u5c42\u7ea7\u7684\u72b6\u6001\u7a7a\u95f4\u4e2d\u7684\u5b50\u76ee\u6807\u6240\u6784\u6210[10,74,85-90]<\/span>\uff08\u56fe5b\uff09<\/span>\u3002\u5728\u8fd9\u91cc\uff0c\u667a\u80fd\u4f53\u901a\u8fc7\u5b8c\u6210\u4e00\u7cfb\u5217\u5b50\u76ee\u6807\uff0c\u6765\u5b9e\u73b0\u6700\u540e\u7684\u76ee\u6807\u3002<\/span><\/p>\n

       <\/span>        <\/span><\/p>\n

<\/span><\/section>\n
\u56fe5. \u4e0d\u540c\u7c7b\u578b\u7684\u52a8\u4f5c\u62bd\u8c61\u3002a. \u57fa\u4e8e\u52a8\u4f5c\u57fa\u5143\uff0c\u667a\u80fd\u4f53\u901a\u8fc7\u9009\u62e9\u4e00\u7cfb\u5217\u9ad8\u5c42\u7ea7\u7684\u52a8\u4f5c\uff0c\u800c\u4e0d\u5177\u4f53\u6307\u5b9a\u6bcf\u4e00\u4e2a\u52a8\u4f5c\u4f1a\u5bfc\u81f4\u7684\u4e2d\u95f4\u72b6\u6001\uff0c\u6765\u51b3\u5b9a\u8fbe\u5230\u6700\u7ec8\u72b6\u6001\u7684\u8def\u5f84\u3002b. \u5b58\u5728\u5b50\u76ee\u6807\u65f6\uff0c\u667a\u80fd\u4f53\u6307\u5b9a\u8981\u8fbe\u6210\u6700\u7ec8\u72b6\u6001\u7684\u5b50\u76ee\u6807\uff0c\u800c\u4e0d\u6307\u5b9a\u8fbe\u6210\u5b50\u76ee\u6807\u7684\u52a8\u4f5c\u3002<\/span><\/section>\n


<\/p>\n

\u5f53\u524d\u5f88\u591a\u57fa\u4e8e\u89c6\u89c9\u8f93\u5165\u7684\u5f3a\u5316\u5b66\u4e60\u65b9\u5f0f\uff0c\u90fd\u4f7f\u7528\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u6765\u63d0\u4f9b\u4e00\u4e2a\u76f8\u5bf9\u901a\u7528\u7684\u62bd\u8c61\uff0c\u5176\u4e2d\u6240\u6709\u5c42\u7ea7\u90fd\u4f7f\u7528\u540c\u4e00\u4e2a\u62bd\u8c61\u7684\u89c6\u89c9\u7279\u5f81\u56fe[11,12,67,91-94]\u3002\u8be5\u7b56\u7565\u7684\u95ee\u9898\u662f\u5b83\u5ffd\u7565\u4e86\u201c\u4e0d\u540c\u5c42\u7ea7\u7684\u63a8\u65ad\u9700\u8981\u4e0d\u540c\u5c42\u7ea7\u7684\u62bd\u8c61\u201d\u8fd9\u4e00\u4e8b\u5b9e\uff1a\u4f8b\u5982\uff0c\u5bf9\u4e8e\u8fd0\u8f93\u4efb\u52a1\uff0c\u7269\u4f53\u7684\u5177\u4f53\u5f62\u72b6\u548c\u91cd\u91cf\uff0c\u53ea\u4e0e\u4f4e\u5c42\u7ea7\u7684\u8fd0\u52a8\u63a7\u5236\u6709\u5173\u3002\u66f4\u9ad8\u5c42\u7ea7\u7684\u89c4\u5212\u5c42\uff0c\u53ea\u9700\u8981\u62bd\u8c61\u4e8b\u5b9e\uff0c\u8bf8\u5982\u67d0\u4e2a\u7269\u4f53\u662f\u5426\u662f\u53ef\u6293\u7684\u6216\u53ef\u8fd0\u8f93\u7684\u3002\u56e0\u6b64\uff0c\u8868\u5f81\u72b6\u6001\u7684\u62bd\u8c61\uff0c\u5e94\u8be5\u548c\u5176\u8981\u4f7f\u7528\u7684\u5c42\u7ea7\u76f8\u4e00\u81f4\u3002\u6211\u4eec\u5c06\u8fd9\u6837\u7684\u62bd\u8c61\u79f0\u4e3a\u9002\u5408\u63a8\u65ad\u7684\u62bd\u8c61\u3002\u5f53\u524d\u7684\u5f3a\u5316\u5b66\u4e60\u65b9\u6cd5\uff0c\u5bf9\u89e3\u51b3\u8be5\u95ee\u9898\u5df2\u6709\u4e0d\u5c0f\u6837\u672c\uff0c\u4f46\u5927\u591a\u57fa\u4e8e\u624b\u52a8\u5b9a\u4e49\u7684\u62bd\u8c61\u51fd\u6570[10,79,82,95]\u3002\u4e5f\u6709\u4e00\u4e9b\u7279\u4f8b\uff0c\u5176\u72b6\u6001\u62bd\u8c61\u662f\u81ea\u52a8\u4ece\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\u4e2d\u5bfc\u51fa\u7684\uff0c\u4f8b\u5982\u901a\u8fc7\u805a\u7c7b[85,96]\uff0c\u7ecf\u7531\u7279\u5f81\u8868\u5f81[97]\u6216\u901a\u8fc7\u56e0\u5b50\u5316<\/span>\uff08factorization\uff09<\/span>[98]\u3002<\/span><\/p>\n


<\/p>\n

\u8ba4\u77e5\u79d1\u5b66\u8fd8\u6307\u51fa\uff0c\u7ec4\u5408\u6027\u662f\u62bd\u8c61\u8868\u5f81\u7684\u4e00\u4e2a\u91cd\u8981\u7279\u5f81\u3002\u4f8b\u5982\uff0c\u4e00\u4e2a\u7b26\u53f7\u5316\u7684\u7ec4\u5408\u5f0f\u7684\u52a8\u4f5c\u8868\u5f81\u6293<\/span>\uff08\u676f\u5b50\uff09<\/span>\uff0c\u5141\u8bb8\u5728\u201c\u6293\u53d6\u201d\u8fd9\u4e00\u52a8\u4f5c\u548c\u201c\u676f\u5b50\u201d\u8fd9\u4e00\u7269\u4f53\u95f4\u8fdb\u884c\u8c03\u63a7\u3002\u5bf9\u4e8e\u5206\u5e03\u5f0f\u6570\u503c\u5316\u7684\u8868\u5f81\uff0c\u4f8b\u5982\u5411\u91cf\uff0c\u7ec4\u5408\u6027\u4e5f\u662f\u9002\u7528\u7684\u3002\u4f8b\u5982\uff0c\u5982\u679cv1<\/sub><\/span>\u53ef\u4ee5\u7531\u5176\u5b83\u5411\u91cf\uff0c\u4f8b\u5982v2<\/sub><\/span>\u548cv3<\/sub><\/span>\uff0c\u7ec4\u5408\u800c\u6210\uff0c\u800c\u4e14\u5b58\u5728\u4e00\u4e2a\u4ee5\u89e3\u91ca\u7ec4\u5408\u89c4\u5219\u7684\u5411\u91cf\u64cd\u4f5c\uff0c\u4f8b\u5982v1<\/sub><\/span>=v2<\/sub><\/span>\u00b7v3<\/sub><\/span>\uff0c\u90a3\u4e48\u53ef\u4ee5\u8bf4\u8fd9\u4e2a\u5411\u91cfv1<\/sub><\/span>\u5177\u6709\u7ec4\u5408\u6027\u3002<\/span><\/p>\n


<\/p>\n

\u8ba4\u77e5\u79d1\u5b66\u4e2d\uff0c\u6709\u5145\u8db3\u7684\u8bc1\u636e\u8868\u660e\uff0c\u7ec4\u5408\u6027\u53ef\u4ee5\u6539\u5584\u8fc1\u79fb\u5b66\u4e60[98]\u3002\u8003\u8651\u5230\u8fc1\u79fb\u5b66\u4e60\u7684\u786e\u4f9d\u8d56\u95ee\u9898\u4e4b\u95f4\u7684\u76f8\u4f3c\u6027\uff0c\u8fd9\u4e9b\u8bc1\u636e\u4ece\u8ba1\u7b97\u89d2\u5ea6\u6765\u8bf4\u662f\u6709\u9053\u7406\u7684\u3002\u5176\u5b83\u7531\u8ba4\u77e5\u79d1\u5b66\u542f\u53d1\u7684\u8ba1\u7b97\u65b9\u6cd5\uff0c\u4f8b\u5982\u6982\u5ff5\u6574\u5408<\/span>\uff08concept binding\uff09<\/span>[41]\uff0c\u8fdb\u4e00\u6b65\u8bc1\u660e\u4e86\uff0c\u8f83\u5c11\u7684\u53ef\u80fd\u7c7b\u6bd4\u6620\u5c04\u63d0\u9ad8\u4e86\u53ef\u8fc1\u79fb\u6027\u3002\u6b64\u5916\uff0cJiang\u7b49\u4eba\u7684\u7814\u7a76\u5de5\u4f5c[67]\u63d0\u4f9b\u4e86\u5b9e\u8bc1\u6570\u636e\uff0c\u6307\u51fa\u5177\u6709\u7ec4\u5408\u5f0f\u7684\u8868\u5f81\u80fd\u63d0\u5347\u5b66\u4e60\u7684\u8fc1\u79fb\u80fd\u529b\u3002\u4ed6\u4eec\u4f7f\u7528\u81ea\u7136\u8bed\u8a00\uff0c\u8fd9\u4e00\u5185\u5728\u7ec4\u5408\u5f0f\u8868\u5f81\uff0c\u6765\u63cf\u8ff0\u5c42\u7ea7\u8bdd\u5f3a\u5316\u5b66\u4e60\u4e2d\u7684\u9ad8\u5c42\u7ea7\u52a8\u4f5c\uff0c\u5e76\u8868\u660e\uff0c\u81ea\u7136\u8bed\u8a00\u7684\u8868\u5f81\uff0c\u76f8\u6bd4\u975e\u7ec4\u5408\u5f0f\u7684\u8868\u5f81\uff0c\u63d0\u5347\u4e86\u8fc1\u79fb\u5b66\u4e60\u7684\u6548\u679c\u3002<\/span><\/p>\n


<\/p>\n

\u5185\u5728\u52a8\u673a\u3002\u4f7f\u5f3a\u5316\u5b66\u4e60\u5177\u6709\u7a33\u5b9a\u6027\u7684\u4e00\u4e2a\u6709\u6548\u65b9\u6cd5\uff0c\u662f\u91c7\u7528\u5185\u5728\u52a8\u673a\uff0c\u5373\uff0c\u5728\u7a00\u758f\u7684\u5916\u5728\u7684\u5956\u52b1\u4e4b\u5916\uff0c\u589e\u52a0\u4e0e\u9886\u57df\u65e0\u5173\u7684\u5185\u5728\u5956\u52b1\u3002\u6709\u5185\u5728\u52a8\u673a\u7684\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\u6700\u5e38\u89c1\u7684\u65b9\u6cd5\u662f\uff1a\u5728\u5b50\u4efb\u52a1\/\u76ee\u6807\u5b8c\u6210\u65f6\u63d0\u4f9b\u5185\u5728\u5956\u52b1[10,11,69,78,82,99-103]\u3002<\/span><\/p>\n


