{"id":40086,"date":"2023-01-28T18:30:25","date_gmt":"2023-01-28T10:30:25","guid":{"rendered":"https:\/\/swarma.org\/?p=40086"},"modified":"2023-01-28T18:30:25","modified_gmt":"2023-01-28T10:30:25","slug":"%e5%89%8d%e6%b2%bf%e7%bb%bc%e8%bf%b0%ef%bc%9a%e9%9d%a2%e5%90%91%e6%b7%b1%e5%ba%a6%e5%ad%a6%e4%b9%a0%e7%9a%84%e9%9b%86%e4%bd%93%e6%99%ba%e8%83%bd","status":"publish","type":"post","link":"https:\/\/swarma.org\/?p=40086","title":{"rendered":"\u524d\u6cbf\u7efc\u8ff0\uff1a\u9762\u5411\u6df1\u5ea6\u5b66\u4e60\u7684\u96c6\u4f53\u667a\u80fd"},"content":{"rendered":"
\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

\u5728\u8fc7\u53bb\u7684\u5341\u5e74\u91cc\uff0c\u6211\u4eec\u89c1\u8bc1\u4e86\u6df1\u5ea6\u5b66\u4e60\u7684\u5d1b\u8d77\uff0c\u9010\u6e10\u5728\u4eba\u5de5\u667a\u80fd\u9886\u57df\u5360\u636e\u4e3b\u5bfc\u5730\u4f4d\u3002\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\u4ee5\u53ca\u5177\u6709\u5927\u5185\u5b58\u7684\u786c\u4ef6\u52a0\u901f\u5668\u7684\u8fdb\u6b65\uff0c\u52a0\u4e0a\u5927\u578b\u6570\u636e\u96c6\u7684\u53ef\u7528\u6027\uff0c\u4f7f\u5f97\u4ece\u4e1a\u8005\u80fd\u591f\u8bad\u7ec3\u548c\u90e8\u7f72\u590d\u6742\u7684\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\uff0c\u5e76\u5728\u8de8\u8d8a\u8ba1\u7b97\u673a\u89c6\u89c9\u3001\u81ea\u7136\u8bed\u8a00\u5904\u7406\u548c\u5f3a\u5316\u5b66\u4e60\u7b49\u591a\u4e2a\u9886\u57df\u7684\u4efb\u52a1\u4e0a\u53d6\u5f97\u6700\u5148\u8fdb\u7684\u6027\u80fd\u3002\u7136\u800c\uff0c\u968f\u7740\u8fd9\u4e9b\u795e\u7ecf\u7f51\u7edc\u53d8\u5f97\u66f4\u5927\u3001\u66f4\u590d\u6742\u3001\u5e94\u7528\u66f4\u5e7f\u6cdb\uff0c\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u7684\u57fa\u672c\u95ee\u9898\u53d8\u5f97\u66f4\u52a0\u660e\u663e\u3002\u4f17\u6240\u5468\u77e5\uff0c\u6700\u5148\u8fdb\u7684\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u5b58\u5728\u5404\u79cd\u95ee\u9898\uff0c\u5305\u62ec\u7a33\u5065\u6027\u5dee\u3001\u65e0\u6cd5\u9002\u5e94\u65b0\u7684\u4efb\u52a1\u8bbe\u7f6e\uff0c\u4ee5\u53ca\u9700\u8981\u4e25\u683c\u548c\u4e0d\u7075\u6d3b\u7684\u914d\u7f6e\u5047\u8bbe\u3002\u901a\u5e38\u5728\u81ea\u7136\u754c\u89c2\u5bdf\u5230\u7684\u96c6\u4f53\u884c\u4e3a\uff0c\u5f80\u5f80\u4f1a\u4ea7\u751f\u7a33\u5065\u3001\u9002\u5e94\u6027\u5f3a\u3001\u5bf9\u73af\u5883\u914d\u7f6e\u7684\u5047\u8bbe\u4e0d\u90a3\u4e48\u4e25\u683c\u7684\u7cfb\u7edf\u3002\u96c6\u4f53\u667a\u80fd\uff08Collective Intelligence\uff09\u4f5c\u4e3a\u4e00\u4e2a\u9886\u57df\uff0c\u7814\u7a76\u7684\u662f\u4ece\u8bb8\u591a\u4e2a\u4f53\u7684\u4e92\u52a8\u4e2d\u6d8c\u73b0\u7684\u7fa4\u4f53\u667a\u80fd\uff08Group Intelligence\uff09\u3002\u5728\u8fd9\u4e2a\u9886\u57df\u4e2d\uff0c\u8bf8\u5982\u81ea\u7ec4\u7ec7\u3001\u6d8c\u73b0\u884c\u4e3a\u3001\u7fa4\u4f18\u5316\u548c\u5143\u80de\u81ea\u52a8\u673a\u7b49\u601d\u60f3\u88ab\u5f00\u53d1\u51fa\u6765\uff0c\u4ee5\u6a21\u62df\u548c\u89e3\u91ca\u590d\u6742\u7cfb\u7edf\u3002\u56e0\u6b64\uff0c\u7814\u7a76\u8005\u5f88\u81ea\u7136\u5730\u4f1a\u5c06\u8fd9\u4e9b\u60f3\u6cd5\u878d\u5165\u5230\u8f83\u65b0\u7684\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\u4e2d\u3002\u5728\u8fd9\u7bc7\u7efc\u8ff0\u4e2d\uff0c\u4f5c\u8005\u5c06\u63d0\u4f9b\u4e00\u4e2a\u6d89\u53ca\u590d\u6742\u7cfb\u7edf\u7684\u795e\u7ecf\u7f51\u7edc\u7814\u7a76\u5386\u53f2\u80cc\u666f\uff0c\u5e76\u5f3a\u8c03\u73b0\u4ee3\u6df1\u5ea6\u5b66\u4e60\u7814\u7a76\u4e2d\u51e0\u4e2a\u6d3b\u8dc3\u7684\u9886\u57df\uff0c\u8fd9\u4e9b\u9886\u57df\u7ed3\u5408\u4e86\u96c6\u4f53\u667a\u80fd\u7684\u539f\u5219\u6765\u63d0\u9ad8\u5176\u80fd\u529b\u3002\u6211\u4eec\u5e0c\u671b\u8fd9\u7bc7\u8bc4\u8bba\u53ef\u4ee5\u4f5c\u4e3a\u590d\u6742\u7cfb\u7edf\u548c\u6df1\u5ea6\u5b66\u4e60\u793e\u533a\u4e4b\u95f4\u7684\u6865\u6881\u3002<\/span><\/strong><\/span><\/strong><\/p>\n


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

\n

\u5173\u952e\u8bcd\uff1a\u96c6\u4f53\u667a\u80fd\uff0c\u6df1\u5ea6\u5b66\u4e60\uff0c\u591a\u4e3b\u4f53\u6a21\u578b\uff0c\u5f3a\u5316\u5b66\u4e60<\/span><\/strong><\/span><\/strong><\/strong><\/strong><\/span><\/strong><\/span><\/strong><\/span><\/strong><\/strong><\/span><\/strong><\/span><\/strong><\/span><\/strong><\/strong><\/strong><\/span><\/strong><\/span><\/strong><\/strong><\/strong><\/strong><\/strong><\/strong><\/strong><\/strong><\/strong><\/strong><\/strong><\/strong><\/strong><\/strong><\/strong><\/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

David Ha,  Yujin Tang<\/span> <\/span>| \u4f5c\u8005<\/span><\/span><\/p>\n

\u5218\u5fd7\u822a<\/span> <\/span>| \u7ffb\u8bd1<\/strong><\/span><\/span><\/p>\n

\u5218\u57f9\u6e90<\/span> <\/span>| \u5ba1\u6821<\/span><\/strong><\/span><\/p>\n

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


<\/p>\n


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

\n
<\/section>\n<\/section>\n
\n
\n
\n
\n
\u8bba\u6587\u9898\u76ee\uff1aCollective intelligence for deep learning: A survey of recent developments<\/span><\/section>\n
\u8bba\u6587\u94fe\u63a5\uff1a<\/span>https:\/\/dl.acm.org\/doi\/10.1177\/26339137221114874<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/blockquote>\n


<\/p>\n


<\/section>\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\u76ee\u5f55<\/strong><\/section>\n

<\/section>\n
\n

1. \u5f15\u8a00<\/span><\/strong><\/p>\n

2. \u80cc\u666f\uff1a\u96c6\u4f53\u667a\u80fd<\/span><\/strong><\/p>\n

3. \u5386\u53f2\u80cc\u666f\uff1a\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc<\/span><\/strong><\/p>\n

4. \u6df1\u5ea6\u5b66\u4e60\u7684\u96c6\u4f53\u667a\u80fd<\/span><\/strong><\/p>\n

5. \u8ba8\u8bba<\/span><\/strong><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n


<\/section>\n

<\/section>\n

\n
\n
\n
\n
\n

<\/section>\n

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

1. \u5f15\u8a00<\/span><\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n
 <\/span><\/section>\n
\u6df1\u5ea6\u5b66\u4e60<\/span>\uff08Deep learning\uff09<\/span>\u662f\u4e00\u7c7b\u673a\u5668\u5b66\u4e60\u65b9\u6cd5\uff0c\u4f7f\u7528\u591a\u5c42<\/span>\uff08\u201c\u6df1\u5ea6\u201d\uff09<\/span>\u795e\u7ecf\u7f51\u7edc\u8fdb\u884c\u8868\u5f81\u5b66\u4e60\u3002\u867d\u7136\u7528\u53cd\u5411\u4f20\u64ad\u7b97\u6cd5\u8bad\u7ec3\u7684\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\u6700\u65e9\u5df2\u7ecf\u572820\u4e16\u7eaa80\u5e74\u4ee3\u51fa\u73b0<\/span>\uff08Schmidhuber, 2014\uff09<\/span>\uff0c\u4f46\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u76f4\u52302012\u5e74\u624d\u53d7\u5230\u5e7f\u6cdb\u5173\u6ce8\uff0c\u5f53\u65f6\u5728GPU\u4e0a\u8bad\u7ec3\u7684\u6df1\u5ea6\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\u89e3\u51b3\u65b9\u6848<\/span>\uff08Krizhevsky et al. 2012\uff09<\/span>\u5728\u5e74\u5ea6\u56fe\u50cf\u8bc6\u522b\u6bd4\u8d5b<\/span>\uff08Deng et al. 2009\uff09<\/span>\u4e2d\u4ee5\u660e\u663e\u4f18\u52bf\u6218\u80dc\u975e\u6df1\u5ea6\u5b66\u4e60\u7684\u65b9\u6cd5\u3002\u8fd9\u4e00\u6210\u529f\u8868\u660e\uff0c\u5f53\u6df1\u5ea6\u5b66\u4e60\u548c\u786c\u4ef6\u52a0\u901f\u4ee5\u53ca\u5927\u578b\u6570\u636e\u96c6\u7684\u53ef\u7528\u6027\u76f8\u7ed3\u5408\u65f6\uff0c\u80fd\u591f\u5728\u975e\u7410\u788e\u7684\u4efb\u52a1\u4e2d\u53d6\u5f97\u6bd4\u4f20\u7edf\u65b9\u6cd5\u66f4\u597d\u7684\u7ed3\u679c\u3002\u5b9e\u8df5\u8005\u4eec\u5f88\u5feb\u5c31\u5c06\u6df1\u5ea6\u5b66\u4e60\u7eb3\u5165\u5230\u5176\u4ed6\u9886\u57df\uff0c\u4ee5\u89e3\u51b3\u957f\u671f\u5b58\u5728\u7684\u95ee\u9898\u3002\u5728\u8ba1\u7b97\u673a\u89c6\u89c9<\/span>\uff08CV\uff09<\/span>\u4e2d\uff0c\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u88ab\u7528\u4e8e\u56fe\u50cf\u8bc6\u522b<\/span>\uff08Simonyan and Zisserman, 2014; He et al. 2016; Radford et al. 2021\uff09<\/span>\u548c\u56fe\u50cf\u751f\u6210<\/span>\uff08Wang et al. 2021; Jabbar et al. 2021\uff09<\/span>\u3002\u5728\u81ea\u7136\u8bed\u8a00\u5904\u7406<\/span>\uff08NLP\uff09<\/span>\u4e2d\uff0c\u6df1\u5ea6\u8bed\u8a00\u6a21\u578b\u53ef\u4ee5\u751f\u6210\u6587\u672c<\/span>\uff08Radford et al. 2018\uff0c2019; Brown et al. 2020\uff09<\/span>\u5e76\u8fdb\u884c\u673a\u5668\u7ffb\u8bd1<\/span>\uff08Stahlberg\uff0c2020\uff09<\/span>\u3002\u6df1\u5ea6\u5b66\u4e60\u4e5f\u88ab\u7eb3\u5165\u5230\u5f3a\u5316\u5b66\u4e60<\/span>\uff08RL\uff09<\/span>\u4e2d\uff0c\u4ee5\u89e3\u51b3\u57fa\u4e8e\u89c6\u89c9\u7684\u8ba1\u7b97\u673a\u6e38\u620f\uff0c\u5982Doom<\/span>\uff08Ha and Schmidhuber, 2018\uff09<\/span>\u548cAtari<\/span>\uff08Mnih et al. 2015\uff09<\/span>\uff0c\u4ee5\u53ca\u5177\u6709\u5927\u578b\u641c\u7d22\u7a7a\u95f4\u7684\u6e38\u620f\uff0c\u5982\u56f4\u68cb<\/span>\uff08Silver et al. 2016\uff09<\/span>\u548c\u661f\u9645\u4e89\u9738<\/span>\uff08Vinyals et al.  2019\uff09<\/span>\u3002\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u4e5f\u88ab\u90e8\u7f72\u5728\u79fb\u52a8\u5e94\u7528\u4e2d\uff0c\u5982\u8bed\u97f3\u8bc6\u522b<\/span>\uff08Alam et al. 2020\uff09<\/span>\u548c\u8bed\u97f3\u5408\u6210<\/span>\uff08Tan et al. 2021\uff09<\/span>\uff0c\u663e\u793a\u51fa\u5176\u5e7f\u6cdb\u7684\u9002\u7528\u6027\u3002<\/span><\/section>\n