<\/p>\n

\u5176\u5b83\u63d0\u4f9b\u5185\u5728\u52a8\u673a\u7684\u65b9\u5f0f\uff0c\u5305\u62ec\u8bc6\u522b\u4e00\u7ec4\u5177\u6709\u591a\u6837\u6027[64,99,104]\u6216\u53ef\u9884\u6d4b\u6027[70]\u7684\u57fa\u672c\u884c\u4e3a\uff0c\u4ee5\u53ca\u8bc6\u522b\u9002\u5408\u4e8e\u91cd\u7ec4\u9ad8\u5c42\u4efb\u52a1\u7684\u57fa\u5143\u3002<\/span><\/p>\n


<\/p>\n

\u53e6\u4e00\u79cd\u5728\u975e\u5c42\u7ea7\u7684\u5f3a\u5316\u5b66\u4e60\u4e2d\u5e38\u89c1\u7684\u5185\u5728\u5956\u52b1\u6a21\u578b\uff0c\u662f\u57fa\u4e8e\u60ca\u8bb6\u5ea6<\/span>\uff08surprise\uff09<\/span>\u548c\u597d\u5947\u5fc3[47,54,105,106]\u3002\u5728\u8fd9\u4e9b\u65b9\u6cd5\u4e2d\uff0c\u60ca\u8bb6\u5ea6\u901a\u5e38\u88ab\u5efa\u6a21\u4e3a\u524d\u5411\u9884\u6d4b\u8bef\u5dee\u7684\u4e00\u4e2a\u51fd\u6570\uff0c\u800c\u597d\u5947\u5fc3\u5219\u662f\u901a\u8fc7\uff0c\u5728\u667a\u80fd\u4f53\u9047\u5230\u60ca\u8bb6\u5ea6\u65f6\u63d0\u4f9b\u5185\u5728\u7684\u5956\u52b1\uff0c\u6765\u5b9e\u73b0\u7684\u3002\u7136\u800c\uff0c\u5728\u5c42\u7ea7\u5316\u7684\u8bbe\u5b9a\u4e0b\uff0c\u53ea\u6709[89,99,107]\u91c7\u7528\u4e86\u60ca\u8bb6\u5ea6\uff0c\u8bc1\u660e\u4e86\u597d\u5947\u5fc3\u80fd\u663e\u8457\u63d0\u5347\u5b66\u4e60\u7684\u6027\u80fd\u3002\u5c42\u7ea7\u5316\u7684\u597d\u5947\u5fc3\u53ef\u80fd\u53ef\u4ee5\u5e26\u6765\u534f\u540c<\/span>\uff08synergy\uff09<\/span>\uff0c\u56e0\u4e3a\u4e00\u4e2a\u6709\u597d\u5947\u5fc3\u7684\u9ad8\u62bd\u8c61\u5c42\u7ea7\uff0c\u4f1a\u81ea\u52a8\u5730\u4e3a\u4f4e\u5c42\u7ea7\u751f\u6210\u5b66\u4e60\u4efb\u52a1\u3002\u4f8b\u5982\uff0c\u5b83\u4f1a\u4ea7\u751f\u96be\u5ea6\u9002\u4e2d\u7684\u5b50\u4efb\u52a1\uff0c\u4f9b\u4f4e\u5c42\u7ea7\u53bb\u5b8c\u6210\u3002\u56f0\u96be\u4e4b\u5904\u5728\u4e8e\u627e\u5230\u96be\u5ea6\u6070\u597d\u7684\u6761\u4ef6\uff0c\u5e76\u627e\u5230\u63a2\u7d22<\/span>\uff08explorative\uff09<\/span>\u548c\u5229\u7528<\/span>\uff08exploitative\uff09<\/span>\u8fd9\u4e24\u4e2a\u5b50\u4efb\u52a1\u4e4b\u95f4\u7684\u9002\u5f53\u5e73\u8861\u3002<\/span><\/p>\n


<\/p>\n

\u5fc3\u667a\u6a21\u62df\u3002\u5fc3\u667a\u6a21\u62df\u4f7f\u667a\u80fd\u4f53\u80fd\u591f\u9884\u6d4b\u81ea\u8eab\u4ee5\u53ca\u4ed6\u8005\u52a8\u4f5c\u7684\u5f71\u54cd\u3002\u56e0\u6b64\uff0c\u8fd9\u662f\u4f7f\u667a\u80fd\u4f53\u5177\u6709\u4e8b\u5148\u89c4\u5212\u80fd\u529b\u7684\u4e00\u9879\u5173\u952e\u673a\u5236\u3002\u5fc3\u667a\u6a21\u62df\u7684\u8ba1\u7b97\u5b9e\u73b0\u65b9\u5f0f\u5305\u62ec\u57fa\u4e8e\u6a21\u578b\u7684\u5f3a\u5316\u5b66\u4e60[76]\uff0c\u52a8\u4f5c\u89c4\u5212[77]\u4ee5\u53ca\u4e24\u8005\u7684\u7efc\u5408[10,108]\u3002\u5c3d\u7ba1\u6709\u8ba4\u77e5\u79d1\u5b66\u7684\u8bc1\u636e\u8868\u660e\u5fc3\u667a\u6a21\u62df\u53d1\u751f\u5728\u591a\u4e2a\u8868\u5f81\u5c42\u9762[2,36]\uff0c\u5c42\u7ea7\u5316\u7684\u5fc3\u667a\u6a21\u62df\u65b9\u6cd5\u5374\u662f\u5f88\u5c11\u7684\u3002\u53ea\u6709\u5c11\u6570\u6574\u5408\u4e86\u5f3a\u5316\u5b66\u4e60\u89c4\u5212\u7684\u5c42\u7ea7\u5316\u65b9\u6cd5\u662f\u5efa\u7acb\u5728\u5fc3\u667a\u6a21\u62df\u4e0a\u7684[10,70,78,79,103,109]\uff0c\u4e0d\u8fc7\uff0c\u8fd9\u4e9b\u6a21\u578b\u4e5f\u4ec5\u5728\u8f83\u9ad8\u89c4\u5212\u5c42\u5e94\u7528\u4e86\u5fc3\u667a\u6a21\u62df\u3002Wu\u7b49\u4eba\u7684\u5de5\u4f5c[110]\u662f\u4e00\u4e2a\u4f8b\u5916\uff0c\u5b83\u5728\u4f4e\u5c42\u7ea7\u4f7f\u7528\u4e86\u5fc3\u667a\u6a21\u62df\uff0c\u51b3\u5b9a\u54ea\u4e9b\u5b50\u7b56\u7565\u5c06\u88ab\u6267\u884c\u3002\u53e6\u4e00\u4e2a\u4f8b\u5916\u662f[111]\uff0c\u5b83\u5728\u591a\u4e2a\u4efb\u52a1\u5c42\u90fd\u4f7f\u7528\u4e86\u5fc3\u667a\u6a21\u62df\u6765\u8fdb\u884c\u89c4\u5212\u3002\u7136\u800c\uff0c\u6211\u4eec\u8fd8\u4e0d\u77e5\u9053\u6709\u4ec0\u4e48\u65b9\u5f0f\u662f\u8fdb\u884c\u4e86\u5c42\u7ea7\u5316\u7684\u540e\u53f0\u89c4\u5212\u7684\uff0c\u5373\uff0c\u901a\u8fc7\u5728\u591a\u5c42\u6267\u884c\u7684\u52a8\u4f5c\uff0c\u6765\u8bad\u7ec3\u4e00\u4e2a\u5c42\u7ea7\u5316\u7684\u7b56\u7565\u3002\u9ad8\u5c42\u7ea7\u7684\u540e\u53f0\u89c4\u5212\u7f3a\u5931\u7684\u4e00\u4e2a\u53ef\u80fd\u7684\u7406\u7531\u662f\uff0c\u5b83\u9700\u8981\u7a33\u5b9a\u7684\u4f4e\u5c42\u7ea7\u7684\u8868\u5f81\uff0c\u800c\u5bf9\u76f8\u5e94\u7684\u7a33\u5b9a\u6807\u51c6\u7684\u5efa\u6a21\u662f\u5f88\u96be\u7684\u3002<\/span><\/p>\n


<\/p>\n

\u611f\u89c9\u8fd0\u52a8\u62bd\u8c61\uff1a\u5185\u5728\u52a8\u673a\u548c\u5fc3\u667a\u6a21\u62df\u7684\u5148\u51b3\u6761\u4ef6<\/span><\/strong><\/span><\/p>\n


<\/span><\/p>\n

\u57fa\u4e8e\u7b56\u7565\u6216\u53cd\u5411\u6a21\u578b\u7684\u5f3a\u5316\u5b66\u4e60\uff0c\u4f1a\u9884\u6d4b\u4e00\u4e2a\u5728\u5f53\u524d\u72b6\u6001\u4e0b\uff0c\u80fd\u591f\u6700\u5927\u5316\u5956\u52b1\u6216\u5b8c\u6210\u7279\u5b9a\u76ee\u6807\u7684\u52a8\u4f5c\u3002\u4e0e\u4e4b\u76f8\u5bf9\u7684\uff0c\u524d\u5411\u6a21\u578b\u4f1a\u6839\u636e\u5f53\u524d\u72b6\u6001\uff0c\u548c\u8981\u91c7\u53d6\u7684\u4e00\u7cfb\u5217\u52a8\u4f5c\uff0c\u9884\u6d4b\u672a\u6765\u7684\u72b6\u6001\u3002\u6587\u732e\u8c03\u7814\u6307\u51fa\uff0c\u5927\u90e8\u5206\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\uff0c\u5728\u9ad8\u5c42\u7ea7\u548c\u4f4e\u5c42\u7ea7\uff0c\u90fd\u4f7f\u7528\u4e86\u53cd\u5411\u6a21\u578b\u3002\u800c\u6709\u7684\u65b9\u6cd5\u53ea\u5728\u4f4e\u5c42\u7ea7\u4f7f\u7528\u53cd\u5411\u6a21\u578b[10,70,78,79,103,109]\uff0c\u5728\u9ad8\u5c42\u7ea7\u51b3\u7b56\u65f6\u5219\u4f7f\u7528\u51b3\u7b56\u65f6\u89c4\u5212\u3002\u6587\u732e\u7efc\u8ff0\u8fd8\u6307\u51fa\uff0c\u6709\u6570\u9879\u673a\u5236\u5bf9\u4e8e\u8fc1\u79fb\u5b66\u4e60\uff0c\u89c4\u5212\u548c\u5c0f\u6837\u672c\u95ee\u9898\u89e3\u51b3\uff0c\u662f\u5fc5\u8981\u7684\uff0c\u6216\u81f3\u5c11\u662f\u80fd\u6709\u5229\u7684\uff0c\u800c\u8fd9\u4e9b\u673a\u5236\u9700\u8981\u524d\u5411\u6a21\u578b\u3002\u6709\u4e00\u4e9b\u975e\u5c42\u7ea7\u7684\u5f3a\u5316\u5b66\u4e60\uff0c\u4f7f\u7528\u4e86\u524d\u5411\u6a21\u578b\u8fdb\u884c\u611f\u89c9\u8fd0\u52a8\u62bd\u8c61[105,112]\u3002\u8fd9\u662f\u901a\u8fc7\u81ea\u76d1\u7763\u7684\u65b9\u5f0f\u8fbe\u6210\u7684\uff1a\u524d\u5411\u9884\u6d4b\u662f\u5728\u4e00\u4e2a\u6f5c\u5728\u7684\u62bd\u8c61\u7a7a\u95f4\u4e2d\u5b66\u4e60\u7684\u3002\u4f46\u76ee\u524d\uff0c\u8fd9\u4e00\u673a\u5236\uff0c\u8fd8\u6ca1\u6709\u7528\u5c42\u7ea7\u5316\u7684\u65b9\u5f0f\u91c7\u7528\u8fc7\u3002<\/span><\/p>\n