<\/section>\n
\u7136\u800c\uff0c\u6df1\u5ea6\u5b66\u4e60\u5e76\u4e0d\u662f\u6ca1\u6709\u526f\u4f5c\u7528\u7684\u7075\u4e39\u5999\u836f\u3002\u867d\u7136\u6211\u4eec\u89c1\u8bc1\u4e86\u8bb8\u591a\u6210\u529f\u7684\u6848\u4f8b\uff0c\u5e76\u4e14\u8d8a\u6765\u8d8a\u591a\u5730\u91c7\u7528\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\uff0c\u4f46\u968f\u7740\u6211\u4eec\u7684\u6a21\u578b\u548c\u8bad\u7ec3\u7b97\u6cd5\u53d8\u5f97\u66f4\u5927\u3001\u66f4\u590d\u6742\uff0c\u6df1\u5ea6\u5b66\u4e60\u7684\u57fa\u672c\u95ee\u9898\u4e5f\u8d8a\u6765\u8d8a\u660e\u663e\u5730\u66b4\u9732\u51fa\u6765\u3002\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u5728\u67d0\u4e9b\u60c5\u51b5\u4e0b\u5e76\u4e0d\u7a33\u5065\u3002\u4f17\u6240\u5468\u77e5\uff0c\u53ea\u8981\u4fee\u6539\u89c6\u9891\u6e38\u620f\u5c4f\u5e55\u4e0a\u7684\u51e0\u4e2a\u50cf\u7d20<\/span>\uff08\u8fd9\u79cd\u4fee\u6539\u5bf9\u4eba\u7c7b\u6765\u8bf4\u6839\u672c\u65e0\u6cd5\u5bdf\u89c9\uff09<\/span>\uff0c\u7528\u672a\u4fee\u6539\u7684\u5c4f\u5e55\u8bad\u7ec3\u51fa\u6765\u7684\u539f\u672c\u8d85\u8d8a\u4eba\u7c7b\u6027\u80fd\u7684\u6a21\u578b\u5c31\u53ef\u80fd\u5931\u8d25<\/span>\uff08Qu et al. 2020\uff09<\/span>\u3002\u53e6\u5916\uff0c\u5728\u6ca1\u6709\u7279\u6b8a\u5904\u7406\u7684\u60c5\u51b5\u4e0b\u8bad\u7ec3\u7684\u8ba1\u7b97\u673a\u89c6\u89c9\u6a21\u578b\uff0c\u53ef\u80fd\u65e0\u6cd5\u8bc6\u522b\u65cb\u8f6c\u7684\u6216\u7c7b\u4f3c\u53d8\u6362\u7684\u4f8b\u5b50\uff0c\u6362\u53e5\u8bdd\u8bf4\uff0c\u6211\u4eec\u76ee\u524d\u7684\u6a21\u578b\u548c\u8bad\u7ec3\u65b9\u6cd5\u4e0d\u9002\u5408\u63a8\u5e7f\u5230\u65b0\u7684\u4efb\u52a1\u4e2d\u53bb\u3002\u6700\u540e\uff0c\u5927\u591a\u6570\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u4e0d\u9002\u5e94\u5b66\u4e60\u4efb\u52a1\u7684\u53d8\u5316\u3002\u6a21\u578b\u5bf9\u8f93\u5165\u8fdb\u884c\u5047\u8bbe\uff0c\u5e76\u671f\u671b\u73af\u5883\u7684\u521a\u6027\u914d\u7f6e\u548c\u9759\u6b62\u6027\uff0c\u5373\u7edf\u8ba1\u5b66\u5bb6\u8ba4\u4e3a\u7684\u6570\u636e\u751f\u6210\u8fc7\u7a0b\u3002\u4f8b\u5982\uff0c\u5728\u4e00\u4e2a\u786e\u5b9a\u7684\u987a\u5e8f\u4e2d\uff0c\u6a21\u578b\u53ef\u80fd\u671f\u671b\u6709\u56fa\u5b9a\u6570\u91cf\u7684\u8f93\u5165\u3002\u6211\u4eec\u4e0d\u80fd\u671f\u671b\u4e3b\u4f53<\/span>\uff08agent\uff09<\/span>\u6709\u80fd\u529b\u8d85\u8d8a\u5b83\u4eec\u5728\u8bad\u7ec3\u671f\u95f4\u5b66\u5230\u7684\u6280\u80fd\uff0c\u4f46\u662f\u4e00\u65e6\u8fd9\u4e9b\u521a\u6027\u914d\u7f6e\u88ab\u8fdd\u53cd\uff0c\u6a21\u578b\u5c31\u4e0d\u4f1a\u6709\u597d\u7684\u8868\u73b0\uff0c\u9664\u975e\u6211\u4eec\u91cd\u65b0\u8bad\u7ec3\u5b83\u4eec\u6216\u8005\u624b\u52a8\u5904\u7406\u8f93\u5165\uff0c\u4f7f\u4e4b\u4e0e\u6a21\u578b\u6700\u521d\u7684\u8bad\u7ec3\u914d\u7f6e\u4e00\u81f4\u3002<\/span><\/section>\n

<\/section>\n
\u6b64\u5916\uff0c\u968f\u7740\u6240\u6709\u8fd9\u4e9b\u8fdb\u5c55\uff0c\u6df1\u5ea6\u5b66\u4e60\u4e2d\u4ee4\u4eba\u5370\u8c61\u6df1\u523b\u7684\u58ee\u4e3e\u662f\u88ab\u5e94\u7528\u5230\u590d\u6742\u7684\u5de5\u7a0b\u9879\u76ee\u4e2d\u3002\u4f8b\u5982\uff0c\u8457\u540d\u7684AlexNet\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u67b6\u6784<\/span>\uff08Krizhevsky et al. 2012\uff09\uff08\u89c1\u56fe1\uff09<\/span>\uff0c\u57282012\u5e74\u8d62\u5f97ImageNet\u7ade\u8d5b\u4e4b\u540e\uff0c\u4f7f\u6df1\u5ea6\u5b66\u4e60\u6210\u4e3a\u8ba1\u7b97\u673a\u89c6\u89c9\u754c\u7684\u7126\u70b9\uff0c\u5b83\u63d0\u51fa\u4e86\u4e00\u4e2a\u7cbe\u5fc3\u8bbe\u8ba1\u7684\u7f51\u7edc\u67b6\u6784\u548c\u4e00\u4e2a\u7cbe\u5fc3\u6821\u51c6\u7684\u8bad\u7ec3\u7a0b\u5e8f\u3002\u73b0\u4ee3\u795e\u7ecf\u7f51\u7edc\u901a\u5e38\u66f4\u52a0\u590d\u6742\uff0c\u9700\u8981\u4e00\u4e2a\u6a2a\u8de8\u7f51\u7edc\u67b6\u6784\u548c\u7cbe\u7ec6\u8bad\u7ec3\u65b9\u6848\u7684\u7ba1\u9053\u3002\u50cf\u8bb8\u591a\u5de5\u7a0b\u9879\u76ee\u4e00\u6837\uff0c\u5728\u4ea7\u751f\u6bcf\u4e00\u4e2a\u7ed3\u679c\u65f6\uff0c\u90fd\u8981\u4ed8\u51fa\u5f88\u591a\u52b3\u52a8\u548c\u5fae\u8c03\u3002<\/span><\/section>\n

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

\u56fe1. AlexNet\uff08Krizhevsky et al. 2012\uff09\u7684\u795e\u7ecf\u7f51\u7edc\u7ed3\u6784\uff0c\u5b83\u662f2012\u5e74ImageNet\u7ade\u8d5b\u7684\u51a0\u519b\u3002<\/span><\/p>\n


<\/section>\n
\u6211\u4eec\u8ba4\u4e3a\uff0c\u6df1\u5ea6\u5b66\u4e60\u7684\u8bb8\u591a\u9650\u5236\u548c\u526f\u4f5c\u7528\u6e90\u4e8e\u8fd9\u6837\u4e00\u4e2a\u4e8b\u5b9e\uff0c\u5373\u76ee\u524d\u6df1\u5ea6\u5b66\u4e60\u7684\u5b9e\u8df5\u4e0e\u5efa\u7b51\u5de5\u7a0b\u7c7b\u4f3c\u3002\u6211\u4eec\u5efa\u9020\u73b0\u4ee3\u795e\u7ecf\u7f51\u7edc\u7cfb\u7edf\u7684\u65b9\u5f0f\u7c7b\u4f3c\u4e8e\u6211\u4eec\u5efa\u9020\u6865\u6881\u548c\u5efa\u7b51\u7684\u65b9\u5f0f\uff0c\u800c\u8fd9\u4e9b\u8bbe\u8ba1\u662f\u4e0d\u5177\u6709\u9002\u5e94\u6027\u7684\u3002\u5f15\u7528\u300a\u63a7\u5236\u8bba\u5927\u8111\u300b<\/span>\uff08The Cybernetic Brain<\/em>\uff0c<\/em>Pickering, 2010\uff09<\/span>\u4e00\u4e66\u4f5c\u8005Pickering\u7684\u8bdd\u8bf4\uff1a\u201c\u6211\u6240\u60f3\u5230\u7684\u5927\u591a\u6570\u5de5\u7a0b\u5b9e\u4f8b\u90fd\u662f\u4e0d\u5177\u6709\u9002\u5e94\u6027\u7684\u3002\u6865\u6881\u548c\u5efa\u7b51\u3001\u8f66\u5e8a\u548c\u53d1\u52a8\u673a\u3001\u6c7d\u8f66\u3001\u7535\u89c6\u4ee5\u53ca\u7535\u8111\uff0c\u90fd\u88ab\u8bbe\u8ba1\u6210\u5bf9\u73af\u5883\u6f20\u4e0d\u5173\u5fc3\uff0c\u627f\u53d7\u73af\u5883\u7684\u6ce2\u52a8\uff0c\u800c\u4e0d\u662f\u9002\u5e94\u73af\u5883\u3002 \u6700\u597d\u7684\u6865\u662f\u4e00\u5ea7\u4e0d\u7ba1\u5929\u6c14\u5982\u4f55\u90fd\u5c79\u7acb\u5728\u90a3\u91cc\u7684\u6865\u3002\u201d<\/span><\/section>\n

<\/section>\n
\u96c6\u4f53\u667a\u80fd<\/span>\uff08Collective Intelligence\uff09<\/span>\u5728\u81ea\u7136\u7cfb\u7edf\u4e2d\u53d1\u6325\u4e86\u5f88\u5927\u7684\u4f5c\u7528\uff0c\u6211\u4eec\u770b\u5230\u7531\u4e8e\u81ea\u7ec4\u7ec7\u800c\u6d8c\u73b0\u7684\u9002\u5e94\u6027\u8bbe\u8ba1\uff0c\u8fd9\u79cd\u8bbe\u8ba1\u5bf9\u5468\u56f4\u4e16\u754c\u7684\u53d8\u5316\u975e\u5e38\u654f\u611f\uff0c\u53cd\u5e94\u8fc5\u901f\u3002\u81ea\u7136\u7cfb\u7edf\u9002\u5e94\u5e76\u6210\u4e3a\u73af\u5883\u7684\u4e00\u90e8\u5206<\/span>\uff08\u89c1\u56fe2\u7684\u6bd4\u55bb\uff09<\/span>\u3002<\/span><\/section>\n

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

\u56fe2. \u5de6\u56fe\uff1a\u4f4d\u4e8e\u963f\u5c14\u574e\u5854\u62c9\u7684\u56fe\u62c9\u771f\u6865\uff0c\u7531\u7f57\u9a6c\u4eba\u4e8e\u516c\u5143106\u5e74\u5efa\u9020\uff08\u7ef4\u57fa\u767e\u79d1\uff0c2022\uff09\u3002\u53f3\u56fe\uff1a\u884c\u519b\u8681\u7ec4\u6210\u7684\u6865\uff08Jenal, 2011\uff09\u3002<\/span><\/p>\n


<\/section>\n
\u6b63\u5982\u884c\u519b\u8681\u96c6\u4f53\u5f62\u6210\u4e00\u5ea7\u9002\u5e94\u73af\u5883\u7684\u6865\u6881\u7684\u4f8b\u5b50\u4e00\u6837\uff0c\u81ea\u7136\u754c\u4e2d\u5e38\u89c1\u7684\u96c6\u4f53\u884c\u4e3a\u5f80\u5f80\u4f1a\u4ea7\u751f\u9002\u5e94\u6027\u5f3a\u3001\u7a33\u5065\u3001\u5bf9\u73af\u5883\u914d\u7f6e\u7684\u5047\u8bbe\u4e0d\u90a3\u4e48\u4e25\u683c\u7684\u7cfb\u7edf\u3002\u96c6\u4f53\u667a\u80fd\u4f5c\u4e3a\u4e00\u4e2a\u9886\u57df\uff0c\u7814\u7a76\u7684\u662f\u4ece\u8bb8\u591a\u4e2a\u4f53\u7684\u4e92\u52a8<\/span>\uff08\u5982\u5408\u4f5c\u3001\u96c6\u4f53\u884c\u52a8\u548c\u7ade\u4e89\uff09<\/span>\u4e2d\u6d8c\u73b0\u7684\u7fa4\u4f53\u667a\u80fd<\/span>\uff08Group Intelligence\uff09<\/span>\u3002\u5728\u8fd9\u4e2a\u9886\u57df\u4e2d\uff0c\u8bf8\u5982\u81ea\u7ec4\u7ec7\u3001\u6d8c\u73b0\u884c\u4e3a\u3001\u7fa4\u4f18\u5316\u548c\u5143\u80de\u81ea\u52a8\u673a\u7b49\u601d\u60f3\u88ab\u5f00\u53d1\u51fa\u6765\uff0c\u4ee5\u6a21\u62df\u548c\u89e3\u91ca\u590d\u6742\u7cfb\u7edf\u3002\u56e0\u6b64\uff0c\u770b\u5230\u8fd9\u4e9b\u601d\u60f3\u88ab\u7eb3\u5165\u8f83\u65b0\u7684\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\u4e2d\u662f\u5f88\u81ea\u7136\u7684\u3002<\/span><\/section>\n

<\/section>\n
\u6211\u4eec\u4e0d\u8ba4\u4e3a\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u5fc5\u987b\u6309\u7167\u5efa\u9020\u6865\u6881\u7684\u601d\u8def\u6765\u6784\u5efa\u3002\u6b63\u5982\u6211\u4eec\u5c06\u5728\u540e\u9762\u8ba8\u8bba\u7684\u90a3\u6837\u3002\u6df1\u5ea6\u5b66\u4e60\u9886\u57df\u4e4b\u6240\u4ee5\u8d70\u4e0a\u8fd9\u6761\u9053\u8def\uff0c\u53ef\u80fd\u53ea\u662f\u5386\u53f2\u4e0a\u4e00\u4e2a\u5076\u7136\u7684\u7ed3\u679c\u3002\u4e8b\u5b9e\u4e0a\uff0c\u6700\u8fd1\u6709\u51e0\u9879\u5de5\u4f5c\u901a\u8fc7\u5c06\u6df1\u5ea6\u5b66\u4e60\u4e0e\u96c6\u4f53\u667a\u80fd\u7684\u7406\u5ff5\u76f8\u7ed3\u5408\uff0c\u4ee5\u89e3\u51b3\u6df1\u5ea6\u5b66\u4e60\u4e2d\u5b58\u5728\u7684\u5c40\u9650\u6027\u3002\u4ece\u5c06\u5143\u80de\u81ea\u52a8\u673a\u5e94\u7528\u4e8e\u57fa\u4e8e\u795e\u7ecf\u7f51\u7edc\u7684\u56fe\u50cf\u5904\u7406\u6a21\u578b<\/span>\uff08Mordvintsev et al. 2020\uff1bRandazzo et al. 2020\uff09<\/span>\uff0c\u5230\u91cd\u65b0\u5b9a\u4e49\u5982\u4f55\u4f7f\u7528\u81ea\u7ec4\u7ec7\u4e3b\u4f53\u6765\u5904\u7406\u5f3a\u5316\u5b66\u4e60\u4e2d\u7684\u95ee\u9898<\/span>\uff08Pathak et al. 2019; Huang et al. 2020; Tang and Ha, 2021\uff09<\/span>\u3002\u968f\u7740\u6211\u4eec\u89c1\u8bc1\u4e86\u5e76\u884c\u8ba1\u7b97\u786c\u4ef6<\/span>\uff08\u9002\u5408\u6a21\u62df\u96c6\u4f53\u884c\u4e3a\uff0c\u89c1\u56fe3\u7684\u4f8b\u5b50\uff09<\/span>\u7684\u4e0d\u65ad\u6280\u672f\u8fdb\u6b65\uff0c\u6211\u4eec\u53ef\u4ee5\u671f\u5f85\u66f4\u591a\u7684\u4f5c\u54c1\u5c06\u96c6\u4f53\u667a\u80fd\u7eb3\u5165\u5230\u4f20\u7edf\u6df1\u5ea6\u5b66\u4e60\u5904\u7406\u7684\u95ee\u9898\u4e2d\u53bb\u3002<\/span><\/section>\n

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

\u56fe3. GPU\u786c\u4ef6\u7684\u6700\u65b0\u8fdb\u5c55\u80fd\u591f\u5bf9\u6210\u5343\u4e0a\u4e07\u7684\u673a\u5668\u4eba\u6a21\u578b\u8fdb\u884c\u903c\u771f\u76843D\u6a21\u62df\uff08Heiden et al. 2021\uff09\uff0c\u5982\u672c\u56fe\u6240\u793a\uff0c\u6765\u81ea\uff08Rudin et al. 2021\uff09\u3002\u8fd9\u6837\u7684\u8fdb\u5c55\u4e3a\u5927\u89c4\u6a21\u7684\u4e09\u7ef4\u6a21\u62df\u4eba\u5de5\u4e3b\u4f53\u6253\u5f00\u4e86\u5927\u95e8\uff0c\u8fd9\u4e9b\u4e3b\u4f53\u53ef\u4ee5\u76f8\u4e92\u4ea4\u6d41\uff0c\u5e76\u96c6\u4f53\u53d1\u5c55\u667a\u80fd\u884c\u4e3a\u3002<\/span><\/p>\n


<\/section>\n
\u8fd9\u7bc7\u7efc\u8ff0\u7684\u76ee\u7684\u662f\u5bf9\u96c6\u4f53\u667a\u80fd\u9886\u57df\u7684\u6838\u5fc3\u601d\u60f3\u3001\u5de5\u5177\u548c\u89c1\u89e3\uff0c\u5c24\u5176\u662f\u81ea\u7ec4\u7ec7\u3001\u6d8c\u73b0\u548c\u7fa4\u4f53\u6a21\u578b\u5982\u4f55\u5f71\u54cd\u6df1\u5ea6\u5b66\u4e60\u7684\u4e0d\u540c\u9886\u57df\uff0c\u4ece\u56fe\u50cf\u5904\u7406\u3001\u5f3a\u5316\u5b66\u4e60\u5230\u5143\u5b66\u4e60\uff0c\u63d0\u4f9b\u4e00\u4e2a\u9ad8\u5c42\u6b21\u7684\u8c03\u67e5\u3002\u6211\u4eec\u5e0c\u671b\u8fd9\u7bc7\u7efc\u8ff0\u80fd\u5bf9\u672a\u6765\u7684\u6df1\u5ea6\u5b66\u4e60\u96c6\u4f53\u667a\u80fd\u7684\u534f\u540c\u4f5c\u7528\u63d0\u4f9b\u4e00\u4e9b\u89c1\u89e3\uff0c\u6211\u4eec\u76f8\u4fe1\u8fd9\u5c06\u5728\u8fd9\u4e24\u4e2a\u9886\u57df\u5e26\u6765\u6709\u610f\u4e49\u7684\u7a81\u7834\u3002<\/span><\/section>\n

<\/section>\n

<\/section>\n

\n
\n
\n
\n
\n

<\/section>\n

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

2. \u80cc\u666f\uff1a\u96c6\u4f53\u667a\u80fd<\/span><\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n
 <\/span><\/section>\n
\u96c6\u4f53\u667a\u80fd\u662f\u4e00\u4e2a\u5728\u793e\u4f1a\u5b66\u3001\u5546\u4e1a\u3001\u901a\u4fe1\u548c\u8ba1\u7b97\u673a\u79d1\u5b66\u7b49\u9886\u57df\u5e7f\u6cdb\u4f7f\u7528\u7684\u672f\u8bed\u3002\u96c6\u4f53\u667a\u80fd\u7684\u5b9a\u4e49\u53ef\u4ee5\u6982\u62ec\u4e3a\u4e00\u79cd\u4e0d\u65ad\u589e\u5f3a\u548c\u534f\u8c03\u7684\u5206\u5e03\u5f0f\u667a\u80fd\uff0c\u5176\u76ee\u6807\u662f\u901a\u8fc7\u76f8\u4e92\u8ba4\u53ef\u548c\u4e92\u52a8\uff0c\u53d6\u5f97\u6bd4\u7fa4\u4f53\u4e2d\u4efb\u4f55\u4e2a\u4f53\u66f4\u597d\u7684\u7ed3\u679c<\/span>\uff08L\u00e9vy, 1997; Leimeister, 2010\uff09<\/span>\u3002\u96c6\u4f53\u667a\u80fd\u5e26\u6765\u7684\u66f4\u597d\u7ed3\u679c\u5f52\u529f\u4e8e\u4e09\u4e2a\u56e0\u7d20\uff1a\u591a\u6837\u6027\u3001\u72ec\u7acb\u6027\u548c\u53bb\u4e2d\u5fc3\u5316<\/span>\uff08Surowiecki, 2005; Tapscott and Williams, 2008\uff09<\/span>\u3002<\/span><\/section>\n

<\/section>\n
\u4e3a\u4e86\u6211\u4eec\u7684\u76ee\u7684\uff0c\u6211\u4eec\u8ba4\u4e3a\u96c6\u4f53\u667a\u80fd\u4f5c\u4e3a\u4e00\u4e2a\u9886\u57df\uff0c\u662f\u5bf9\u8bb8\u591a\u4e2a\u4f53\u4e4b\u95f4\u7684\u4e92\u52a8<\/span>\uff08\u53ef\u4ee5\u662f\u5408\u4f5c\u6027\u7684\uff0c\u4e5f\u53ef\u4ee5\u662f\u7ade\u4e89\u6027\u7684\uff09<\/span>\u4e2d\u6d8c\u73b0\u7684\u7fa4\u4f53\u667a\u80fd\u7684\u7814\u7a76\u3002\u8fd9\u79cd\u7fa4\u4f53\u667a\u80fd\u662f\u6d8c\u73b0\u7684\u4ea7\u7269\uff0c\u5f53\u89c2\u5bdf\u5230\u7fa4\u4f53\u5177\u6709\u7ec4\u6210\u7fa4\u4f53\u7684\u4e2a\u4f53\u6240\u4e0d\u5177\u5907\u7684\u7279\u6027\u65f6\uff0c\u5b83\u624d\u4f1a\u51fa\u73b0\uff0c\u5e76\u4e14\u53ea\u6709\u5f53\u7fa4\u4f53\u4e2d\u7684\u4e2a\u4f53\u5728\u66f4\u5e7f\u6cdb\u7684\u6574\u4f53\u4e2d\u4e92\u52a8\u65f6\u624d\u4f1a\u4ea7\u751f\u6d8c\u73b0\u73b0\u8c61\u3002<\/span><\/section>\n

<\/section>\n
\u8fd9\u79cd\u7cfb\u7edf\u7684\u4f8b\u5b50\u5728\u81ea\u7136\u754c\u4e2d\u6bd4\u6bd4\u7686\u662f\uff0c\u4ece\u4e2a\u4f53\u4e4b\u95f4\u7b80\u5355\u7684\u5c40\u90e8\u4e92\u52a8\/\u534f\u4f5c\u4e2d\u4ea7\u751f\u4e86\u590d\u6742\u7684\u5168\u5c40\u884c\u4e3a\uff0c\u4ee5\u5b9e\u73b0\u5171\u540c\u7684\u76ee\u6807<\/span>\uff08Deneubourg and  Goss\uff0c1989; Toner, 2005\uff1bSumpter\uff0c2010; Lajad et al. 2021\uff09<\/span>\u3002\u5728\u8fd9\u7bc7\u8bc4\u8bba\u4e2d\uff0c\u6211\u4eec\u4ec5\u9650\u4e8e\u5173\u6ce8\u96c6\u4f53\u667a\u80fd\u7684\u6a21\u62df\uff0c\u800c\u4e0d\u662f\u5206\u6790\u5728\u81ea\u7136\u548c\u793e\u4f1a\u4e2d\u89c2\u5bdf\u5230\u7684\u96c6\u4f53\u667a\u80fd\u3002\u65e9\u671f\u6570\u5341\u5e74\u7684\u5de5\u4f5c\u4e5f\u63a2\u7d22\u4e86\u96c6\u4f53\u884c\u4e3a\u7684\u6a21\u62df\uff0c\u5e76\u4ece\u8fd9\u79cd\u6a21\u62df\u4e2d\u6536\u96c6\u89c1\u89e3\u3002Mataric<\/span>\uff081993\uff09<\/span>\u4f7f\u7528\u7269\u7406\u79fb\u52a8\u673a\u5668\u4eba\u6765\u7814\u7a76\u5bfc\u81f4\u7fa4\u4f53\u884c\u4e3a\u7684\u793e\u4f1a\u4e92\u52a8\u3002\u4ed6\u4eec\u63d0\u51fa\u4e86\u4e00\u5957\u57fa\u672c\u7684\u76f8\u4e92\u4f5c\u7528\u673a\u5236<\/span>\uff08\u4f8b\u5982\uff0c\u907f\u514d\u78b0\u649e\u3001\u8ddf\u968f\u548c\u805a\u96c6\uff09<\/span>\uff0c\u5e0c\u671b\u8fd9\u4e9b\u57fa\u672c\u8981\u7d20\u80fd\u591f\u4f7f\u4e00\u7fa4\u81ea\u7ec4\u7ec7\u4e3b\u4f53\u5b8c\u6210\u4e00\u4e2a\u5171\u540c\u7684\u76ee\u6807\u6216\u76f8\u4e92\u5b66\u4e60\u3002\u53d7\u5728\u771f\u5b9e\u7684\u8682\u8681\u7fa4\u4e2d\u89c2\u5bdf\u5230\u7684\u7fa4\u4f53\u884c\u4e3a\u7684\u542f\u53d1\uff0cDorigo\u7b49\u4eba<\/span>\uff082000\uff09<\/span>\u63d0\u51fa\u4e86stigmergy<\/span>\uff08\u793e\u4f1a\u6027\u6606\u866b\u4f7f\u7528\u7684\u4e00\u79cd\u7279\u6b8a\u5f62\u5f0f\u7684\u95f4\u63a5\u901a\u4fe1\uff09<\/span>\u4f5c\u4e3a\u5206\u5e03\u5f0f\u901a\u4fe1\u8303\u5f0f\uff0c\u5e76\u5c55\u793a\u4e86\u5b83\u662f\u5982\u4f55\u542f\u53d1\u89e3\u51b3\u5206\u5e03\u5f0f\u4f18\u5316\u548c\u63a7\u5236\u95ee\u9898\u7684\u65b0\u578b\u7b97\u6cd5\u7684\u3002\u6b64\u5916\uff0cSchweitzer\u548cFarmer<\/span>\uff082003\uff09<\/span>\u5728\u8bb8\u591a\u4e0d\u540c\u7684\u80cc\u666f\u4e0b\u5e94\u7528\u4e86\u5e03\u6717\u4e3b\u4f53\u6a21\u578b\u3002\u7ed3\u5408\u591a\u4e3b\u4f53\u7cfb\u7edf\u548c\u7edf\u8ba1\u65b9\u6cd5\uff0c\u4f5c\u8005\u63d0\u51fa\u4e86\u4e00\u4e2a\u7406\u89e3\u590d\u6742\u7cfb\u7edf\u7684\u8fde\u8d2f\u6846\u67b6\u7684\u613f\u666f\u3002<\/span><\/section>\n