<\/p>\n

\u5bf9\u4e8e\u91c7\u7528\u597d\u5947\u5fc3\u4f5c\u4e3a\u5185\u5728\u52a8\u673a\u7684\u6a21\u578b\uff0c\u524d\u5411\u6a21\u578b\u4e5f\u662f\u6709\u7528\u7684\uff0c\u4f8b\u5982\uff0c\u5bf9\u667a\u80fd\u4f53\u5728\u60ca\u8bb6\u51fa\u73b0\u65f6\u7ed9\u4e0e\u5956\u52b1\uff0c\u5176\u4e2d\u60ca\u8bb6\u5ea6\u662f\u9884\u6d4b\u8bef\u5dee\u7684\u51fd\u6570[55]\u3002\u597d\u5947\u5fc3\u5df2\u88ab\u8bc1\u660e\uff0c\u80fd\u7f13\u89e3\u5f3a\u5316\u5b66\u4e60\u4e2d\u7684\u7a00\u758f\u5956\u52b1\u95ee\u9898[89,105]\u3002\u4f46\u9664\u4e86[89,107]\u7b49\u5c11\u8bb8\u4f8b\u5916\uff0c\u5f88\u5c11\u6709\u4f7f\u7528\u5c42\u7ea7\u5316\u524d\u5411\u6a21\u578b\uff0c\u6765\u4ea7\u751f\u5c42\u7ea7\u5316\u597d\u5947\u5fc3\u7684\u5c1d\u8bd5\u3002<\/span><\/p>\n


<\/p>\n


<\/p>\n

\n
\n
\n
\n
\n

<\/section>\n

<\/section>\n
\n
<\/svg><\/section>\n<\/section>\n<\/section>\n
\n

6. \u8ba8\u8bba\u548c\u5c55\u671b<\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n


<\/span><\/p>\n

\u5f53\u524d\u7684\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\u7b56\u7565\uff0c\u6709\u5982\u4e0b\u7f3a\u70b9\u3002\u8fd9\u4e9b\u7f3a\u70b9\u662f\u7531\u878d\u5408\u4e0d\u540c\u673a\u5236\u8fc7\u7a0b\u4e2d\u7684\u7279\u5b9a\u6311\u6218\u4ea7\u751f\u7684<\/span>\uff08\u56fe6\uff09<\/span>\u3002\u5728\u6b64\uff0c\u6211\u4eec\u901a\u8fc7\u5177\u4f53\u7684\u7b97\u6cd5\u5efa\u8bae\u6765\u89e3\u51b3\u8fd9\u4e9b\u7f3a\u70b9\u3002<\/span><\/p>\n


<\/p>\n

       <\/span>        <\/span><\/p>\n

\u56fe6. \u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\u7684\u7f3a\u9677\uff0c\u6311\u6218\u548c\u5efa\u8bae\u3002\u8ba1\u7b97\u5b9e\u73b0\u90e8\u5206\u6982\u8ff0\u4e86\u5f53\u4ee3\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\u65b9\u6cd5\u7684\u51e0\u4e2a\u7f3a\u70b9\u3002\u6211\u4eec\u8ba4\u4e3a\uff0c\u8fd9\u4e9b\u7f3a\u70b9\u81f3\u5c11\u6765\u81ea\u56db\u4e2a\u6838\u5fc3\u6311\u6218\uff0c\u5e76\u5c31\u6709\u52a9\u4e8e\u89e3\u51b3\u8fd9\u4e9b\u6311\u6218\u7684\u5173\u952e\u65b9\u6cd5\u63d0\u51fa\u4e86\u4e09\u70b9\u5efa\u8bae\u3002<\/span><\/section>\n


<\/p>\n

\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\u67b6\u6784\u7684\u7f3a\u70b9\u548c\u6311\u6218<\/span><\/strong><\/span><\/p>\n


<\/span><\/p>\n

\u4ee5\u4e0b\u5404\u8282\u6307\u51fa\u4e86\u5f53\u524d\u65b9\u6cd5\u7684\u4e3b\u8981\u6311\u6218\u3002<\/span><\/p>\n


<\/span><\/p>\n

\u5c42\u7ea7\u5316\u7684\u5fc3\u667a\u6a21\u62df\uff0c\u9700\u8981\u7a33\u5b9a\u7684\u5c42\u7ea7\u8868\u5f81\u548c\u7cbe\u786e\u7684\u9884\u6d4b\u3002\u5927\u90e8\u5206\u5f53\u524d\u7684\u5c0f\u6837\u672c\u89e3\u51b3\u95ee\u9898\uff0c\u90fd\u4f9d\u8d56\u8fc1\u79fb\u5b66\u4e60\uff0c\u4f46\u5b83\u4eec\u65e2\u6ca1\u6709\u5229\u7528\u51b3\u7b56\u65f6\u7684\u89c4\u5212\uff0c\u4e5f\u6ca1\u6709\u8003\u8651\u540e\u53f0\u89c4\u5212\u6765\u53d1\u6325\u5c42\u7ea7\u5316\u5fc3\u667a\u6a21\u62df\u7684\u597d\u5904\u3002\u5c3d\u7ba1\u5f88\u591a\u975e\u5c42\u7ea7\u5316\u7684\u65b9\u6cd5\u6307\u51fa\uff0c\u5fc3\u667a\u6a21\u62df\u53ef\u4ee5\u63d0\u5347\u91c7\u6837\u6548\u7387[113,114]\uff0c\u4f46\u4ec5\u6709\u4e00\u9879\u7814\u7a76\u5728\u591a\u5c42\u7ea7\u4e0a\u4f7f\u7528\u4e86\u5c42\u7ea7\u5316\u7684\u5fc3\u667a\u6a21\u62df[111]\u3002\u8fd9\u4e00\u7f3a\u5931\u7684\u4e3b\u8981\u539f\u56e0\uff0c\u662f\u9ad8\u5c42\u7ea7\u7684\u524d\u5411\u6a21\u578b\u5728\u5f00\u59cb\u8bad\u7ec3\u65f6\uff0c\u4f4e\u5c42\u7ea7\u7684\u7b56\u7565\u5e94\u8be5\u5df2\u7ecf\u6536\u655b\u3002\u8be5\u95ee\u9898\u5728\u65e0\u6a21\u578b\u7684\u5f3a\u5316\u5b66\u4e60\u4e2d\u5df2\u88ab\u89e3\u51b3[64]\uff0c\u4f46\u5bf9\u4e8e\u9884\u6d4b\u6027\u7684\u524d\u5411\u6a21\u578b\uff0c\u8fd8\u6ca1\u6709\u76f8\u5173\u7814\u7a76\u3002<\/span><\/p>\n


<\/p>\n

\u5185\u5728\u52a8\u673a\u9700\u8981\u6a21\u578b\u5177\u6709\u7cbe\u786e\u7684\u4e0d\u786e\u5b9a\u6027\u5e94\u5bf9\u80fd\u529b\u3002\u572850\u7bc7\u8c03\u7814\u7684\u6587\u732e\u4e2d\uff0c\u53ea\u67099\u7bc7\u5e94\u7528\u4e86\u591a\u6837\u6027\u548c\u597d\u5947\u5fc3\uff0c\u8fd9\u4e2a\u6bd4\u4f8b\u662f\u504f\u4f4e\u7684\u3002\u795e\u7ecf\u8ba4\u77e5\u57fa\u7840\u4e00\u8282\u4ecb\u7ecd\u4e86\u5927\u91cf\u7684\u8ba4\u77e5\u8bc1\u636e\uff0c\u8868\u660e\u8fd9\u4e9b\u673a\u5236\u5bf9\u751f\u7269\u667a\u80fd\u4f53\u7684\u667a\u80fd\u884c\u4e3a\u81f3\u5173\u91cd\u8981[56]\u3002<\/span><\/p>\n


<\/p>\n

\u8fd9\u4e00\u7f3a\u9677\u7684\u4e00\u4e2a\u6f5c\u5728\u539f\u56e0\uff0c\u662f\u57fa\u4e8e\u9884\u6d4b\u7684\u5185\u5728\u52a8\u673a\u6a21\u578b\uff0c\u53d7\u5230\u4e86\u5f53\u524d\u524d\u5411\u8ba1\u7b97\u6a21\u578b\u4e2d\uff0c\u5bf9\u4e0d\u786e\u5b9a\u6027\u8fc7\u4e8e\u7b80\u5355\u7684\u5904\u7406\u7684\u5f71\u54cd\uff0c\u8fd9\u662f\u5e26\u566a\u58f0\u7684\u89c2\u5bdf\u6216\u6df7\u6c8c\u4e0d\u53ef\u9884\u6d4b\u7684\u52a8\u6001\u7cfb\u7edf\u8981\u9762\u5bf9\u7684\u95ee\u9898\u3002\u4f8b\u5982\uff0c\u901a\u8fc7\u6700\u5927\u5316\u524d\u5411\u6a21\u578b\u7684\u9884\u6d4b\u8bef\u5dee\uff0cRoder\u7b49\u4eba[89]\u5b9e\u73b0\u4e86\u4e00\u79cd\u597d\u5947\u5fc3\u9a71\u52a8\u7684\u5185\u5728\u52a8\u673a\u3002\u4f46\u8fd9\u4f1a\u5bfc\u81f4\u667a\u80fd\u4f53\u9047\u5230\u6240\u8c13\u7684\u201c\u4e0d\u6e05\u6670\u7684\u7535\u89c6\u95ee\u9898\u201d\uff0c\u5373\u667a\u80fd\u4f53\u603b\u662f\u8bd5\u56fe\u9047\u5230\u4e0d\u786e\u5b9a\u7684\u89c2\u5bdf<\/span>\uff08\u4f8b\u5982\u7535\u89c6\u603b\u662f\u663e\u793a\u767d\u566a\u97f3\uff09<\/span>\u3002\u8fd9\u4f1a\u5e72\u6270\u5b66\u5230\u5c42\u7ea7\u5316\u8868\u5f81\u7684\u95ee\u9898\uff0c\u56e0\u4e3a\u5982\u679c\u5bf9\u8fd9\u4e9b\u8868\u5f81\u7684\u5b66\u4e60\u8fd8\u6ca1\u6709\u6536\u655b\uff0c\u5c31\u65e0\u6cd5\u5e94\u5bf9\u9ad8\u5c42\u7ea7\u8868\u5f81\u7684\u4e0d\u786e\u5b9a\u6027\u3002<\/span><\/p>\n