<\/section>\n
\u867d\u7136\u8fd9\u4e9b\u65e9\u671f\u7684\u5de5\u4f5c\u63a8\u52a8\u4e86\u9002\u7528\u4e8e\u4f18\u5316\u95ee\u9898\u7b97\u6cd5\u7684\u53d1\u73b0<\/span>\uff08\u5982\u89e3\u51b3\u65c5\u884c\u5546\u95ee\u9898\u7684\u8681\u7fa4\u4f18\u5316\u7b97\u6cd5\uff09<\/span>\uff0c\u4f46\u8bb8\u591a\u8fd9\u4e9b\u5de5\u4f5c\u7684\u76ee\u7684\u662f\u4f7f\u7528\u8fd9\u4e9b\u6a21\u578b\u6765\u7406\u89e3\u96c6\u4f53\u667a\u80fd\u7684\u6d8c\u73b0\u73b0\u8c61\u3002\u8fd9\u6307\u51fa\u4e86\u96c6\u4f53\u667a\u80fd\u548c\u4eba\u5de5\u667a\u80fd\u9886\u57df\u7684\u76ee\u6807\u4e4b\u95f4\u7684\u4e00\u4e2a\u6839\u672c\u533a\u522b\u3002\u5728\u96c6\u4f53\u667a\u80fd\u9886\u57df\uff0c\u76ee\u6807\u662f\u5efa\u7acb\u590d\u6742\u7cfb\u7edf\u7684\u6a21\u578b\uff0c\u4ee5\u5e2e\u52a9\u6211\u4eec\u89e3\u91ca\u548c\u7406\u89e3\u6d8c\u73b0\u73b0\u8c61\uff0c\u8fd9\u53ef\u80fd\u88ab\u5e94\u7528\u4e8e\u7406\u89e3\u81ea\u7136\u548c\u793e\u4f1a\u4e2d\u7684\u771f\u5b9e\u7cfb\u7edf\u3002\u800c\u4eba\u5de5\u667a\u80fd\uff08\u5c24\u5176\u662f\u673a\u5668\u5b66\u4e60\u9886\u57df\uff09\u5219\u5173\u6ce8\u4f18\u5316\u3001\u5206\u7c7b\u3001\u9884\u6d4b\u548c\u89e3\u51b3\u4e00\u4e2a\u5177\u4f53\u95ee\u9898\u3002<\/strong><\/span><\/section>\n

<\/section>\n
\u5728\u6211\u4eec\u63d0\u5230\u7684\u65e9\u671f\u4f5c\u54c1\u4e2d\uff0c\u5e76\u6ca1\u6709\u5145\u5206\u5229\u7528\u6df1\u5ea6\u5b66\u4e60\u7684\u5efa\u6a21\u80fd\u529b\u6216\u786c\u4ef6\u7684\u53d1\u5c55\u3002\u5c3d\u7ba1\u5982\u6b64\uff0c\u8fd9\u4e9b\u5de5\u4f5c\u8fd8\u662f\u4e0d\u65ad\u5c55\u793a\u4e86\u96c6\u4f53\u667a\u80fd\u7684\u60ca\u4eba\u6548\u679c\uff0c\u5373\u7cfb\u7edf\u662f\u81ea\u7ec4\u7ec7\u7684\uff0c\u80fd\u591f\u901a\u8fc7\u7fa4\u4f53\u667a\u80fd\u8fdb\u884c\u4f18\u5316\uff0c\u5e76\u5448\u73b0\u51fa\u6d8c\u73b0\u884c\u4e3a\u3002\u8fd9\u4e9b\u7814\u7a76\u4eba\u5458\u8ba4\u4e3a\uff0c\u96c6\u4f53\u667a\u80fd\u7684\u6982\u5ff5\u662f\u6709\u5f88\u597d\u7684\u524d\u666f\u7684\uff0c\u53ef\u4ee5\u5e94\u7528\u4e8e\u6df1\u5ea6\u5b66\u4e60\uff0c\u4ee5\u4ea7\u751f\u7a33\u5065\u3001\u9002\u5e94\u6027\u5f3a\u3001\u5bf9\u73af\u5883\u914d\u7f6e\u7684\u5047\u8bbe\u4e0d\u90a3\u4e48\u4e25\u683c\u7684\u89e3\u51b3\u65b9\u6848\uff0c\u8fd9\u4e5f\u662f\u672c\u7efc\u8ff0\u7684\u91cd\u70b9\u3002<\/span><\/section>\n

<\/section>\n

<\/section>\n

\n
\n
\n
\n
\n

<\/section>\n

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

3. \u5386\u53f2\u80cc\u666f\uff1a\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc<\/span><\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n
 <\/span><\/section>\n
\u590d\u6742\u7cfb\u7edf\u7684\u601d\u60f3\uff0c\u5982\u81ea\u7ec4\u7ec7\uff0c\u88ab\u7528\u6765\u6a21\u62df\u548c\u7406\u89e3\u6d8c\u73b0\u548c\u96c6\u4f53\u884c\u4e3a\uff0c\u4e0e\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\u7684\u53d1\u5c55\u6709\u7740\u957f\u671f\u800c\u6709\u8da3\u7684\u5386\u53f2\u5173\u7cfb\u3002\u867d\u7136\u4eff\u751f\u5b66\u6d3e\u548c\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\u662f\u572820\u4e16\u7eaa50\u5e74\u4ee3\u968f\u7740\u4eba\u5de5\u667a\u80fd\u7684\u8bde\u751f\u800c\u51fa\u73b0\u7684\uff0c\u4f46\u6211\u4eec\u7684\u6545\u4e8b\u5f00\u59cb\u4e8e20\u4e16\u7eaa70\u5e74\u4ee3\uff0c\u5f53\u65f6\u7684\u5148\u9a71Leon Chua\u9886\u5bfc\u7684\u4e00\u7fa4\u7535\u5b50\u5de5\u7a0b\u5e08\u5f00\u59cb\u53d1\u5c55\u975e\u7ebf\u6027\u7535\u8def\u7406\u8bba\uff0c\u5e76\u5c06\u5176\u5e94\u7528\u4e8e\u8ba1\u7b97\u3002\u4ed6\u56e0\u572820\u4e16\u7eaa70\u5e74\u4ee3\u6784\u601d\u4e86\u8bb0\u5fc6\u4f53<\/span>\uff08\u4e00\u4e2a\u6700\u8fd1\u624d\u5b9e\u73b0\u7684\u8bbe\u5907\uff09<\/span>\uff0c\u5e76\u8bbe\u8ba1\u4e86\u8521\u6c0f\u7535\u8def\uff0c\u8fd9\u662f\u7b2c\u4e00\u6279\u56e0\u8868\u73b0\u51fa\u6df7\u6c8c\u884c\u4e3a\u800c\u95fb\u540d\u7535\u8def\u3002\u572820\u4e16\u7eaa80\u5e74\u4ee3\uff0c\u4ed6\u7684\u5c0f\u7ec4\u5f00\u53d1\u4e86\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc<\/span>\uff08Cellular Neural Networks\uff09<\/span>\uff0c\u8fd9\u662f\u4e00\u79cd\u7c7b\u4f3c\u4e8e\u5143\u80de\u81ea\u52a8\u673a<\/span>\uff08Cellular Automata, CA\uff09<\/span>\u7684\u8ba1\u7b97\u7cfb\u7edf\uff0c\u4f46\u4f7f\u7528\u795e\u7ecf\u7f51\u7edc\u6765\u4ee3\u66ff\u5143\u80de\u81ea\u52a8\u673a\u7cfb\u7edf\u4e2d\u5178\u578b\u7684\u7b97\u6cd5\u5355\u5143\uff0c\u5982\u5eb7\u5a01\u7684\u751f\u547d\u6e38\u620f<\/span>\uff08Conway et al. 1970\uff09<\/span>\u6216\u521d\u7b49\u7684\u5143\u80de\u81ea\u52a8\u673a\u89c4\u5219<\/span>\uff08Wolfram, 2002\uff09<\/span>\u3002<\/span><\/section>\n

<\/section>\n
\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc<\/span>\uff08CeNNs\uff09\uff08Chua and Yang, 1988a, 1988b\uff09<\/span>\u662f\u4e00\u79cd\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\uff0c\u6bcf\u4e2a\u795e\u7ecf\u5143\u6216\u7ec6\u80de\u53ea\u80fd\u4e0e\u5b83\u4eec\u7684\u8fd1\u90bb\u4e92\u52a8\u3002\u5728\u6700\u57fa\u672c\u7684\u8bbe\u7f6e\u4e2d\uff0c\u6bcf\u4e2a\u7ec6\u80de\u7684\u72b6\u6001\u90fd\u662f\u901a\u8fc7\u5176\u90bb\u5c45\u548c\u81ea\u8eab\u72b6\u6001\u7684\u975e\u7ebf\u6027\u51fd\u6570\u6765\u6301\u7eed\u66f4\u65b0\u7684\u3002\u4e0e\u4f9d\u9760\u6570\u5b57\u3001\u79bb\u6563\u65f6\u95f4\u8ba1\u7b97\u7684\u73b0\u4ee3\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\u4e0d\u540c\uff0c\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u662f\u8fde\u7eed\u7684\u65f6\u95f4\u7cfb\u7edf\uff0c\u901a\u5e38\u7528\u975e\u7ebf\u6027\u6a21\u62df\u7535\u5b50\u5143\u4ef6\u5b9e\u73b0<\/span>\uff08\u89c1\u56fe4\uff0c\u5de6\uff09<\/span>\uff0c\u4f7f\u5176\u901f\u5ea6\u975e\u5e38\u5feb\u3002\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u7684\u52a8\u529b\u5b66\u4f9d\u8d56\u4e8e\u72ec\u7acb\u7684\u5c40\u90e8\u4fe1\u606f\u5904\u7406\u548c\u5904\u7406\u5355\u5143\u4e4b\u95f4\u7684\u4e92\u52a8\uff0c\u4e0e\u5143\u80de\u81ea\u52a8\u673a\u4e00\u6837\uff0c\u5b83\u4eec\u4e5f\u8868\u73b0\u51fa\u6d8c\u73b0\u884c\u4e3a\uff0c\u5e76\u53ef\u4ee5\u88ab\u505a\u6210\u901a\u7528\u56fe\u7075\u673a\u3002\u7136\u800c\uff0c\u5b83\u4eec\u6bd4\u79bb\u6563\u7684\u5143\u80de\u81ea\u52a8\u673a\u548c\u6570\u5b57\u8ba1\u7b97\u673a\u8981\u666e\u904d\u5f97\u591a\u3002\u7531\u4e8e\u8fde\u7eed\u7684\u72b6\u6001\u7a7a\u95f4\uff0c\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u8868\u73b0\u51fa\u524d\u6240\u672a\u6709\u7684\u6d8c\u73b0\u884c\u4e3a\u3002<\/span>(GoraS et al. 1995)<\/span><\/section>\n

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

\u56fe4. \u5de6\u56fe\uff1a\u4e8c\u7ef4\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7684\u5178\u578b\u914d\u7f6e\uff08Liu et al. 2020\uff09\u3002\u53f3\u56fe\uff1a\u6df1\u5ea6\u5b66\u4e60\u548c\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u8fd9\u4e24\u4e2a\u8bcd\u5728\u4e00\u6bb5\u65f6\u95f4\u5185\u7684\u8c37\u6b4c\u8d8b\u52bf\u3002<\/span><\/p>\n


<\/section>\n
\u4ece20\u4e16\u7eaa90\u5e74\u4ee3\u523021\u4e16\u7eaa\u4e2d\u671f\uff0c\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u6210\u4e3a\u4eba\u5de5\u667a\u80fd\u7814\u7a76\u7684\u4e00\u4e2a\u5b8c\u6574\u5b50\u9886\u57df\u3002\u7531\u4e8e\u5176\u5f3a\u5927\u800c\u9ad8\u6548\u7684\u5206\u5e03\u5f0f\u8ba1\u7b97\uff0c\u5b83\u5728\u56fe\u50cf\u5904\u7406\u3001\u7eb9\u7406\u5206\u6790\u4e2d\u627e\u5230\u4e86\u5e94\u7528\uff0c\u5176\u56fa\u6709\u7684\u6a21\u62df\u8ba1\u7b97\u88ab\u5e94\u7528\u4e8e\u6c42\u89e3\u504f\u5fae\u5206\u65b9\u7a0b\uff0c\u751a\u81f3\u4e3a\u751f\u7269\u7cfb\u7edf\u548c\u5668\u5b98\u5efa\u6a21<\/span>\uff08Chua and Roska, 2002\uff09<\/span>\u3002\u6709\u6570\u4ee5\u5343\u8ba1\u7684\u540c\u884c\u8bc4\u8bae\u7684\u8bba\u6587\u3001\u6559\u79d1\u4e66\u548c\u4e00\u4e2aIEEE\u4f1a\u8bae\u4e13\u95e8\u8ba8\u8bba\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\uff0c\u6709\u8bb8\u591a\u5efa\u8bae\u6269\u5927\u5b83\u4eec\u7684\u89c4\u6a21\uff0c\u5806\u53e0\u5b83\u4eec\uff0c\u5c06\u5b83\u4eec\u4e0e\u6570\u5b57\u7535\u8def\u76f8\u7ed3\u5408\uff0c\u5e76\u7814\u7a76\u4e0d\u540c\u7684\u65b9\u6cd5\u6765\u8bad\u7ec3\u5b83\u4eec<\/span>\uff08\u5c31\u50cf\u6211\u4eec\u76ee\u524d\u5728\u6df1\u5ea6\u5b66\u4e60\u4e2d\u770b\u5230\u7684\u90a3\u6837\uff09<\/span>\u3002\u81f3\u5c11\u6709\u4e24\u5bb6\u786c\u4ef6\u521b\u4e1a\u516c\u53f8\u6210\u7acb\uff0c\u751f\u4ea7\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u4ef6\u548c\u8bbe\u5907\u3002<\/span><\/section>\n