<\/p>\n

\u7ec4\u5408\u6027\u8981\u6c42\u7cbe\u51c6\u7684\u89d2\u8272\u586b\u5145<\/span>\uff08role-filler\uff09<\/span>\u8868\u5f81\u3002\u53ea\u6709\u5c11\u6570\u7684\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\u7814\u7a76\u8003\u8651\u4e86\u7ec4\u5408\u6027\uff0c\u4f8b\u5982\u4f7f\u7528\u81ea\u7136\u8bed\u8a00\uff0c\u5bf9\u52a8\u4f5c\u8fdb\u884c\u8868\u5f81[67]\uff0c\u6216\u57fa\u4e8e\u7b26\u53f7\u903b\u8f91\u8fdb\u884c\u7ec4\u5408\u5f0f\u8868\u5f81[11]\u3002\u7531\u4e8e\u7ec4\u6210\u5f0f\u8868\u5f81\u901a\u5e38\u5efa\u7acb\u5728\u53d8\u503c\u8868\u5f81\u7684\u57fa\u7840\u4e0a\uff0c\u56e0\u6b64\uff0c\u8981\u7f13\u89e3\u8fd9\u4e00\u7f3a\u9677\uff0c\u9700\u8981\u9996\u5148\u5728\u66f4\u57fa\u7840\u7684\u5c42\u9762\u4e0a\u5bf9\u8fd9\u4e9b\u8868\u5f81\u8fdb\u884c\u7814\u7a76\u3002\u56fe\u795e\u7ecf\u7f51\u7edc<\/span>\uff08graph neural networks\uff09<\/span>\uff0c\u4e00\u79cd\u5bf9\u5b9e\u4f53\u548c\u5176\u95f4\u7684\u5173\u7cfb\u901a\u8fc7\u4e00\u4e2a\u56fe\u4e2d\u7684\u8282\u70b9<\/span>\uff08node\uff09<\/span>\u548c\u8fde\u8fb9<\/span>\uff08edge\uff09<\/span>\u7f16\u7801\u7684\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\uff0c\u662f\u5b9e\u73b0\u7ec4\u5408\u5f0f\u8868\u5f81\u7684\u53ef\u884c\u65b9\u5f0f\u3002\u5f53\u63d0\u4f9b\u4e86\u80cc\u540e\u7684\u56fe\u7ed3\u6784\u540e\uff0c\u56fe\u795e\u7ecf\u7f51\u7edc\u80fd\u591f\u66f4\u597d\u5730\u6cdb\u5316\uff0c\u5e76\u652f\u6301\u5173\u7cfb\u63a8\u7406[115]\u3002\u7136\u800c\uff0c\u4ece\u6570\u636e\u4e2d\u53d1\u73b0\u4e00\u4e2a\u5408\u9002\u7684\u62bd\u8c61\u56fe\uff0c\u4ecd\u7136\u662f\u4e00\u4e2a\u5f00\u653e\u95ee\u9898\u3002<\/span><\/p>\n


<\/p>\n

\u9002\u914d\u63a8\u65ad\u7684\u72b6\u6001\u62bd\u8c61\u9700\u8981\u5f3a\u5b66\u4e60\u4fe1\u53f7\u3002\u6700\u8fd1\u7684\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\u65b9\u5f0f[86,89]\uff0c\u5bf9\u5c42\u7ea7\u4e2d\u7684\u6bcf\u4e00\u5c42\uff0c\u4f7f\u7528\u76f8\u540c\u7684\u72b6\u6001\u8868\u5f81\uff0c\u7136\u800c\uff0c\u5728\u8ba4\u77e5\u79d1\u5b66\u4e0a\uff0c\u66f4\u53ef\u884c\u7684\u662f\u8ba9\u4e0d\u540c\u7684\u62bd\u8c61\u5c42\u7ea7\uff0c\u6709\u8be5\u5c42\u7ea7\u63a8\u65ad\u6240\u9700\u8981\u7684\u62bd\u8c61\u6c34\u5e73\u3002\u4f8b\u5982\uff0c\u9ad8\u5c42\u7ea7\u7684\u63a8\u65ad\uff0c\u53ea\u9700\u8981\u77e5\u9053\u7269\u4f53\u662f\u5426\u80fd\u591f\u88ab\u6293\u53d6\uff0c\u800c\u4e0d\u5fc5\u8003\u8651\u7269\u4f53\u7684\u5177\u4f53\u5f62\u72b6\u3002\u5bf9\u4e8e\u4f4e\u5c42\u7ea7\u7684\u63a8\u65ad\uff0c\u8fd9\u6837\u7684\u7ec6\u8282\u4fe1\u606f\u662f\u5fc5\u987b\u7684\u3002\u6709\u7684\u5f3a\u5316\u5b66\u4e60\u65b9\u6cd5\uff0c\u901a\u8fc7\u5377\u79ef\u81ea\u7f16\u7801\u5668\uff0c\u4ece\u89c6\u89c9\u6570\u636e\u4e2d\u5b66\u4e60\u72b6\u6001\u62bd\u8c61[114]\u3002\u7136\u800c\u8fd9\u6837\u7684\u8868\u5f81\uff0c\u662f\u4e3a\u4e86\u91cd\u73b0\u8f93\u5165\u7684\u89c6\u89c9\u7279\u5f81\u6700\u4f18\u5316\u7684\uff0c\u800c\u4e0d\u662f\u4e3a\u51b3\u7b56\u6700\u4f18\u5316\u7684\u3002\u6b64\u5916\uff0c\u5728\u8fd1\u671f\u57fa\u4e8e\u89c6\u89c9\u7684\u67b6\u6784[11,12,67,91-94]\u4e2d\uff0c\u5f3a\u5316\u5b66\u4e60\u667a\u80fd\u4f53\u6240\u6709\u7684\u5c42\u7ea7\uff0c\u90fd\u63a5\u6536\u76f8\u540c\u7684\u56fe\u50cf\u5d4c\u5165\u3002\u56e0\u6b64\uff0c\u8fd9\u91cc\u7684\u62bd\u8c61\u5e76\u4e0d\u662f\u4e3a\u63a8\u65ad\u9002\u914d\u7684\u3002\u5b66\u4e60\u5f97\u5230\u9002\u914d\u63a8\u65ad\u4fe1\u53f7\u7684\u95ee\u9898\u5728\u4e8e\uff1a\u4e0e\u5176\u5b83\u65b9\u6cd5\u5982\u76d1\u7763\u5b66\u4e60\u76f8\u6bd4\uff0c\u5f3a\u5316\u5b66\u4e60\u7684\u5956\u52b1\u4fe1\u53f7\u5bf9\u5b66\u4e60\u8fc7\u7a0b\u7684\u63a8\u52a8\u4f5c\u7528\u76f8\u5bf9\u8f83\u5f31\u3002<\/span><\/p>\n


<\/p>\n


<\/p>\n

\n
\n
\n
\n
\n

<\/section>\n

<\/section>\n
\n
<\/svg><\/section>\n<\/section>\n<\/section>\n
\n

7. \u5c55\u671b\u548c\u5bf9\u4e0a\u8ff0\u6311\u6218\u7684\u5efa\u8bae<\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n

<\/span>
<\/p>\n

\u5728\u4e0b\u6587\u4e2d\uff0c\u6211\u4eec\u6982\u8ff0\u4e86\u8ba1\u7b97\u6027\u5173\u952e\u65b9\u6cd5\u5982\u4f55\u6709\u52a9\u4e8e\u89e3\u51b3\u4e0a\u8ff0\u6311\u6218\u3002<\/span><\/p>\n


<\/p>\n

\u5bf9\u4f53\u9a8c\u7684\u56de\u6eaf\u5f0f\u5206\u6790\u3002\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\u7684\u4e00\u4e2a\u6838\u5fc3\u95ee\u9898\uff0c\u662f\u53ea\u6709\u5728\u4f4e\u5c42\u7684\u5b66\u4e60\u8fc7\u7a0b\u5df2\u6536\u655b\u4e4b\u540e\uff0c\u9ad8\u5c42\u7684\u8868\u5f81\u624d\u4f1a\u51fa\u73b0[64]\u3002\u6211\u4eec\u60f3\u6307\u51fa\uff0c\u5bf9\u4f53\u9a8c\u7684\u56de\u987e\u5f0f\u5206\u6790\uff0c\u53ef\u4ee5\u7f13\u89e3\u8be5\u95ee\u9898\u3002\u901a\u8fc7\u540e\u89c1<\/span>\uff08hindsight\uff09<\/span>\u6765\u8bad\u7ec3\u5f3a\u5316\u5b66\u4e60\u4e2d\u7684\u7b56\u7565[86,116]\uff0c\u5373\u5c06\u4e00\u4e2a\u7b56\u7565\u6267\u884c\u540e\u8fbe\u5230\u7684\u4e16\u754c\u72b6\u6001\uff0c\u5f53\u6210\u662f\u60f3\u8981\u7684\u76ee\u6807\u72b6\u6001\uff0c\u5df2\u7ecf\u8bc1\u660e\u4e86\u4e0a\u8ff0\u8bba\u70b9\u3002\u6211\u4eec\u6307\u51fa\uff0c\u540e\u89c1\u5b66\u4e60<\/span>\uff08hindsight learning\uff09<\/span>\u662f\u66f4\u4e3a\u4e00\u822c\u7684\u5bf9\u4e0e\u7ecf\u9a8c\u7684\u56de\u6eaf\u5f0f\u5206\u6790\u7684\u7279\u6b8a\u5f62\u5f0f\u3002\u4f8b\u5982\uff0c\u53ef\u4ee5\u5728\u4e8b\u540e\u4e5f\u5b66\u4e60\u5c42\u7ea7\u5316\u7684\u524d\u5411\u6a21\u578b\uff0c\u800c\u4e0d\u9700\u8981\u9ad8\u7ea7\u522b\u5c42\u7b49\u5f85\u4f4e\u7ea7\u522b\u5c42\u7684\u6536\u655b\u3002\u6211\u4eec\u8bbe\u60f3\uff0c\u5c06\u8fd8\u6ca1\u6709\u5b8c\u5168\u8bad\u7ec3\u597d\u7684\u4f4e\u5c42\uff0c\u5728\u8fc7\u53bb\u9519\u8bef\u52a8\u4f5c\u7684\u7ed3\u679c\uff0c\u91cd\u65b0\u770b\u4f5c\u60f3\u8981\u7684\u7ed3\u679c\uff0c\u80fd\u591f\u7a33\u5b9a\u9ad8\u5c42\u7ea7\u7684\u8868\u5f81\u3002\u8fd9\u4e00\u6280\u672f\u80fd\u591f\u6d88\u9664\u6a21\u578b\u7684\u504f\u5dee\u3002<\/span><\/p>\n


<\/p>\n

\u5143\u4e0d\u786e\u5b9a\u6a21\u578b<\/span>\uff08Meta-uncentainty model\uff09<\/span>\u3002\u5728\u8ba4\u77e5\u6a21\u62df\u548c\u5185\u5728\u52a8\u673a\u4e2d\uff0c\u9884\u6d4b\u8bef\u5dee\u53d1\u6325\u4e86\u5173\u952e\u4f5c\u7528\u3002\u5728\u5173\u4e8e\u795e\u7ecf\u8ba4\u77e5\u57fa\u7840\u7684\u7ae0\u8282\uff0c\u6211\u4eec\u63a2\u8ba8\u4e86\u81ea\u7531\u80fd\u539f\u7406\uff0c\u5176\u4e2d\u5305\u542b\u4e86\u4e24\u4e2a\u57fa\u4e8e\u9884\u6d4b\u5904\u7406\u6a21\u578b\u7684\u770b\u4f3c\u77db\u76fe\u7684\u673a\u5236\u3002\u5728\u4e00\u65b9\u9762\uff0c\u662f\u4e3b\u52a8\u63a8\u65ad[117]\uff0c\u8bd5\u56fe\u6700\u5c0f\u5316\u957f\u671f\u7684\u60ca\u8bb6\u5ea6\u3002\u8fd9\u901a\u8fc7\u6700\u5c0f\u5316\u524d\u5411\u6a21\u578b\u7684\u9884\u6d4b\u8bef\u5dee\u5b9e\u73b0\u3002\u53e6\u4e00\u65b9\u9762\uff0c\u8fd8\u6709\u4e3b\u52a8\u5b66\u4e60\uff0c\u5b83\u5bfb\u6c42\u4fe1\u606f\u589e\u76ca\u7684\u6700\u5927\u5316\u3002\u8fd9\u53ef\u4ee5\u901a\u8fc7\u4e0d\u540c\u7684\u5185\u5728\u5956\u52b1\u51fd\u6570\uff0c\u4f7f\u4e0d\u786e\u5b9a\u6027[118]\u6216\u9884\u6d4b\u8bef\u5dee[89]\u6700\u5927\u5316\u6765\u5b9e\u73b0\u3002Schwartenbeck\u7b49\u4eba[117]\uff0c\u901a\u8fc7\u5efa\u7acb\u9884\u671f\u7684\u4e0d\u786e\u5b9a\u6027[119]\uff0c\u6765\u89e3\u51b3\u8fd9\u4e00\u77db\u76fe\uff0c\u7edf\u4e00\u4e24\u79cd\u673a\u5236\u3002\u8fd9\u6837\uff0c\u9884\u671f\u7684\u548c\u975e\u9884\u671f\u7684\u4e0d\u786e\u5b9a\u6027\uff0c\u5c31\u53ef\u4ee5\u901a\u8fc7\u5bf9\u4e0d\u786e\u5b9a\u6027\u7684\u5143\u5206\u6790\u6765\u5e94\u5bf9\u3002\u6211\u4eec\u63d0\u8bae\uff0c\u5bf9\u9ad8\u7684\u5143\u4e0d\u786e\u5b9a\u6027\u8fdb\u884c\u5185\u5728\u5956\u52b1\uff0c\u80fd\u591f\u5e26\u6765\u63a2\u7d22\u884c\u4e3a\uff0c\u5e26\u6765\u77ed\u671f\u8f83\u9ad8\u4f46\u957f\u671f\u53ef\u9884\u671f\u8f83\u4f4e\u7684\u5143\u4e0d\u786e\u5b9a\u6027\u3002\u6b64\u5916\uff0c\u65b0\u83b7\u5f97\u7684\u5bf9\u524d\u5411\u6a21\u578b\u4e0d\u786e\u5b9a\u6027\u7684\u786e\u5b9a\u6027\uff0c\u80fd\u591f\u5e26\u6765\u66f4\u51c6\u786e\u7684\uff0c\u80fd\u591f\u77e5\u6653\u81ea\u8eab\u5c40\u9650\u7684\u9884\u6d4b\u8fc7\u7a0b[120]\u3002<\/span><\/p>\n


<\/p>\n

\u53cc\u5411\u52a8\u529b\u5b66\u6a21\u578b\u3002\u53cc\u5411\u52a8\u529b\u5b66\u6a21\u578b\u5305\u62ec\u4e86\u524d\u5411\u53ca\u540e\u5411\u63a8\u65ad\u7684\u6a21\u578b\u3002\u6211\u4eec\u8ba4\u4e3a\uff0c\u8fd9\u6837\u7684\u7ec4\u5408\uff0c\u6709\u5229\u4e8e\u5c42\u7ea7\u5316\u67b6\u6784\u4e2d\u7684\u8868\u5f81\u62bd\u8c61\u3002\u8fd9\u4e00\u731c\u60f3\u662f\u57fa\u4e8ePathak\u7b49\u4eba[105]\u548cHafner\u7b49\u4eba[112]\u7684\u5de5\u4f5c\u7684\u3002\u4ed6\u4eec\u7ed3\u5408\u524d\u5411\u548c\u53cd\u5411\u6a21\u578b\uff0c\u901a\u8fc7\u81ea\u76d1\u7763\u5b66\u4e60\uff0c\u6765\u4ea7\u751f\u9690\u85cf\u8868\u5f81\u7684\u7814\u7a76\u3002\u3002\u53cc\u5411\u6a21\u578b\u5728\u4e00\u4e2a\u6f5c\u5728\u7684\u7a7a\u95f4\u4e0a\u8fd0\u884c\uff0c\u7531\u4e00\u4e2a\u62bd\u8c61\u51fd\u6570\u751f\u6210\u3002\u901a\u8fc7\u5728\u9690\u85cf\u5c42\u4e0a\uff0c\u540c\u65f6\u8fdb\u884c\u524d\u5411\u548c\u540e\u5411\u7684\u9884\u6d4b\uff0c\u62bd\u8c61\u51fd\u6570\u5177\u6709\u4e86\u4e24\u4e2a\u91cd\u8981\u7684\u7279\u5f81\u3002\u4e00\uff0c\u6f5c\u7a7a\u95f4\u5177\u6709\u964d\u566a\u80fd\u529b\uff0c\u56e0\u4e3a\u53ea\u6709\u6f5c\u89c2\u5bdf\u7684\u53ef\u9884\u6d4b\u90e8\u5206\u51b3\u5b9a\u4e86\u524d\u5411\u548c\u540e\u5411\u9884\u6d4b\u3002\u4e8c\uff0c\u7531\u4e8e\u6a21\u578b\u662f\u81ea\u76d1\u7763\u7684\uff0c\u56e0\u6b64\u63d0\u4f9b\u4e86\u4e00\u4e2a\u6bd4\u76f8\u5bf9\u8f83\u5f31\u7684\u5956\u52b1\u4fe1\u53f7\u66f4\u7a33\u5b9a\u7684\u53cd\u9988\u4fe1\u53f7\u3002\u8fd9\u7f13\u89e3\u4e86\u9884\u6d4b\u8bef\u5dee\uff0c\u5e76\u52a0\u5f3a\u4e86\u6574\u4f53\u7684\u5b66\u4e60\u4fe1\u53f7\u3002<\/span><\/p>\n


<\/p>\n


<\/p>\n

\n
\n
\n
\n
\n

<\/section>\n

<\/section>\n
\n
<\/svg><\/section>\n<\/section>\n<\/section>\n
\n

8. \u603b\u7ed3<\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n


<\/span><\/p>\n

\u8be5\u7efc\u8ff0\u5bf9\u5c42\u7ea7\u5316\u95ee\u9898\u89e3\u51b3\u7684\u8ba4\u77e5\u57fa\u7840\uff0c\u53ca\u5982\u4f55\u5728\u5f53\u524d\u7684\u5c42\u7ea7\u5316\u5f3a\u5316\u5b66\u4e60\u4e2d\u5b9e\u73b0\u8fd9\u4e9b\u673a\u5236\uff0c\u8fdb\u884c\u4e86\u6c47\u603b\u3002\u6211\u4eec\u5c06\u5c0f\u6837\u672c\u7684\u95ee\u9898\u89e3\u51b3\uff0c\u770b\u6210\u662f\u667a\u80fd\u4f53\u7684\u7ec8\u6781\u76ee\u6807\uff0c\u5373\u667a\u80fd\u4f53\u5728\u5c3d\u53ef\u80fd\u5c11\u7684\u5c1d\u8bd5\u6b21\u6570\u4e0b\uff0c\u5b66\u4f1a\u56e0\u679c\u6027\u5730\u89e3\u51b3\u6709\u610f\u4e49\u7684\u95ee\u9898\u3002\u4f5c\u4e3a\u6211\u4eec\u7684\u7684\u6838\u5fc3\u7814\u7a76\u95ee\u9898\uff0c\u6211\u4eec\u547c\u5401\u90a3\u4e9b\u8ba1\u7b97\u4e0a\u7684\u5148\u51b3\u6761\u4ef6\u548c\u673a\u5236\uff0c\u4f7f\u4eba\u5de5\u667a\u80fd\u4f53\u7684\u95ee\u9898\u89e3\u51b3\u80fd\u529b\u80fd\u8fbe\u5230\u667a\u80fd\u52a8\u7269\u7684\u6c34\u5e73\u3002\u6211\u4eec\u8fd8\u6307\u51fa\u7ed3\u5408\u4e0d\u540c\u673a\u5236\u65f6\u9047\u5230\u7684\u6311\u6218\u3002<\/span><\/p>\n


<\/p>\n

\u6211\u4eec\u610f\u8bc6\u5230\uff0c\u6211\u4eec\u603b\u7ed3\u7684\u8ba4\u77e5\u79d1\u5b66\u7406\u8bba\uff0c\u901a\u5e38\u662f\u6709\u4e89\u8bae\u4e14\u77db\u76fe\u7684<\/span>\uff08\u5728\u795e\u7ecf\u8ba4\u77e5\u57fa\u7840\u4e00\u7ae0\uff09<\/span>\u3002\u7136\u800c\u6211\u4eec\u7684\u8ba8\u8bba\u6307\u51fa\uff0c\u7279\u5b9a\u7406\u8bba\u7684\u7ec4\u5408\uff0c\u80fd\u591f\u4ea7\u751f\u7b97\u6cd5\u4e0a\u7684\u534f\u540c\u4f5c\u7528\u3002\u6362\u53e5\u8bdd\u8bf4\uff0c\u6211\u4eec\u63a8\u65ad\uff0c\u7279\u5b9a\u7684\u673a\u5236\u5982\u679c\u5bf9\u67d0\u4e00\u8ba1\u7b97\u67b6\u6784\u6709\u5229\uff0c\u90a3\u4e48\u5b83\u5bf9\u751f\u7269\u7684\u795e\u7ecf\u7cfb\u7edf\u4e5f\u6709\u5229\uff0c\u53cd\u4e4b\u4ea6\u7136\u3002\u8fd9\u6837\u8ba4\u77e5\u79d1\u5b66\u548c\u4eba\u5de5\u667a\u80fd\u53ef\u76f8\u4e92\u542f\u53d1\uff0c\u6c47\u96c6\u6210\u5bf9\u4e24\u4e2a\u9886\u57df\u90fd\u6709\u610f\u4e49\u7684\u95f4\u63a5\uff0c\u52a9\u76ca\u4eba\u5de5\u667a\u80fd\u4e0e\u4eba\u7c7b\u7684\u672a\u6765\u3002<\/span><\/p>\n