<\/section>\n
\u4f46\u57282000\u5e74\u4ee3\u7684\u540e\u534a\u671f\uff0c\u4ed6\u4eec\u7a81\u7136\u4ece\u821e\u53f0\u4e0a\u6d88\u5931\u4e86\uff012006\u5e74\u4e4b\u540e\uff0c\u4eba\u5de5\u667a\u80fd\u754c\u51e0\u4e4e\u6ca1\u6709\u4eba\u63d0\u53ca\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u3002\u800c\u4ece2010\u5e74\u4ee3\u5f00\u59cb\uff0cGPU\u6210\u4e3a\u4e86\u795e\u7ecf\u7f51\u7edc\u7814\u7a76\u7684\u4e3b\u6d41\u5e73\u53f0\uff0c\u8fd9\u5bfc\u81f4\u4e86\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\u88ab\u91cd\u65b0\u547d\u540d\u4e3a\u6df1\u5ea6\u5b66\u4e60\u3002\u8bf7\u53c2\u9605\u56fe4<\/span>\uff08\u53f3\uff09<\/span>\uff0c\u4e86\u89e3\u4e0d\u540c\u65f6\u671f\u7684\u8d8b\u52bf\u5bf9\u6bd4\u3002<\/span><\/section>\n

<\/section>\n
\u6ca1\u6709\u4eba\u80fd\u591f\u771f\u6b63\u6307\u51fa\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u5728\u4eba\u5de5\u667a\u80fd\u7814\u7a76\u4e2d\u6d88\u4ea1\u7684\u786e\u5207\u539f\u56e0\u3002\u5c31\u50cf\u8bb0\u5fc6\u4f53\u4e00\u6837\uff0c\u4e5f\u8bb8\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u9886\u5148\u4e8e\u5b83\u7684\u65f6\u4ee3\u3002\u6216\u8005\u662f\u6d88\u8d39\u7ea7GPU\u7684\u6700\u7ec8\u5d1b\u8d77\u4f7f\u5176\u6210\u4e3a\u4e00\u4e2a\u5f15\u4eba\u6ce8\u76ee\u7684\u6df1\u5ea6\u5b66\u4e60\u5e73\u53f0\u3002\u4eba\u4eec\u53ea\u80fd\u60f3\u8c61\u5728\u4e00\u4e2a\u5e73\u884c\u5b87\u5b99\u4e2d\uff0c\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u7684\u6a21\u62df\u8ba1\u7b97\u673a\u82af\u7247\u8d62\u5f97\u4e86\u786c\u4ef6\u5f69\u7968<\/span>\uff08Hooker, 2020\uff09<\/span>\uff0c\u4eba\u5de5\u667a\u80fd\u7684\u72b6\u6001\u53ef\u80fd\u4f1a\u975e\u5e38\u4e0d\u540c\uff0c\u56e0\u4e3a\u90a3\u4e2a\u4e16\u754c\u7684\u6240\u6709\u8bbe\u5907\u90fd\u5d4c\u5165\u4e86\u5f3a\u5927\u7684\u5206\u5e03\u5f0f\u6a21\u62df\u5143\u80de\u81ea\u52a8\u673a\u3002<\/span><\/section>\n

<\/section>\n
\u7136\u800c\uff0c\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u548c\u6df1\u5ea6\u5b66\u4e60\u7684\u4e00\u4e2a\u5173\u952e\u533a\u522b\u662f\u53ef\u53ca\u6027\uff0c\u5728\u6211\u4eec\u770b\u6765\uff0c\u8fd9\u662f\u5b83\u6ca1\u6709\u6d41\u884c\u8d77\u6765\u7684\u4e3b\u8981\u539f\u56e0\u3002\u5728\u76ee\u524d\u7684\u6df1\u5ea6\u5b66\u4e60\u8303\u5f0f\u4e2d\uff0c\u6709\u4e00\u4e2a\u5b8c\u6574\u7684\u5de5\u5177\u751f\u6001\u7cfb\u7edf\uff0c\u65e8\u5728\u4f7f\u8bad\u7ec3\u548c\u90e8\u7f72\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\u53d8\u5f97\u5bb9\u6613\u3002\u901a\u8fc7\u5411\u6df1\u5ea6\u5b66\u4e60\u6846\u67b6\u63d0\u4f9b\u6570\u636e\u96c6<\/span>\uff08Chollet et al. 2015\uff09<\/span>\uff0c\u6216\u6a21\u62df\u4efb\u52a1\u73af\u5883<\/span>\uff08Hill et al. 2018\uff09<\/span>\uff0c\u7528\u6df1\u5ea6\u5b66\u4e60\u6846\u67b6\u8bad\u7ec3\u795e\u7ecf\u7f51\u7edc\u7684\u53c2\u6570\u4e5f\u76f8\u5bf9\u7b80\u5355\u3002\u6df1\u5ea6\u5b66\u4e60\u5de5\u5177\u88ab\u8bbe\u8ba1\u6210\u53ef\u4ee5\u88ab\u4efb\u4f55\u6709\u57fa\u672c\u7f16\u7a0b\u80cc\u666f\u7684\u4eba\u4f7f\u7528\u3002\u53e6\u4e00\u65b9\u9762\uff0c\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u662f\u4e3a\u7535\u5b50\u5de5\u7a0b\u5e08\u8bbe\u8ba1\u7684\uff0c\u5f53\u65f6\u5927\u591a\u6570\u7535\u5b50\u5de5\u7a0b\u7684\u5b66\u751f\u5bf9\u6a21\u62df\u7535\u8def\u7684\u4e86\u89e3\u591a\u4e8e\u7f16\u7a0b\u8bed\u8a00\u3002<\/span><\/section>\n

<\/section>\n
\u4e3a\u4e86\u8bf4\u660e\u8fd9\u4e00\u56f0\u96be\uff0c\u201c\u8bad\u7ec3\u201d\u4e00\u4e2a\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u9700\u8981\u89e3\u51b3\u4e00\u4e2a\u81f3\u5c11\u6709\u4e5d\u4e2a\u5e38\u5fae\u5206\u65b9\u7a0b\u7684\u7cfb\u7edf\uff0c\u4ee5\u786e\u5b9a\u652f\u914d\u6a21\u62df\u7535\u8def\u7684\u7cfb\u6570\uff0c\u4ece\u800c\u5b9a\u4e49\u7cfb\u7edf\u7684\u884c\u4e3a\uff01\u5728\u5b9e\u8df5\u4e2d\uff0c\u8bb8\u591a\u4ece\u4e1a\u8005\u9700\u8981\u4f9d\u9760\u4e00\u672c\u5df2\u77e5\u95ee\u9898\u89e3\u51b3\u65b9\u6848\u7684\u6307\u5357<\/span>\uff08Chua and Roska, 2002\uff09<\/span>\uff0c\u7136\u540e\u9488\u5bf9\u65b0\u95ee\u9898\u624b\u52a8\u8c03\u6574\u89e3\u51b3\u65b9\u6848\u3002\u6700\u7ec8\uff0c\u9057\u4f20\u7b97\u6cd5<\/span>\uff08\u548c\u65e9\u671f\u7248\u672c\u7684\u53cd\u5411\u4f20\u64ad\uff09<\/span>\u88ab\u63d0\u51fa\u6765\u8bad\u7ec3\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc<\/span>\uff08Kozek et al. 1993\uff09<\/span>\uff0c\u4f46\u5b83\u4eec\u9700\u8981\u6a21\u62df\u8f6f\u4ef6\u6765\u8bad\u7ec3\u548c\u6d4b\u8bd5\u7535\u8def\uff0c\u7136\u540e\u518d\u90e8\u7f72\u5230\u5b9e\u9645<\/span>\uff08\u548c\u9ad8\u5ea6\u5b9a\u5236\u7684\uff09<\/span>\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u786c\u4ef6\u4e0a\u3002<\/span><\/section>\n

<\/section>\n
\u4ece\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u4e2d\u6211\u4eec\u53ef\u4ee5\u5f97\u5230\u5f88\u591a\u6559\u8bad\u3002\u5b83\u4eec\u662f\u6a21\u62df\u548c\u6570\u5b57\u8ba1\u7b97\u7684\u6781\u5176\u5f3a\u5927\u7684\u6df7\u5408\u4f53\uff0c\u771f\u6b63\u7efc\u5408\u4e86\u5143\u80de\u81ea\u52a8\u673a\u548c\u795e\u7ecf\u7f51\u7edc\u3002\u4e0d\u5e78\u7684\u662f\uff0c\u5728\u5176\u6d88\u4ea1\u4e4b\u524d\uff0c\u6211\u4eec\u53ef\u80fd\u53ea\u89c1\u8bc1\u4e86\u5176\u5168\u90e8\u6f5c\u529b\u7684\u5f00\u59cb\u3002\u6700\u7ec8\uff0c\u5546\u4e1a\u5316\u7684GPU\u548c\u5c06\u795e\u7ecf\u7f51\u7edc\u62bd\u8c61\u4e3a\u7b80\u5355\u7684Python\u4ee3\u7801\u7684\u8f6f\u4ef6\u5de5\u5177\uff0c\u4f7f\u6df1\u5ea6\u5b66\u4e60\u5f97\u4ee5\u4f20\u64ad\u3002\u867d\u7136\u7ec6\u80de\u5f0f\u7c7b\u795e\u7ecf\u7f51\u7edc\u5df2\u7ecf\u6d88\u5931\u4e86\uff0c\u4f46\u6765\u81ea\u590d\u6742\u7cfb\u7edf\u7684\u6982\u5ff5\u548c\u60f3\u6cd5\uff0c\u5982\u5143\u80de\u81ea\u52a8\u673a\u3001\u81ea\u7ec4\u7ec7\u548c\u6d8c\u73b0\u884c\u4e3a\uff0c\u5e76\u6ca1\u6709\u6d88\u5931\u3002\u5c3d\u7ba1\u4ec5\u9650\u4e8e\u6570\u5b57\u786c\u4ef6\uff0c\u6211\u4eec\u6b63\u5728\u89c1\u8bc1\u96c6\u4f53\u667a\u80fd\u7684\u6982\u5ff5\u5728\u6df1\u5ea6\u5b66\u4e60\u7684\u8bb8\u591a\u9886\u57df\u91cd\u65b0\u51fa\u73b0\uff0c\u4ece\u56fe\u50cf\u751f\u6210\uff0c\u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60\uff0c\u5230\u96c6\u4f53\u548c\u5206\u5e03\u5f0f\u5b66\u4e60\u7b97\u6cd5\u3002\u6b63\u5982\u6211\u4eec\u5c06\u770b\u5230\u7684\uff0c\u8fd9\u4e9b\u6982\u5ff5\u901a\u8fc7\u4e3a\u4f20\u7edf\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\u7684\u4e00\u4e9b\u5c40\u9650\u548c\u9650\u5236\u63d0\u4f9b\u89e3\u51b3\u65b9\u6848\uff0c\u63a8\u8fdb\u6df1\u5ea6\u5b66\u4e60\u7684\u53d1\u5c55\u3002<\/span><\/section>\n

<\/section>\n

<\/section>\n

\n
\n
\n
\n
\n

<\/section>\n

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

4. \u6df1\u5ea6\u5b66\u4e60\u7684\u96c6\u4f53\u667a\u80fd<\/span><\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n
 <\/span><\/section>\n
\u96c6\u4f53\u667a\u80fd\u81ea\u7136\u4ea7\u751f\u4e8e\u7f51\u7edc\u4e2d\u591a\u4e2a\u4e2a\u4f53\u7684\u76f8\u4e92\u4f5c\u7528\uff0c\u81ea\u7ec4\u7ec7\u884c\u4e3a\u4ece\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\u4e2d\u6d8c\u73b0\u51fa\u6765\u4e5f\u5c31\u4e0d\u8db3\u4e3a\u5947\u4e86\u3002\u5f53\u6211\u4eec\u5728\u6574\u4e2a\u7f51\u7edc\u4e2d\u91c7\u7528\u76f8\u540c\u7684\u6743\u91cd\u53c2\u6570\u91cd\u590d\u8ba1\u7b97\u76f8\u540c\u7684\u6a21\u5757\u65f6\u5c24\u5176\u5982\u6b64\u3002\u4f8b\u5982\uff0cGilpin<\/span>\uff082019\uff09<\/span>\u89c2\u5bdf\u5230\u5143\u80de\u81ea\u52a8\u673a\u548c\u5377\u79ef\u795e\u7ecf\u7f51\u7edc<\/span>\uff08CNN\uff09<\/span>\u4e4b\u95f4\u7684\u5bc6\u5207\u8054\u7cfb\uff0c\u8fd9\u662f\u4e00\u79cd\u7ecf\u5e38\u7528\u4e8e\u56fe\u50cf\u5904\u7406\u7684\u795e\u7ecf\u7f51\u7edc\uff0c\u5b83\u5bf9\u6240\u6709\u7684\u8f93\u5165\u65bd\u52a0\u76f8\u540c\u7684\u6743\u91cd<\/span>\uff08\u6216\u8fc7\u6ee4\u5668\uff09<\/span>\u3002\u4e8b\u5b9e\u4e0a\uff0c\u4ed6\u4eec\u8868\u660e\u4efb\u4f55\u5143\u80de\u81ea\u52a8\u673a\u90fd\u53ef\u4ee5\u7528\u67d0\u79cd\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u6765\u8868\u793a\uff0c\u5e76\u4e14\u7528\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u4f18\u96c5\u5730\u6f14\u793a\u4e86\u5eb7\u5a01\u7684\u751f\u547d\u6e38\u620f<\/span>\uff08Conway et al. 1970\uff09<\/span>\u3002\u8fd9\u8bf4\u660e\u5728\u67d0\u4e9b\u60c5\u51b5\u4e0b\uff0c\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u53ef\u4ee5\u8868\u73b0\u51fa\u6709\u8da3\u7684\u81ea\u7ec4\u7ec7\u884c\u4e3a\u3002\u6211\u4eec\u5c06\u5728\u540e\u9762\u8ba8\u8bba\u7684Mordvintsev\u7b49\u4eba<\/span>\uff082020\uff09<\/span>\u7684\u51e0\u9879\u5de5\u4f5c\u90fd\u5229\u7528\u4e86\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u7684\u81ea\u7ec4\u7ec7\u7279\u6027\uff0c\u5e76\u4e3a\u56fe\u50cf\u518d\u751f\u7b49\u5e94\u7528\u5f00\u53d1\u4e86\u57fa\u4e8e\u795e\u7ecf\u7f51\u7edc\u7684\u5143\u80de\u81ea\u52a8\u673a\u3002<\/span><\/section>\n