<\/p>\n

\n
\n
\n

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

\n
\n
\n
\n
    \n
  1. \n

    Gruber, R. et al. New Caledonian crows use mental representations to solve metatool problems. Curr. Biol. 29, 686\u2013692 (2019).<\/p>\n<\/li>\n

  2. \n

    Butz, M. V. & Kutter, E. F. How the Mind Comes into Being (Oxford Univ. Press, 2017).<\/p>\n<\/li>\n

  3. \n

    Perkins, D. N. & Salomon, G. in International Encyclopedia of Education (eds. Husen T. & Postelwhite T. N.) 6452\u20136457 (Pergamon Press, 1992).<\/p>\n<\/li>\n

  4. \n

    Botvinick, M. M., Niv, Y. & Barto, A. C. Hierarchically organized behavior and its neural foundations: a reinforcement learning perspective. Cognition 113, 262\u2013280 (2009).<\/p>\n<\/li>\n

  5. \n

    Tomov, M. S., Yagati, S., Kumar, A., Yang, W. & Gershman, S. J. Discovery of hierarchical representations for efficient planning.PLoS Comput. Biol. 16, e1007594 (2020).<\/p>\n<\/li>\n

  6. \n

    Arulkumaran, K., Deisenroth, M. P., Brundage, M. & Bharath, A. A. Deep reinforcement learning: a brief survey. IEEE Signal Process. Mag. 34, 26\u201338 (2017).<\/p>\n<\/li>\n

  7. \n

    Li, Y. Deep reinforcement learning: an overview. Preprint at https:\/\/arxiv.org\/abs\/1701.07274 (2018).<\/p>\n<\/li>\n

  8. \n

    Sutton, R. S. & Barto, A. G. Reinforcement Learning: An Introduction 2nd edn (MIT Press, 2018).<\/p>\n<\/li>\n

  9. \n

    Neftci, E. O. & Averbeck, B. B. Reinforcement learning in artificial and biological systems. Nat. Mach. Intell. 1, 133\u2013143 (2019).<\/p>\n<\/li>\n

  10. \n

    Eppe, M., Nguyen, P. D. H. & Wermter, S. From semantics to execution: integrating action planning with reinforcement learning for robotic causal problem-solving. Front. Robot. AI 6, 123 (2019).<\/p>\n<\/li>\n

  11. \n

    Oh, J., Singh, S., Lee, H. & Kohli, P. Zero-shot task generalization with multi-task deep reinforcement learning. In Proc. 34th International Conference on Machine Learning (ICML) (eds. Precup, D. & Teh, Y. W.) 2661\u20132670 (PMLR, 2017).<\/p>\n<\/li>\n

  12. \n

    Sohn, S., Oh, J. & Lee, H. Hierarchical reinforcement learning for zero-shot generalization with subtask dependencies. In Proc. 32nd International Conference on Neural Information Processing Systems (NeurIPS) (eds Bengio S. et al.) Vol. 31, 7156\u20137166 (ACM, 2018).<\/p>\n<\/li>\n

  13. \n

    Hegarty, M. Mechanical reasoning by mental simulation. Trends Cogn. Sci. 8, 280\u2013285 (2004).<\/p>\n<\/li>\n

  14. \n

    Klauer, K. J. Teaching for analogical transfer as a means of improving problem-solving, thinking and learning. Instruct. Sci. 18, 179\u2013192 (1989).<\/p>\n<\/li>\n

  15. \n

    Duncker, K. & Lees, L. S. On problem-solving. Psychol. Monographs 58, No.5 (whole No. 270), 85\u2013101 https:\/\/doi.org\/10.1037\/h0093599 (1945).<\/p>\n<\/li>\n

  16. \n

    Dayan, P. Goal-directed control and its antipodes. Neural Netw. 22, 213\u2013219 (2009).<\/p>\n<\/li>\n

  17. \n

    Dolan, R. J. & Dayan, P. Goals and habits in the brain. Neuron 80, 312\u2013325 (2013).<\/p>\n<\/li>\n

  18. \n

    O\u2019Doherty, J. P., Cockburn, J. & Pauli, W. M. Learning, reward, and decision making. Annu. Rev. Psychol. 68, 73\u2013100 (2017).<\/p>\n<\/li>\n

  19. \n

    Tolman, E. C. & Honzik, C. H. Introduction and removal of reward, and maze performance in rats. Univ. California Publ. Psychol. 4, 257\u2013275 (1930).<\/p>\n<\/li>\n

  20. \n

    Butz, M. V. & Hoffmann, J. Anticipations control behavior: animal behavior in an anticipatory learning classifier system. Adaptive Behav. 10, 75\u201396 (2002).<\/p>\n<\/li>\n

  21. \n

    Miller, G. A., Galanter, E. & Pribram, K. H. Plans and the Structure of Behavior (Holt, Rinehart & Winston, 1960).<\/p>\n<\/li>\n

  22. \n

    Botvinick, M. & Weinstein, A. Model-based hierarchical reinforcement learning and human action control. Philos. Trans. R. Soc. B Biol. Sci. 369, 20130480 (2014).<\/p>\n<\/li>\n

  23. \n

    Wiener, J. M. & Mallot, H. A. \u2019Fine-to-coarse\u2019 route planning and navigation in regionalized environments. Spatial Cogn. Comput. 3, 331\u2013358 (2003).<\/p>\n<\/li>\n

  24. \n

    Stock, A. & Stock, C. A short history of ideo-motor action. Psychol. Res. 68, 176\u2013188 (2004).<\/p>\n<\/li>\n

  25. \n

    Hommel, B., M\u00fcsseler, J., Aschersleben, G. & Prinz, W. The theory of event coding (TEC): a framework for perception and action planning. Behav. Brain Sci. 24, 849\u2013878 (2001).<\/p>\n<\/li>\n

  26. \n

    Hoffmann, J. in Anticipatory Behavior in Adaptive Learning Systems: Foundations, Theories and Systems (eds Butz, M. V. et al.) 44\u201365 (Springer, 2003).<\/p>\n<\/li>\n

  27. \n

    Kunde, W., Elsner, K. & Kiesel, A. No anticipation-no action: the role of anticipation in action and perception. Cogn. Process. 8, 71\u201378 (2007).<\/p>\n<\/li>\n

  28. \n

    Barsalou, L. W. Grounded cognition. Annu. Rev. Psychol. 59, 617\u2013645 (2008).<\/p>\n<\/li>\n

  29. \n

    Butz, M. V. Toward a unified sub-symbolic computational theory of cognition. Front. Psychol. 7, 925 (2016).<\/p>\n<\/li>\n

  30. \n

    Pulverm\u00fcller, F. Brain embodiment of syntax and grammar: discrete combinatorial mechanisms spelt out in neuronal circuits. Brain Lang. 112, 167\u2013179 (2010).<\/p>\n<\/li>\n

  31. \n

    Sutton, R. S., Precup, D. & Singh, S. Between MDPs and semi-MDPs: a framework for temporal abstraction in reinforcement learning. Artif. Intell. 112, 181\u2013211 (1999).<\/p>\n<\/li>\n

  32. \n

    Flash, T. & Hochner, B. Motor primitives in vertebrates and invertebrates. Curr. Opin. Neurobiol. 15, 660\u2013666 (2005).<\/p>\n<\/li>\n

  33. \n

    Schaal, S. in Adaptive Motion of Animals and Machines (eds. Kimura, H. et al.) 261\u2013280 (Springer, 2006).<\/p>\n<\/li>\n

  34. \n

    Feldman, J., Dodge, E. & Bryant, J. in The Oxford Handbook of Linguistic Analysis (eds Heine, B. & Narrog, H.) 111\u2013138 (Oxford Univ. Press, 2009).<\/p>\n<\/li>\n

  35. \n

    Fodor, J. A. Language, thought and compositionality. Mind Lang. 16, 1\u201315 (2001).<\/p>\n<\/li>\n

  36. \n

    Frankland, S. M. & Greene, J. D. Concepts and compositionality: in search of the brain\u2019s language of thought. Annu. Rev. Psychol. 71, 273\u2013303 (2020).<\/p>\n<\/li>\n

  37. \n

    Hummel, J. E. Getting symbols out of a neural architecture. Connection Sci. 23, 109\u2013118 (2011).<\/p>\n<\/li>\n

  38. \n

    Haynes, J. D., Wisniewski, D., Gorgen, K., Momennejad, I. & Reverberi, C. FMRI decoding of intentions: compositionality, hierarchy and prospective memory. In Proc. 3rd International Winter Conference on Brain-Computer Interface (BCI), 1-3 (IEEE, 2015).<\/p>\n<\/li>\n

  39. \n

    G\u00e4rdenfors, P. The Geometry of Meaning: Semantics Based on Conceptual Spaces (MIT Press, 2014).<\/p>\n<\/li>\n

  40. \n

    Lakoff, G. & Johnson, M. Philosophy in the Flesh (Basic Books, 1999).<\/p>\n<\/li>\n

  41. \n

    Eppe, M. et al. A computational framework for concept blending. Artif. Intell. 256, 105\u2013129 (2018).<\/p>\n<\/li>\n

  42. \n

    Turner, M. The Origin of Ideas (Oxford Univ. Press, 2014).<\/p>\n<\/li>\n

  43. \n

    Deci, E. L. & Ryan, R. M. Self-determination theory and the facilitation of intrinsic motivation. Am. Psychol. 55, 68\u201378 (2000).<\/p>\n<\/li>\n

  44. \n

    Friston, K. et al. Active inference and epistemic value. Cogn. Neurosci. 6, 187\u2013214 (2015).<\/p>\n<\/li>\n

  45. \n

    Berlyne, D. E. Curiosity and exploration. Science 153, 25\u201333 (1966).<\/p>\n<\/li>\n

  46. \n

    Loewenstein, G. The psychology of curiosity: a review and reinterpretation. Psychol. Bull. 116, 75\u201398 (1994).<\/p>\n<\/li>\n

  47. \n

    Oudeyer, P.-Y., Kaplan, F. & Hafner, V. V. Intrinsic motivation systems for autonomous mental development. In IEEE Transactions on Evolutionary Computation (eds. Coello, C. A. C. et al.) Vol. 11, 265\u2013286 (IEEE, 2007).<\/p>\n<\/li>\n

  48. \n

    Pisula, W. Play and exploration in animals\u2014a comparative analysis. Polish Psychol. Bull. 39, 104\u2013107 (2008).<\/p>\n<\/li>\n

  49. \n

    Jeannerod, M. Mental imagery in the motor context. Neuropsychologia 33, 1419\u20131432 (1995).<\/p>\n<\/li>\n

  50. \n

    Kahnemann, D. & Tversky, A. in Judgement under Uncertainty: Heuristics and Biases (eds Kahneman, D. et al.) Ch. 14, 201\u2013208 (Cambridge Univ. Press, 1982).<\/p>\n<\/li>\n

  51. \n

    Wells, G. L. & Gavanski, I. Mental simulation of causality. J. Personal. Social Psychol. 56, 161\u2013169 (1989).<\/p>\n<\/li>\n

  52. \n

    Taylor, S. E., Pham, L. B., Rivkin, I. D. & Armor, D. A. Harnessing the imagination: mental simulation, self-regulation and coping. Am. Psychol. 53, 429\u2013439 (1998).<\/p>\n<\/li>\n

  53. \n

    Kaplan, F. & Oudeyer, P.-Y. in Embodied Artificial Intelligence, Lecture Notes in Computer Science Vol. 3139 (eds Iida, F. et al.) 259\u2013270 (Springer, 2004).<\/p>\n<\/li>\n

  54. \n

    Schmidhuber, J. Formal theory of creativity, fun, and intrinsic motivation. IEEE Trans. Auton. Mental Dev. 2, 230\u2013247 (2010).<\/p>\n<\/li>\n

  55. \n

    Friston, K., Mattout, J. & Kilner, J. Action understanding and active inference. Biol. Cybern. 104, 137\u2013160 (2011).<\/p>\n<\/li>\n

  56. \n

    Oudeyer, P.-Y. Computational theories of curiosity-driven learning. In The New Science of Curiosity (ed. Goren Gordon), 43-72 (Nova Science Publishers, 2018); https:\/\/arxiv.org\/abs\/1802.10546<\/p>\n<\/li>\n

  57. \n

    Colombo, M. & Wright, C. First principles in the life sciences: the free-energy principle, organicism and mechanism. Synthese 198, 3463\u20133488 (2021).<\/p>\n<\/li>\n

  58. \n

    Huang, Y. & Rao, R. P. Predictive coding. WIREs Cogn. Sci. 2, 580\u2013593 (2011).<\/p>\n<\/li>\n

  59. \n

    Friston, K. The free-energy principle: a unified brain theory? Nat. Rev. Neurosci. 11, 127\u2013138 (2010).<\/p>\n<\/li>\n

  60. \n

    Knill, D. C. & Pouget, A. The Bayesian brain: the role of uncertainty in neural coding and computation. Trends Neurosci. 27, 712\u2013719 (2004).<\/p>\n<\/li>\n

  61. \n

    Clark, A. Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behav. Brain Sci. 36, 181\u2013204 (2013).<\/p>\n<\/li>\n

  62. \n

    Clark, A. Surfing Uncertainty: Prediction, Action and the Embodied Mind (Oxford Univ. Press, 2016).<\/p>\n<\/li>\n

  63. \n

    Zacks, J. M., Speer, N. K., Swallow, K. M., Braver, T. S. & Reyonolds, J. R. Event perception: a mind\/brain perspective. Psychol. Bull. 133, 273\u2013293 (2007).<\/p>\n<\/li>\n

  64. \n

    Eysenbach, B., Ibarz, J., Gupta, A. & Levine, S. Diversity is all you need: learning skills without a reward function. In International Conference on Learning Representations (ICLR, 2019).<\/p>\n<\/li>\n

  65. \n

    Frans, K., Ho, J., Chen, X., Abbeel, P. & Schulman, J. Meta learning shared hierarchies. In Proc. International Conference on Learning Representations https:\/\/openreview.net\/pdf?id=SyX0IeWAW (ICLR, 2018).<\/p>\n<\/li>\n

  66. \n

    Heess, N. et al. Learning and transfer of modulated locomotor controllers. Preprint at https:\/\/arxiv.org\/abs\/1610.05182 (2016).<\/p>\n<\/li>\n

  67. \n

    Jiang, Y., Gu, S., Murphy, K. & Finn, C. Language as an abstraction for hierarchical deep reinforcement learning. In Neural Information Processing Systems (NeurIPS) (eds. Wallach, H. et al.) 9414\u20139426 (ACM, 2019).<\/p>\n<\/li>\n

  68. \n

    Li, A. C., Florensa, C., Clavera, I. & Abbeel, P. Sub-policy adaptation for hierarchical reinforcement learning. In Proc. International Conference on Learning Representations https:\/\/openreview.net\/forum?id=ByeWogStDS (ICLR, 2020).<\/p>\n<\/li>\n

  69. \n

    Qureshi, A. H. et al. Composing task-agnostic policies with deep reinforcement learning. In Proc. International Conference on Learning Representations https:\/\/openreview.net\/forum?id=H1ezFREtwH (ICLR, 2020).<\/p>\n<\/li>\n

  70. \n

    Sharma, A., Gu, S., Levine, S., Kumar, V. & Hausman, K. Dynamics-aware unsupervised discovery of skills. In Proc. International Conference on Learning Representations https:\/\/openreview.net\/forum?id=HJgLZR4KvH (ICLR, 2020).<\/p>\n<\/li>\n

  71. \n

    Tessler, C., Givony, S., Zahavy, T., Mankowitz, D. J. & Mannor, S. A deep hierarchical approach to lifelong learning in minecraft. In Proc. 31st AAAI Conference on Artificial Intelligence 1553\u20131561 (AAAI, 2017).<\/p>\n<\/li>\n

  72. \n

    Vezhnevets, A. et al. Strategic attentive writer for learning macro-actions. In Neural Information Processing Systems (NIPS) (eds. Lee, D. et al.) 3494\u20133502 (NIPS, 2016).<\/p>\n<\/li>\n

  73. \n

    Devin, C., Gupta, A., Darrell, T., Abbeel, P. & Levine, S. Learning modular neural network policies for multi-task and multi-robot transfer. In Proc. International Conference on Robotics and Automation (ICRA) (eds. Okamura, A. et al.) 2169\u20132176 (IEEE, 2017).<\/p>\n<\/li>\n

  74. \n

    Hejna, D. J., Abbeel, P. & Pinto, L. Hierarchically decoupled morphological transfer. In Proc. International Conference on Machine Learning (ICML) (eds. Daum\u00e9 III, H. & Singh, A.) 11409\u201311420 (PMLR, 2020).<\/p>\n<\/li>\n

  75. \n

    Hamrick, J. B. et al. On the role of planning in model-based deep reinforcement learning. In Proc. International Conference on Learning Representations https:\/\/openreview.net\/pdf?id=IrM64DGB21 (ICLR, 2021).<\/p>\n<\/li>\n

  76. \n

    Sutton, R. S. Integrated architectures for learning, planning, and reacting based on approximating dynamic programming. In Proc. 7th International Conference on Machine Learning (ICML) (eds. Porter, B. W. & Mooney, R. J.) 216\u2013224 (Morgan Kaufmann, 1990).<\/p>\n<\/li>\n

  77. \n

    Nau, D. et al. SHOP2: an HTN planning system. J. Artif. Intell. Res. 20, 379\u2013404 (2003).<\/p>\n<\/li>\n

  78. \n

    MATH  Lyu, D., Yang, F., Liu, B. & Gustafson, S. SDRL: interpretable and data-efficient deep reinforcement learning leveraging symbolic planning. In Proc. AAAI Conference on Artificial Intelligence Vol. 33, 2970\u20132977 (AAAI, 2019).<\/p>\n<\/li>\n

  79. \n

    Ma, A., Ouimet, M. & Cort\u00e9s, J. Hierarchical reinforcement learning via dynamic subspace search for multi-agent planning. Auton. Robot. 44, 485\u2013503 (2020).<\/p>\n<\/li>\n

  80. \n

    Bacon, P.-L., Harb, J. & Precup, D. The option-critic architecture. In Proc. 31st AAAI Conference on Artificial Intelligence 1726\u20131734 (AAAI, 2017).<\/p>\n<\/li>\n

  81. \n

    Dietterich, T. G. State abstraction in MAXQ hierarchical reinforcement learning. In Advances in Neural Information Processing Systems (NIPS) (eds. Solla, S. et al.) Vol. 12, 994\u20131000 (NIPS, 1999).<\/p>\n<\/li>\n

  82. \n

    Kulkarni, T. D., Narasimhan, K. R., Saeedi, A. & Tenenbaum, J. B. Hierarchical deep reinforcement learning: integrating temporal abstraction and intrinsic motivation. In Neural Information Processing Systems (NIPS) (eds. Lee, D. et al.) 3675\u20133683 (NIPS, 2016).<\/p>\n<\/li>\n

  83. \n

    Shankar, T., Pinto, L., Tulsiani, S. & Gupta, A. Discovering motor programs by recomposing demonstrations. In Proc. International Conference on Learning Representations https:\/\/openreview.net\/attachment?id=rkgHY0NYwr&name=original_pdf (ICLR, 2020).<\/p>\n<\/li>\n

  84. \n

    Vezhnevets, A. S., Wu, Y. T., Eckstein, M., Leblond, R. & Leibo, J. Z. Options as responses: grounding behavioural hierarchies in multi-agent reinforcement learning. In Proc. International Conference on Machine Learning (ICML) (eds. Daum\u00e9 III, H. & Singh, A.) 9733\u20139742 (PMLR, 2020).<\/p>\n<\/li>\n

  85. \n

    Ghazanfari, B., Afghah, F. & Taylor, M. E. Sequential association rule mining for autonomously extracting hierarchical task structures in reinforcement learning. IEEE Access 8, 11782\u201311799 (2020).<\/p>\n<\/li>\n

  86. \n

    Levy, A., Konidaris, G., Platt, R. & Saenko, K. Learning multi-level hierarchies with hindsight. In Proc. International Conference on Learning Representations https:\/\/openreview.net\/pdf?id=ryzECoAcY7 (ICLR, 2019).<\/p>\n<\/li>\n

  87. \n

    Nachum, O., Gu, S., Lee, H. & Levine, S. Data-efficient hierarchical reinforcement learning. In Proc. 32nd International Conference on Neural Information Processing Systems (NIPS) (eds. Bengio, S. et al.) 3307\u20133317 (NIPS, 2018).<\/p>\n<\/li>\n

  88. \n

    Rafati, J. & Noelle, D. C. Learning representations in model-free hierarchical reinforcement learning. In Proc. 33rd AAAI Conference on Artificial Intelligence 10009\u201310010 (AAAI, 2019).<\/p>\n<\/li>\n

  89. \n

    R\u00f6der, F., Eppe, M., Nguyen, P. D. H. & Wermter, S. Curious hierarchical actor-critic reinforcement learning. In Proc. International Conference on Artificial Neural Networks (ICANN) (eds. Farka\u0161, I. et al.) 408\u2013419 (Springer, 2020).<\/p>\n<\/li>\n

  90. \n

    Zhang, T., Guo, S., Tan, T., Hu, X. & Chen, F. Generating adjacency-constrained subgoals in hierarchical reinforcement learning. In Neural Information Processing Systems (NIPS) (eds. Larochelle, H. et al.) 21579-21590 (NIPS, 2020).<\/p>\n<\/li>\n