<\/section>\n
\u5176\u4ed6\u7c7b\u578b\u7684\u795e\u7ecf\u7f51\u7edc\u67b6\u6784\uff0c\u5982\u56fe\u795e\u7ecf\u7f51\u7edc<\/span>\uff08Wu et al. 2020; Sanchez-Lengeling et al. 2021; Daigavane et al. 2021\uff09<\/span>\u660e\u786e\u5730\u5c06\u81ea\u7ec4\u7ec7\u4f5c\u4e3a\u4e2d\u5fc3\u7279\u5f81\uff0c\u5c06\u56fe\u7684\u6bcf\u4e2a\u8282\u70b9\u7684\u884c\u4e3a\u5efa\u6a21\u4e3a\u76f8\u540c\u7684\u795e\u7ecf\u7f51\u7edc\u6a21\u5757\uff0c\u5e76\u5c06\u4fe1\u606f\u4f20\u9012\u7ed9\u7531\u56fe\u7684\u8fde\u8fb9\u5b9a\u4e49\u7684\u90bb\u5c45\u3002\u4f20\u7edf\u4e0a\uff0c\u56fe\u795e\u7ecf\u7f51\u7edc\u88ab\u7528\u6765\u5206\u6790\u76f8\u5173\u7684\u7f51\u7edc\uff0c\u5982\u793e\u4ea4\u7f51\u7edc\u548c\u5206\u5b50\u7ed3\u6784\u3002\u6700\u8fd1\u7684\u5de5\u4f5c<\/span>\uff08Grattarola et al. 2021\uff09<\/span>\u4e5f\u8bc1\u660e\u4e86\u56fe\u795e\u7ecf\u7f51\u7edc\u6709\u80fd\u529b\u5b66\u4e60\u65e2\u5b9a\u5143\u80de\u81ea\u52a8\u673a\u7cfb\u7edf\u7684\u89c4\u5219\uff0c\u5982Voronoi\u56fe\uff0c\u6216\u7fa4\u4f53\u7684\u96c6\u805a\u884c\u4e3a<\/span>\uff08Schoenholz and Cubuk, 2020\uff09<\/span>\u3002\u6b63\u5982\u6211\u4eec\u5c06\u5728\u540e\u9762\u8ba8\u8bba\u7684\u90a3\u6837\uff0c\u56fe\u795e\u7ecf\u7f51\u7edc\u7684\u81ea\u7ec4\u7ec7\u7279\u6027\u6700\u8fd1\u88ab\u5e94\u7528\u4e8e\u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60\u9886\u57df\uff0c\u521b\u9020\u51fa\u4e86\u5177\u6709\u8fdc\u8d85\u4e00\u822c\u80fd\u529b\u7684\u4e3b\u4f53\u3002<\/span><\/section>\n

<\/section>\n
\u6211\u4eec\u5df2\u7ecf\u786e\u5b9a\u4e86\u56db\u4e2a\u6df1\u5ea6\u5b66\u4e60\u9886\u57df\uff0c\u5b83\u4eec\u5df2\u7ecf\u5f00\u59cb\u7eb3\u5165\u4e0e\u96c6\u4f53\u667a\u80fd\u76f8\u5173\u7684\u60f3\u6cd5\uff1a\uff081\uff09\u56fe\u50cf\u5904\u7406\uff0c\uff082\uff09\u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60\uff0c\uff083\uff09\u591a\u4e3b\u4f53\u5b66\u4e60\uff0c\u4ee5\u53ca\uff084\uff09\u5143\u5b66\u4e60\u3002\u6211\u4eec\u5c06\u5728\u672c\u8282\u4e2d\u8be6\u7ec6\u8ba8\u8bba\u6bcf\u4e2a\u9886\u57df\u5e76\u63d0\u4f9b\u76f8\u5e94\u7684\u5b9e\u4f8b\u3002<\/span><\/section>\n

<\/section>\n
4.1 \u56fe\u50cf\u5904\u7406<\/span><\/strong><\/span><\/section>\n

<\/section>\n
\u81ea\u7136\u754c\u4e2d\u7684\u9690\u542b\u5173\u7cfb\u548c\u91cd\u590d\u51fa\u73b0\u7684\u6a21\u5f0f<\/span>\uff08\u5982\u7eb9\u7406\u548c\u666f\u7269\uff09<\/span>\u53ef\u4ee5\u91c7\u7528\u5143\u80de\u81ea\u52a8\u673a\u7684\u65b9\u6cd5\uff0c\u4ee5\u5b66\u4e60\u81ea\u7136\u56fe\u50cf\u7684\u66ff\u4ee3\u8868\u793a\u3002\u4e0e\u7ec6\u80de\u7c7b\u5f0f\u795e\u7ecf\u7f51\u7edc\u4e00\u6837\uff0c<\/span>\uff08Mordvintsev et al. 2020\uff09<\/span>\u63d0\u51fa\u7684\u795e\u7ecf\u5143\u80de\u81ea\u52a8\u673a<\/span>\uff08neural CA\uff09<\/span>\u6a21\u578b\u5c06\u56fe\u50cf\u7684\u6bcf\u4e2a\u72ec\u7acb\u50cf\u7d20\u89c6\u4e3a\u4e00\u4e2a\u795e\u7ecf\u7f51\u7edc\u7ec6\u80de\u3002\u8fd9\u4e9b\u7f51\u7edc\u88ab\u8bad\u7ec3\u6210\u6839\u636e\u5176\u8fd1\u90bb\u7684\u72b6\u6001\u6765\u9884\u6d4b\u5176\u989c\u8272\uff0c\u4ece\u800c\u5f00\u53d1\u51fa\u4e00\u4e2a\u7528\u4e8e\u751f\u6210\u56fe\u50cf\u7684\u5f62\u6001\u53d1\u751f\u6a21\u578b\u3002\u4ed6\u4eec\u8bc1\u660e\u4e86\u6709\u53ef\u80fd\u901a\u8fc7\u8fd9\u79cd\u65b9\u5f0f\u8bad\u7ec3\u795e\u7ecf\u7f51\u7edc\u6765\u91cd\u5efa\u6574\u4e2a\u56fe\u50cf\uff0c\u5373\u4f7f\u6bcf\u4e2a\u7ec6\u80de\u7f3a\u4e4f\u5173\u4e8e\u5176\u4f4d\u7f6e\u7684\u4fe1\u606f\uff0c\u53ea\u4f9d\u9760\u6765\u81ea\u5176\u90bb\u5c45\u7684\u5c40\u90e8\u4fe1\u606f\u3002\u8fd9\u79cd\u65b9\u6cd5\u4f7f\u751f\u6210\u7b97\u6cd5\u80fd\u591f\u62b5\u6297\u566a\u58f0\u3002\u6b64\u5916\uff0c\u5141\u8bb8\u56fe\u50cf\u5728\u53d7\u635f\u65f6\u6062\u590d\u3002\u795e\u7ecf\u5143\u80de\u81ea\u52a8\u673a\u7684\u6269\u5c55<\/span>\uff08Randazzo et al. 2020\uff09<\/span>\u4f7f\u5355\u4e2a\u7ec6\u80de\u80fd\u591f\u6267\u884c\u56fe\u50cf\u5206\u7c7b\u4efb\u52a1\uff0c\u5982\u624b\u5199\u6570\u5b57\u5206\u7c7b<\/span>\uff08MNIST\uff09<\/span>\uff0c\u53ea\u68c0\u67e5\u5355\u4e2a\u50cf\u7d20\u7684\u5185\u5bb9\uff0c\u5e76\u5c06\u4fe1\u606f\u4f20\u9012\u7ed9\u7ec6\u80de\u7684\u8fd1\u90bb<\/span>\uff08\u89c1\u56fe5\uff09<\/span>\u3002\u968f\u7740\u65f6\u95f4\u7684\u63a8\u79fb\uff0c\u5c06\u5f62\u6210\u5173\u4e8e\u54ea\u4e2a\u6570\u5b57\u6700\u6709\u53ef\u80fd\u7684\u50cf\u7d20\u7684\u5171\u8bc6\uff0c\u4f46\u6709\u8da3\u7684\u662f\uff0c\u6839\u636e\u50cf\u7d20\u7684\u4f4d\u7f6e\u53ef\u80fd\u4f1a\u4ea7\u751f\u5206\u6b67\uff0c\u7279\u522b\u662f\u5982\u679c\u56fe\u50cf\u88ab\u6545\u610f\u7ed8\u5236\u6210\u4ee3\u8868\u4e0d\u540c\u7684\u6570\u5b57\u3002<\/span><\/section>\n

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

\u56fe5. \u7531\uff08Randazzo et al. 2020\uff09\u521b\u5efa\u7684\u8bad\u7ec3\u6709\u7d20\u7684\u795e\u7ecf\u5143\u80de\u81ea\u52a8\u673a\u8bc6\u522bMNIST\u6570\u5b57\uff0c\u4e5f\u53ef\u4f5c\u4e3a\u4e92\u52a8\u7f51\u7edc\u6f14\u793a\u3002\u6bcf\u4e2a\u7ec6\u80de\u53ea\u5141\u8bb8\u770b\u5230\u4e00\u4e2a\u50cf\u7d20\u7684\u5185\u5bb9\uff0c\u5e76\u4e0e\u5b83\u7684\u90bb\u5c45\u4ea4\u6d41\u3002\u968f\u7740\u65f6\u95f4\u7684\u63a8\u79fb\uff0c\u5c06\u5f62\u6210\u4e00\u4e2a\u5171\u8bc6\uff0c\u5373\u54ea\u4e2a\u6570\u5b57\u662f\u6700\u6709\u53ef\u80fd\u7684\u50cf\u7d20\uff0c\u4f46\u6709\u8da3\u7684\u662f\uff0c\u6839\u636e\u9884\u6d4b\u7684\u50cf\u7d20\u7684\u4f4d\u7f6e\uff0c\u53ef\u80fd\u4f1a\u4ea7\u751f\u5206\u6b67\u3002<\/span><\/p>\n


<\/section>\n
\u795e\u7ecf\u5143\u80de\u81ea\u52a8\u673a\u7684\u518d\u751f\u529f\u80fd\u5df2\u7ecf\u88ab\u63a2\u7d22\u5230\u4e8c\u7ef4\u56fe\u50cf\u4e4b\u5916\u3002\u5728\u540e\u6765\u7684\u5de5\u4f5c\u4e2d\uff0c<\/span>\uff08Zhang et al. 2021\uff09<\/span>\u91c7\u7528\u4e86\u7c7b\u4f3c\u7684\u65b9\u6cd5\u6765\u751f\u6210\u4e09\u7ef4\u4f53\u7d20\u3002\u8fd9\u5bf9\u9ad8\u5206\u8fa8\u7387\u7684\u4e09\u7ef4\u626b\u63cf\u7279\u522b\u6709\u7528\uff0c\u56e0\u4e3a\u4e09\u7ef4\u5f62\u72b6\u6570\u636e\u5f80\u5f80\u662f\u7528\u7a00\u758f\u548c\u4e0d\u5b8c\u6574\u7684\u70b9\u6765\u63cf\u8ff0\u7684\u3002\u4f7f\u7528\u751f\u6210\u5f0f\u5143\u80de\u81ea\u52a8\u673a\uff0c\u4ed6\u4eec\u53ef\u4ee5\u53ea\u4ece\u90e8\u5206\u70b9\u7684\u96c6\u5408\u4e2d\u6062\u590d\u5b8c\u6574\u7684\u4e09\u7ef4\u5f62\u72b6\u3002\u8fd9\u79cd\u65b9\u6cd5\u4e5f\u9002\u7528\u4e8e\u7eaf\u751f\u6210\u9886\u57df\u4e4b\u5916\uff0c\u4e5f\u53ef\u4ee5\u5e94\u7528\u4e8e\u6d3b\u52a8\u73af\u5883\u4e2d\u4eba\u5de5\u4e3b\u4f53\u7684\u6784\u5efa\uff0c\u5982\u6e38\u620f\u300a\u6211\u7684\u4e16\u754c\u300b<\/span>\uff08Minecraft\uff09(Sudhakaran et al. 2021\uff09<\/span>\u3002\u8bad\u7ec3\u795e\u7ecf\u5143\u80de\u81ea\u52a8\u673a\u4eceMinecraft\u4e2d\u751f\u957f\u51fa\u590d\u6742\u7684\u5b9e\u4f53\uff0c\u5982\u57ce\u5821\u3001\u516c\u5bd3\u533a\u548c\u6811\uff0c\u5176\u4e2d\u4e00\u4e9b\u662f\u7531\u6210\u5343\u4e0a\u4e07\u7684\u5757\u4f53\u7d20\u7ec4\u6210\u7684\u3002\u9664\u4e86\u518d\u751f\uff0c\u4ed6\u4eec\u7684\u7cfb\u7edf\u8fd8\u80fd\u91cd\u65b0\u751f\u6210\u51fa\u7b80\u5355\u529f\u80fd\u673a\u5668\u7684\u90e8\u4ef6<\/span>\uff08\u5982\u6e38\u620f\u4e2d\u7684\u865a\u62df\u751f\u7269\uff09<\/span>\uff0c\u4ed6\u4eec\u6f14\u793a\u4e86\u4e00\u4e2a\u5f62\u6001\u751f\u7269\u5728\u865a\u62df\u4e16\u754c\u4e2d\u88ab\u5207\u6210\u4e24\u534a\u65f6\u4f1a\u957f\u6210\u4e24\u4e2a\u4e0d\u540c\u7684\u751f\u7269<\/span>\uff08\u89c1\u56fe6\uff09<\/span>\u3002<\/span><\/section>\n

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

\u56fe6.  \u795e\u7ecf\u5143\u80de\u81ea\u52a8\u673a\u4e5f\u88ab\u5e94\u7528\u4e8eMinecraft\u5b9e\u4f53\u7684\u518d\u751f\u3002\u5728\u8fd9\u9879\u5de5\u4f5c\u4e2d\uff0c\u4f5c\u8005\u7684\u8868\u660e\u4e0d\u4ec5\u4f7fMinecraft\u5efa\u7b51\u3001\u6811\u6728\u7684\u518d\u751f\uff0c\u800c\u4e14\u53ef\u4ee5\u4f7f\u6e38\u620f\u4e2d\u7684\u7b80\u5355\u529f\u80fd\u673a\u5668\uff0c\u5982\u8815\u866b\u72b6\u751f\u7269\u7684\u518d\u751f\uff0c\u5728\u5207\u6210\u4e24\u534a\u65f6\u751a\u81f3\u53ef\u4ee5\u518d\u751f\u4e3a\u4e24\u4e2a\u4e0d\u540c\u7684\u751f\u7269\u3002<\/span><\/p>\n


<\/section>\n
\u5143\u80de\u81ea\u52a8\u673a\u4e5f\u81ea\u7136\u9002\u7528\u4e8e\u4e3a\u56fe\u50cf\u63d0\u4f9b\u89c6\u89c9\u89e3\u91ca\u3002<\/span>(Qin et al. 2018\uff09<\/span>\u7814\u7a76\u4e86\u4f7f\u7528\u5c42\u6b21\u5143\u80de\u81ea\u52a8\u673a\u6a21\u578b\u8fdb\u884c\u89c6\u89c9\u663e\u8457\u6027\u5206\u6790\uff0c\u4ee5\u8bc6\u522b\u56fe\u50cf\u4e2d\u7a81\u51fa\u7684\u9879\u76ee\u3002\u901a\u8fc7\u8ba9\u5143\u80de\u81ea\u52a8\u673a\u5bf9\u4ece\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u4e2d\u63d0\u53d6\u7684\u89c6\u89c9\u7279\u5f81\u8fdb\u884c\u64cd\u4f5c\uff0c\u4ed6\u4eec\u80fd\u591f\u8fed\u4ee3\u5730\u6784\u5efa\u56fe\u50cf\u7684\u591a\u5c3a\u5ea6\u663e\u8457\u6027\u5730\u56fe\uff0c\u6700\u7ec8\u7684\u56fe\u50cf\u63a5\u8fd1\u4e8e\u76ee\u6807\u9879\u76ee\u3002Sandler\u7b49\u4eba<\/span>\uff08 2020\uff09<\/span>\u540e\u6765\u7814\u7a76\u4e86\u5143\u80de\u81ea\u52a8\u673a\u5728\u56fe\u50cf\u5206\u5272\u4efb\u52a1\u4e2d\u7684\u5e94\u7528\uff0c\u8fd9\u662f\u4e00\u4e2a\u6df1\u5ea6\u5b66\u4e60\u4eab\u6709\u5de8\u5927\u6210\u529f\u7684\u9886\u57df\u3002\u4ed6\u4eec\u8bc1\u660e\u4e86\u4f7f\u7528\u89c4\u5219\u76f8\u5bf9\u7b80\u5355\u7684\u5143\u80de\u81ea\u52a8\u673a\u6267\u884c\u590d\u6742\u7684\u5206\u5272\u4efb\u52a1\u7684\u53ef\u884c\u6027<\/span>\uff08\u53ea\u670910K\u7684\u795e\u7ecf\u7f51\u7edc\u53c2\u6570\uff09<\/span>\uff0c\u5176\u4f18\u70b9\u662f\u8be5\u65b9\u6cd5\u80fd\u591f\u6269\u5c55\u5230\u4ee4\u4eba\u96be\u4ee5\u7f6e\u4fe1\u7684\u5927\u56fe\u50cf\u5c3a\u5bf8\uff0c\u8fd9\u5bf9\u4e8e\u62e5\u6709\u6570\u767e\u4e07\u751a\u81f3\u6570\u5341\u4ebf\u6a21\u578b\u53c2\u6570\u7684\u4f20\u7edf\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u6765\u8bf4\u662f\u4e00\u4e2a\u6311\u6218\uff0c\u56e0\u4e3a\u5b83\u53d7\u5230\u4e86GPU\u5185\u5b58\u7684\u9650\u5236\u3002<\/span><\/section>\n

<\/section>\n
4.2 \u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60<\/span><\/strong><\/span><\/section>\n

<\/section>\n
\u6df1\u5ea6\u5b66\u4e60\u7684\u5174\u8d77\u50ac\u751f\u4e86\u4f7f\u7528\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u8fdb\u884c\u5f3a\u5316\u5b66\u4e60\uff0c\u5373\u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60<\/span>\uff08Deep RL\uff09<\/span>\uff0c\u4e3a\u5f3a\u5316\u5b66\u4e60\u4e3b\u4f53\u914d\u5907\u4e86\u73b0\u4ee3\u795e\u7ecf\u7f51\u7edc\u67b6\u6784\uff0c\u53ef\u4ee5\u89e3\u51b3\u66f4\u590d\u6742\u7684\u95ee\u9898\uff0c\u5982\u9ad8\u7ef4\u8fde\u7eed\u63a7\u5236\u6216\u57fa\u4e8e\u89c6\u89c9\u7684\u50cf\u7d20\u89c2\u6d4b\u4efb\u52a1\u3002\u867d\u7136\u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60\u4e0e\u6df1\u5ea6\u5b66\u4e60\u6709\u7740\u76f8\u540c\u7684\u6210\u529f\u7279\u8bc1\uff0c\u5373\u91c7\u7528\u8db3\u591f\u7684\u8ba1\u7b97\u8d44\u6e90\u4e00\u822c\u4f1a\u5bfc\u81f4\u76ee\u6807\u8bad\u7ec3\u4efb\u52a1\u7684\u89e3\u51b3\u65b9\u6848\u88ab\u53d1\u73b0\u3002\u4f46\u4e0e\u6df1\u5ea6\u5b66\u4e60\u4e00\u6837\uff0c\u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60\u4e5f\u6709\u5176\u5c40\u9650\u6027\u3002\u5f53\u4efb\u52a1\u7a0d\u6709\u6539\u53d8\u65f6\uff0c\u4e3a\u6267\u884c\u67d0\u4e00\u7279\u5b9a\u4efb\u52a1\u800c\u8bad\u7ec3\u7684\u4e3b\u4f53\u5f80\u5f80\u4f1a\u5931\u8d25\u3002\u6b64\u5916\uff0c\u795e\u7ecf\u7f51\u7edc\u89e3\u51b3\u65b9\u6848\u4e00\u822c\u53ea\u9002\u7528\u4e8e\u5177\u6709\u660e\u786e\u7684\u8f93\u5165\u548c\u8f93\u51fa\u6620\u5c04\u7684\u7279\u5b9a\u5f62\u6001\u3002\u4f8b\u5982\uff0c\u4e3a\u56db\u6761\u817f\u7684\u8682\u8681\u8bad\u7ec3\u7684\u8fd0\u52a8\u7b56\u7565\u53ef\u80fd\u5bf9\u516d\u6761\u817f\u7684\u8682\u8681\u4e0d\u8d77\u4f5c\u7528\uff0c\u800c\u4e00\u4e2a\u671f\u671b\u63a5\u653610\u4e2a\u8f93\u5165\u7684\u63a7\u5236\u5668\uff0c\u5982\u679c\u4f60\u7ed9\u5b835\u4e2a\u621620\u4e2a\u8f93\u5165\uff0c\u5c31\u4e0d\u4f1a\u8d77\u4f5c\u7528\u3002<\/span><\/section>\n

<\/section>\n
\u8fdb\u5316\u8ba1\u7b97\u9886\u57df\u8f83\u65e9\u5730\u5f00\u59cb\u5e94\u5bf9\u8fd9\u4e9b\u6311\u6218\uff0c\u5728\u652f\u914d\u4eba\u5de5\u4e3b\u4f53\u8bbe\u8ba1\u7684\u8fdb\u5316\u8fc7\u7a0b\u4e2d\u52a0\u5165\u4e86\u6a21\u5757\u5316<\/span>\uff08Schilling, 2000; Schilling and Steensma, 2001\uff09<\/span>\u3002\u8ba9\u7531\u76f8\u540c\u4f46\u72ec\u7acb\u7684\u6a21\u5757\u7ec4\u6210\u7684\u4e3b\u4f53\u901a\u8fc7\u6a21\u5757\u95f4\u7684\u5c40\u90e8\u76f8\u4e92\u4f5c\u7528\u4fc3\u8fdb\u81ea\u7ec4\u7ec7\uff0c\u4f7f\u7cfb\u7edf\u5bf9\u4e3b\u4f53\u5f62\u6001\u7684\u53d8\u5316\u5177\u6709\u7a33\u5065\u6027\uff0c\u8fd9\u662f\u8fdb\u5316\u7cfb\u7edf\u7684\u4e00\u4e2a\u57fa\u672c\u8981\u6c42\u3002\u8fd9\u4e9b\u60f3\u6cd5\u5df2\u7ecf\u5728\u8f6f\u4f53\u673a\u5668\u4eba\u7684\u5de5\u4f5c\u6587\u732e\u4e2d\u63d0\u51fa<\/span>\uff08Cheney et al. 2014\uff09<\/span>\uff0c\u5176\u4e2d\u673a\u5668\u4eba\u7531\u7531\u4f53\u7d20\u7ec6\u80de\u7f51\u683c\u7ec4\u6210\uff0c\u6bcf\u4e2a\u7ec6\u80de\u7531\u4e00\u4e2a\u72ec\u7acb\u7684\u795e\u7ecf\u7f51\u7edc\u63a7\u5236\uff0c\u5177\u6709\u5c40\u90e8\u611f\u77e5\u529f\u80fd\uff0c\u53ef\u4ee5\u4ea7\u751f\u4e00\u4e2a\u5c40\u90e8\u7684\u884c\u52a8\u3002\u901a\u8fc7\u4fe1\u606f\u4f20\u9012\uff0c\u6784\u6210\u673a\u5668\u4eba\u7684\u7ec6\u80de\u7fa4\u80fd\u591f\u81ea\u7ec4\u7ec7\uff0c\u5e76\u6267\u884c\u4e00\u7cfb\u5217\u7684\u8fd0\u52a8\u4efb\u52a1<\/span>\uff08\u89c1\u56fe7\uff09<\/span>\u3002\u540e\u6765\u7684\u5de5\u4f5c<\/span>\uff08Joachimczak et al. 2016\uff09<\/span>\u751a\u81f3\u63d0\u51fa\u5728\u7ec6\u80de\u653e\u7f6e\u7684\u8fdb\u5316\u8fc7\u7a0b\u4e2d\u52a0\u5165\u53d8\u5f62\uff0c\u4ee5\u4ea7\u751f\u5bf9\u4e00\u7cfb\u5217\u73af\u5883\u7684\u7a33\u5065\u914d\u7f6e\u3002<\/span><\/section>\n

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

\u56fe7. \u4e8c\u7ef4\u548c\u4e09\u7ef4\u7684\u8f6f\u4f53\u673a\u5668\u4eba\u6a21\u62df\u5b9e\u4f8b\u3002\u6bcf\u4e2a\u7ec6\u80de\u4ee3\u8868\u4e00\u4e2a\u5355\u72ec\u7684\u795e\u7ecf\u7f51\u7edc\uff0c\u5177\u6709\u5c40\u90e8\u611f\u77e5\u529f\u80fd\uff0c\u4ea7\u751f\u5c40\u90e8\u52a8\u4f5c\uff0c\u5305\u62ec\u4e0e\u90bb\u8fd1\u7684\u7ec6\u80de\u4ea4\u6d41\u3002\u8bad\u7ec3\u8fd9\u4e9b\u7cfb\u7edf\u6765\u5b8c\u6210\u5404\u79cd\u8fd0\u52a8\u4efb\u52a1\uff0c\u4e0d\u4ec5\u6d89\u53ca\u5230\u8bad\u7ec3\u795e\u7ecf\u7f51\u7edc\uff0c\u800c\u4e14\u8fd8\u6d89\u53ca\u5230\u5f62\u6210\u4e3b\u4f53\u5f62\u6001\u7684\u8f6f\u4f53\u7ec6\u80de\u7684\u8bbe\u8ba1\u548c\u653e\u7f6e\u3002\u56fe\u81ea\uff08Horibe et al. 2021\uff09\u3002<\/span><\/p>\n