  91. \n

    Lample, G. & Chaplot, D. S. Playing FPS games with deep reinforcement learning. In Proc. 31st AAAI Conference on Artificial Intelligence 2140\u20132146 (AAAI, 2017).<\/p>\n<\/li>\n

  92. \n

    Vezhnevets, A. S. et al. FeUdal networks for hierarchical reinforcement learning. In Proc. 34th International Conference on Machine Learning (ICML) (eds. Precup, D. & Teh, Y. W.) Vol. 70, 3540\u20133549 (PMLR, 2017).<\/p>\n<\/li>\n

  93. \n

    Wulfmeier, M. et al. Compositional Transfer in Hierarchical Reinforcement Learning. In Robotics: Science and System XVI (RSS) (eds. Toussaint M. et al.) (Robotics: Science and Systems Foundation, 2020); https:\/\/arxiv.org\/abs\/1906.11228<\/p>\n<\/li>\n

  94. \n

    Yang, Z., Merrick, K., Jin, L. & Abbass, H. A. Hierarchical deep reinforcement learning for continuous action control. IEEE Trans. Neural Netw. Learn. Syst. 29, 5174\u20135184 (2018).<\/p>\n<\/li>\n

  95. \n

    Toussaint, M., Allen, K. R., Smith, K. A. & Tenenbaum, J. B. Differentiable physics and stable modes for tool-use and manipulation planning. In Proc. Robotics: Science and Systems XIV (RSS) (eds. Kress-Gazit, H. et al.) https:\/\/ipvs.informatik.uni-stuttgart.de\/mlr\/papers\/18-toussaint-RSS.pdf (Robotics: Science and Systems Foundation, 2018).<\/p>\n<\/li>\n

  96. \n

    Akrour, R., Veiga, F., Peters, J. & Neumann, G. Regularizing reinforcement learning with state abstraction. In Proc. IEEE\/RSJ International Conference on Intelligent Robots and Systems (IROS) 534\u2013539 (IEEE, 2018).<\/p>\n<\/li>\n

  97. \n

    Schaul, T. & Ring, M. Better generalization with forecasts. In Proc. 23rd International Joint Conference on Artificial Intelligence (IJCAI) (ed. Rossi, F.) 1656\u20131662 (AAAI, 2013).<\/p>\n<\/li>\n

  98. \n

    Colas, C., Akakzia, A., Oudeyer, P.-Y., Chetouani, M. & Sigaud, O. Language-conditioned goal generation: a new approach to language grounding for RL. Preprint at https:\/\/arxiv.org\/abs\/2006.07043 (2020).<\/p>\n<\/li>\n

  99. \n

    Blaes, S., Pogancic, M. V., Zhu, J. J. & Martius, G. Control what you can: intrinsically motivated task-planning agent. Neural Inf. Process. Syst. 32, 12541\u201312552 (2019).<\/p>\n<\/li>\n

  100. \n

    Haarnoja, T., Hartikainen, K., Abbeel, P. & Levine, S. Latent space policies for hierarchical reinforcement learning. In Proc. International Conference on Machine Learning (ICML) (eds. Dy, J. & Krause, A.) Vol. 4, 2965\u20132975 (PMLR, 2018).<\/p>\n<\/li>\n

  101. \n

    Rasmussen, D., Voelker, A. & Eliasmith, C. A neural model of hierarchical reinforcement learning. PLoS ONE 12, e0180234 (2017).<\/p>\n<\/li>\n

  102. \n

    Riedmiller, M. et al. Learning by playing\u2014solving sparse reward tasks from scratch. In Proc. International Conference on Machine Learning (ICML) (eds. Dy, J. & Krause, A.) Vol. 10, 6910\u20136919 (PMLR, 2018).<\/p>\n<\/li>\n

  103. \n

    Yang, F., Lyu, D., Liu, B. & Gustafson, S. PEORL: integrating symbolic planning and hierarchical reinforcement learning for robust decision-making. In Proc. 27th International Joint Conference on Artificial Intelligence (IJCAI) (ed. Lang, J.) 4860\u20134866 (IJCAI, 2018).<\/p>\n<\/li>\n

  104. \n

    Machado, M. C., Bellemare, M. G. & Bowling, M. A Laplacian framework for option discovery in reinforcement learning. In Proc. International Conference on Machine Learning (ICML) (eds. Precup, D. & Teh, Y. W.) Vol. 5, 3567\u20133582 (PMLR, 2017).<\/p>\n<\/li>\n

  105. \n

    Pathak, D., Agrawal, P., Efros, A. A. & Darrell, T. Curiosity-driven exploration by self-supervised prediction. In Proc. 34th International Conference on Machine Learning (ICML) (eds. Precup, D. & Teh, Y. W.) 2778\u20132787 (PMLR, 2017).<\/p>\n<\/li>\n

  106. \n

    Schillaci, G. et al. Intrinsic motivation and episodic memories for robot exploration of high-dimensional sensory spaces. Adaptive Behav. 29 549\u2013566 (2020).<\/p>\n<\/li>\n

  107. \n

    Colas, C., Fournier, P., Sigaud, O., Chetouani, M. & Oudeyer, P.-Y. CURIOUS: intrinsically motivated modular multi-goal reinforcement learning. In Proc. International Conference on Machine Learning (ICML) (eds. Chaudhuri, K. & Salakhutdinov, R.) 1331\u20131340 (PMLR, 2019).<\/p>\n<\/li>\n

  108. \n

    Hafez, M. B., Weber, C., Kerzel, M. & Wermter, S. Improving robot dual-system motor learning with intrinsically motivated meta-control and latent-space experience imagination. Robot. Auton. Syst. 133, 103630 (2020).<\/p>\n<\/li>\n

  109. \n

    Yamamoto, K., Onishi, T. & Tsuruoka, Y. Hierarchical reinforcement learning with abductive planning. In Proc. ICML\/IJCAI\/AAMAS 2018 Workshop on Planning and Learning (PAL-18) (2018).<\/p>\n<\/li>\n

  110. \n

    Wu, B., Gupta, J. K. & Kochenderfer, M. J. Model primitive hierarchical lifelong reinforcement learning. In Proc. International Joint Conference on Autonomous Agents and Multiagent Systems (AAMAS) (eds. Agmon, N. et al.) Vol. 1, 34\u201342 (IFAAMAS, 2019).<\/p>\n<\/li>\n

  111. \n

    Li, Z., Narayan, A. & Leong, T. Y. An efficient approach to model-based hierarchical reinforcement learning. In Proc. 31st AAAI Conference on Artificial Intelligence 3583\u20133589 (AAAI, 2017).<\/p>\n<\/li>\n

  112. \n

    Hafner, D., Lillicrap, T. & Norouzi, M. Dream to control: learning behaviors by latent imagination. In Proc. International Conference on Learning Representations https:\/\/openreview.net\/pdf?id=S1lOTC4tDS (ICLR, 2020).<\/p>\n<\/li>\n

  113. \n

    Deisenroth, M. P., Rasmussen, C. E. & Fox, D. Learning to control a low-cost manipulator using data-efficient reinforcement learning. In Robotics: Science and Systems VII (RSS) (eds. Durrant-Whyte, H. et al.) 57\u201364 (Robotics: Science and Systems Foundation, 2011).<\/p>\n<\/li>\n

  114. \n

    Ha, D. & Schmidhuber, J. Recurrent world models facilitate policy evolution. In Proc. 32nd International Conference on Neural Information Processing Systems (NeurIPS) (eds. Bengio, S. et al.) 2455\u20132467 (NIPS, 2018).<\/p>\n<\/li>\n

  115. \n

    Battaglia, P. W. et al. Relational inductive biases, deep learning and graph networks. Preprint at https:\/\/arxiv.org\/abs\/1806.01261 (2018).<\/p>\n<\/li>\n

  116. \n

    Andrychowicz, M. et al. Hindsight experience replay. In Proc. Neural Information Processing Systems (NIPS) (eds. Guyon I. et al.) 5048\u20135058 (NIPS, 2017); https:\/\/papers.nips.cc\/paper\/7090-hindsight-experience-replay.pdf<\/p>\n<\/li>\n

  117. \n

    Schwartenbeck, P. et al. Computational mechanisms of curiosity and goal-directed exploration. eLife 8, e41703 (2019).<\/p>\n<\/li>\n

  118. \n

    Haarnoja, T., Zhou, A., Abbeel, P. & Levine, S. Soft actor-critic: off-policy maximum entropy deep reinforcement learning with a stochastic actor. In Proc. International Conference on Machine Learning (ICML) (eds. Dy, J. & Krause, A.) 1861\u20131870 (PMLR, 2018).<\/p>\n<\/li>\n

  119. \n

    Yu, A. J. & Dayan, P. Uncertainty, neuromodulation and attention. Neuron 46, 681\u2013692 (2005).<\/p>\n<\/li>\n

  120. \n

    Baldwin, D. A. & Kosie, J. E. How does the mind render streaming experience as events? Top. Cogn. Sci. 13, 79\u2013105 (2021).<\/p>\n<\/li>\n<\/ol>\n

    <\/span>
    <\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

    \n
    \n

    \uff08\u53c2\u8003\u6587\u732e\u53ef\u4e0a\u4e0b\u6ed1\u52a8<\/strong>\u67e5\u770b\uff09<\/p>\n<\/section>\n<\/section>\n


    <\/span><\/section>\n

    <\/span><\/section>\n
    \n
    \n
    \n
    \n

    \u590d\u6742\u79d1\u5b66\u6700\u65b0\u8bba\u6587<\/strong><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

    \n


    <\/p>\n<\/section>\n

    \u96c6\u667a\u6591\u56fe\u9876\u520a\u8bba\u6587\u901f\u9012\u680f\u76ee\u4e0a\u7ebf\u4ee5\u6765\uff0c\u6301\u7eed\u6536\u5f55\u6765\u81eaNature\u3001Science<\/span>\u7b49\u9876\u520a\u7684\u6700\u65b0\u8bba\u6587\uff0c\u8ffd\u8e2a\u590d\u6742\u7cfb\u7edf\u3001\u7f51\u7edc\u79d1\u5b66\u3001\u8ba1\u7b97\u793e\u4f1a\u79d1\u5b66\u7b49\u9886\u57df\u7684\u524d\u6cbf\u8fdb\u5c55\u3002\u73b0\u5728\u6b63\u5f0f\u63a8\u51fa\u8ba2\u9605\u529f\u80fd\uff0c\u6bcf\u5468\u901a\u8fc7\u5fae\u4fe1\u670d\u52a1\u53f7\u300c\u96c6\u667a\u6591\u56fe\u300d\u63a8\u9001\u8bba\u6587\u4fe1\u606f\u3002\u626b\u63cf\u4e0b\u65b9\u4e8c\u7ef4\u7801\u5373\u53ef\u4e00\u952e\u8ba2\u9605\uff1a<\/span>
    <\/p>\n

    <\/p>\n

    \n


    <\/p>\n<\/section>\n


    <\/p>\n

    \n

    \u63a8\u8350\u9605\u8bfb<\/span><\/strong><\/strong><\/p>\n