<\/section>\n
\u6700\u8fd1\uff0c\u8f6f\u4f53\u673a\u5668\u4eba\u751a\u81f3\u4e0e\u524d\u9762\u8ba8\u8bba\u7684\u795e\u7ecf\u5143\u80de\u81ea\u52a8\u673a\u65b9\u6cd5\u76f8\u7ed3\u5408\uff0c\u4f7f\u8fd9\u4e9b\u673a\u5668\u4eba\u80fd\u591f\u81ea\u6211\u518d\u751f\u3002Horibe\u7b49\u4eba<\/span>\uff082021\uff09<\/span>\u5f25\u5408\u4e86\u901a\u5e38\u5728\u6df1\u5ea6\u5f3a\u5316\u5b66\u793e\u533a\u5b8c\u6210\u7684\u7b56\u7565\u4f18\u5316<\/span>\uff08\u76ee\u6807\u662f\u627e\u5230\u7b56\u7565\u795e\u7ecf\u7f51\u7edc\u7684\u6700\u4f73\u53c2\u6570\uff09<\/span>\u548c\u5728\u8f6f\u4f53\u6587\u732e\u4e2d\u5b8c\u6210\u7684\u90a3\u79cd\u5f62\u6001-\u7b56\u7565\u5171\u540c\u8fdb\u5316<\/span>\uff08\u5f62\u6001\u548c\u7b56\u7565\u795e\u7ecf\u7f51\u7edc\u4e00\u8d77\u88ab\u4f18\u5316\uff09<\/span>\u5de5\u4f5c\u4e4b\u95f4\u7684\u5dee\u8ddd\u3002Bhatia\u7b49\u4eba<\/span>\uff082021\uff09<\/span>\u6700\u8fd1\u5f00\u53d1\u4e86\u4e00\u4e2a\u7c7b\u4f3c OpenAI Gym<\/span>\uff08Brockman et al. 2016\uff09<\/span>\u7684\u73af\u5883\uff0c\u540d\u4e3aEvolution Gym\uff0c\u8fd9\u662f\u4e00\u4e2a\u5f00\u53d1\u548c\u6bd4\u8f83\u5171\u540c\u4f18\u5316\u8bbe\u8ba1\u548c\u63a7\u5236\u7684\u7b97\u6cd5\u7684\u57fa\u51c6\uff0c\u5b83\u63d0\u4f9b\u4e86\u4e00\u4e2a\u9ad8\u6548\u7684\u8f6f\u4f53\u673a\u5668\u4eba\u6a21\u62df\u5668\uff0c\u7528C++\u7f16\u5199\uff0c\u5e26\u6709Python\u63a5\u53e3\u3002<\/span><\/section>\n

<\/section>\n
\u6a21\u5757\u5316\u3001\u53bb\u4e2d\u5fc3\u5316\u7684\u81ea\u7ec4\u7ec7\u63a7\u5236\u5668\u4e5f\u5f00\u59cb\u5728\u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60\u793e\u533a\u8fdb\u884c\u63a2\u7d22\u3002Wang\u7b49\u4eba<\/span>\uff082018\uff09<\/span>\u548cHuang\u7b49\u4eba<\/span>\uff082020\uff09<\/span>\u63a2\u7d22\u4e86\u4f7f\u7528\u6a21\u5757\u5316\u795e\u7ecf\u7f51\u7edc\u6765\u63a7\u5236\u6a21\u62df\u673a\u5668\u4eba\u7684\u6bcf\u4e2a\u5355\u72ec\u7684\u6267\u884c\u5668\uff0c\u4ee5\u5b9e\u73b0\u8fde\u7eed\u63a7\u5236\u3002\u4ed6\u4eec\u5c06\u5168\u5c40\u8fd0\u52a8\u7b56\u7565\u8868\u8fbe\u4e3a\u6a21\u5757\u5316\u795e\u7ecf\u7f51\u7edc\u7684\u96c6\u5408<\/span>\uff08\u5728\uff08Huang et al. 2020\uff09\u7684\u60c5\u51b5\u4e0b\uff0c\u4e5f\u662f\u76f8\u540c\u7684\u7f51\u7edc\uff09<\/span>\uff0c\u5bf9\u5e94\u4e8e\u6bcf\u4e2a\u4e3b\u4f53\u7684\u6267\u884c\u5668\uff0c\u5e76\u4f7f\u7528\u5f3a\u5316\u5b66\u4e60\u8bad\u7ec3\u8be5\u7cfb\u7edf\u3002\u50cf\u8f6f\u4f53\u673a\u5668\u4eba\u4e00\u6837\uff0c\u6bcf\u4e2a\u6a21\u5757\u53ea\u8d1f\u8d23\u63a7\u5236\u5176\u76f8\u5e94\u7684\u6267\u884c\u5668\uff0c\u5e76\u4e14\u53ea\u4ece\u5176\u672c\u5730\u4f20\u611f\u5668\u63a5\u6536\u4fe1\u606f<\/span>\uff08\u89c1\u56fe8\uff09<\/span>\u3002\u6d88\u606f\u5728\u76f8\u90bb\u7684\u6a21\u5757\u4e4b\u95f4\u4f20\u9012\uff0c\u5728\u9065\u8fdc\u7684\u6a21\u5757\u4e4b\u95f4\u4f20\u64ad\u4fe1\u606f\u3002\u4ed6\u4eec\u8868\u660e\uff0c\u4e00\u4e2a\u5355\u4e00\u7684\u6a21\u5757\u7b56\u7565\u53ef\u4ee5\u4e3a\u51e0\u79cd\u4e0d\u540c\u7684\u673a\u5668\u4eba\u5f62\u6001\u4ea7\u751f\u8fd0\u52a8\u884c\u4e3a\uff0c\u5e76\u8868\u660e\u8fd9\u4e9b\u7b56\u7565\u53ef\u4ee5\u63a8\u5e7f\u5230\u8bad\u7ec3\u4e2d\u672a\u77e5\u7684\u5f62\u6001\u53d8\u5316\uff0c\u5982\u6709\u989d\u5916\u817f\u7684\u751f\u7269\u3002\u4e0e\u8f6f\u4f53\u673a\u5668\u4eba\u7684\u60c5\u51b5\u4e00\u6837\uff0c\u8fd9\u4e9b\u7ed3\u679c\u4e5f\u8bc1\u660e\u4e86\u901a\u8fc7\u53bb\u4e2d\u5fc3\u5316\u7684\u6a21\u5757\u4e4b\u95f4\u7684\u4fe1\u606f\u4f20\u9012\uff0c\u51fa\u73b0\u4e86\u96c6\u4e2d\u7684\u534f\u8c03\uff0c\u8fd9\u4e9b\u6a21\u5757\u6b63\u5728\u96c6\u4f53\u4f18\u5316\u4e00\u4e2a\u5171\u540c\u7684\u5956\u52b1\u3002<\/span><\/section>\n

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

\u56fe8. \u4f20\u7edf\u7684\u5f3a\u5316\u5b66\u4e60\u65b9\u6cd5\u4e3a\u5177\u6709\u56fa\u5b9a\u5f62\u6001\u7684\u7279\u5b9a\u673a\u5668\u4eba\u8bad\u7ec3\u4e00\u4e2a\u7279\u5b9a\u7684\u7b56\u7565\u3002\u4f46\u6700\u8fd1\u7684\u5de5\u4f5c\uff0c\u5982\u8fd9\u91cc\u663e\u793a\u7684\uff08Huang et al. 2020\uff09\u8bd5\u56fe\u8bad\u7ec3\u4e00\u4e2a\u5355\u4e00\u7684\u6a21\u5757\u5316\u795e\u7ecf\u7f51\u7edc\uff0c\u8d1f\u8d23\u63a7\u5236\u673a\u5668\u4eba\u7684\u5355\u4e00\u90e8\u5206\u3002\u56e0\u6b64\uff0c\u6bcf\u4e2a\u673a\u5668\u4eba\u7684\u5168\u5c40\u7b56\u7565\u662f\u8fd9\u4e9b\u76f8\u540c\u7684\u6a21\u5757\u5316\u795e\u7ecf\u7f51\u7edc\u534f\u8c03\u7684\u7ed3\u679c\u3002\u4ed6\u4eec\u8868\u660e\uff0c\u8fd9\u6837\u7684\u7cfb\u7edf\u53ef\u4ee5\u5728\u5404\u79cd\u4e0d\u540c\u7684\u9aa8\u67b6\u7ed3\u6784\u4e0a\u8fdb\u884c\u6cdb\u5316\uff0c\u4ece\u8df3\u8dc3\u8005\u5230\u56db\u8db3\u52a8\u7269\uff0c\u751a\u81f3\u662f\u4e00\u4e9b\u672a\u89c1\u8fc7\u7684\u5f62\u6001\u3002<\/span><\/p>\n


<\/section>\n
\u4e0a\u8ff0\u5de5\u4f5c\u6697\u793a\u4e86\u5177\u8eab\u8ba4\u77e5\u7684\u529b\u91cf\uff0c\u5b83\u5f3a\u8c03\u4e86\u4e3b\u4f53\u7684\u8eab\u4f53\u5728\u4ea7\u751f\u884c\u4e3a\u4e2d\u7684\u4f5c\u7528\u3002\u867d\u7136\u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60\u7684\u5927\u90e8\u5206\u5de5\u4f5c\u91cd\u70b9\u662f\u4e3a\u5177\u6709\u56fa\u5b9a\u8bbe\u8ba1\u7684\u4e3b\u4f53<\/span>\uff08\u5982\u53cc\u8db3\u673a\u5668\u4eba\u3001\u4eba\u5f62\u673a\u5668\u4eba\u6216\u673a\u68b0\u81c2\uff09<\/span>\u5b66\u4e60\u795e\u7ecf\u7f51\u7edc\u7b56\u7565\uff0c\u4f46\u5177\u8eab\u667a\u80fd\u662f\u8be5\u5b50\u9886\u57df\u4e2d\u6b63\u5728\u805a\u96c6\u5174\u8da3\u7684\u9886\u57df<\/span>\uff08Ha\uff0c2018; Pathak et al. 2019\uff09<\/span>\u3002\u53d7<\/span>\uff08Stoy et al. 2010; Rubenstein et al. 2014; Hamann\uff0c2018\uff09<\/span>\u4e4b\u524d\u5173\u4e8e\u81ea\u6211\u914d\u7f6e\u6a21\u5757\u5316\u673a\u5668\u4eba\u7684\u51e0\u9879\u5de5\u4f5c\u7684\u542f\u53d1\uff0cPathak\u7b49\u4eba<\/span>\uff082019\uff09<\/span>\u7814\u7a76\u4e86\u4e00\u4e2a\u539f\u59cb\u4e3b\u4f53\u7684\u96c6\u5408\uff0c\u8fd9\u4e9b\u4e3b\u4f53\u5b66\u4e60\u81ea\u7ec4\u88c5\u6210\u4e00\u4e2a\u590d\u6742\u7684\u8eab\u4f53\uff0c\u540c\u65f6\u4e5f\u5b66\u4e60\u4e00\u4e2a\u5c40\u90e8\u7b56\u7565\u6765\u63a7\u5236\u8eab\u4f53\uff0c\u800c\u6ca1\u6709\u4e00\u4e2a\u660e\u786e\u7684\u96c6\u4e2d\u63a7\u5236\u5355\u5143\u3002\u6bcf\u4e2a\u539f\u59cb\u4e3b\u4f53<\/span>\uff08\u7531\u4e00\u4e2a\u80a2\u4f53\u548c\u4e00\u4e2a\u9a6c\u8fbe\u7ec4\u6210\uff09<\/span>\u53ef\u4ee5\u4e0e\u9644\u8fd1\u7684\u4e3b\u4f53\u8054\u7cfb\u8d77\u6765\uff0c\u5141\u8bb8\u51fa\u73b0\u590d\u6742\u7684\u5f62\u6001\u3002\u4ed6\u4eec\u7684\u7ed3\u679c\u8868\u660e\uff0c\u8fd9\u4e9b\u52a8\u6001\u548c\u6a21\u5757\u5316\u7684\u4e3b\u4f53\u5bf9\u6761\u4ef6\u7684\u53d8\u5316\u5177\u6709\u7a33\u5065\u6027\u3002\u8fd9\u4e9b\u7b56\u7565\u4e0d\u4ec5\u53ef\u4ee5\u63a8\u5e7f\u5230\u672a\u77e5\u7684\u73af\u5883\uff0c\u800c\u4e14\u53ef\u4ee5\u63a8\u5e7f\u5230\u672a\u77e5\u7684\u7531\u66f4\u591a\u6a21\u5757\u7ec4\u6210\u7684\u5f62\u6001\u3002\u6211\u4eec\u6ce8\u610f\u5230\uff0c\u8fd9\u4e9b\u60f3\u6cd5\u53ef\u4ee5\u7528\u6765\u8ba9\u4e00\u822c\u7684\u6df1\u5ea6\u5b66\u4e60\u7cfb\u7edf<\/span>\uff08\u4e0d\u5c40\u9650\u4e8e\u5f3a\u5316\u5b66\u4e60\uff09<\/span>\u62e5\u6709\u66f4\u7075\u6d3b\u7684\u67b6\u6784\uff0c\u751a\u81f3\u53ef\u4ee5\u5b66\u4e60\u673a\u5668\u5b66\u4e60\u7b97\u6cd5\uff0c\u6211\u4eec\u5c06\u5728\u540e\u9762\u7684\u5143\u5b66\u4e60\u90e8\u5206\u8ba8\u8bba\u8fd9\u4e2a\u95ee\u9898\u3002<\/span><\/section>\n

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

\u56fe9. \u81ea\u7ec4\u7ec7\u4e5f\u4f7f\u5f3a\u5316\u5b66\u4e60\u73af\u5883\u4e2d\u7684\u7cfb\u7edf\u80fd\u591f\u4e3a\u7279\u5b9a\u7684\u4efb\u52a1\u81ea\u6211\u914d\u7f6e\u5b83\u81ea\u5df1\u7684\u8bbe\u8ba1\u3002\u5728\uff08Pathak et al. 2019\uff09\u4e2d\uff0c\u4f5c\u8005\u63a2\u7d22\u4e86\u8fd9\u79cd\u52a8\u6001\u548c\u6a21\u5757\u5316\u7684\u4e3b\u4f53\uff0c\u5e76\u8868\u660e\u5b83\u4eec\u4e0d\u4ec5\u53ef\u4ee5\u6cdb\u5316\u5230\u672a\u77e5\u7684\u73af\u5883\uff0c\u8fd8\u53ef\u4ee5\u6cdb\u5316\u5230\u7531\u989d\u5916\u6a21\u5757\u7ec4\u6210\u7684\u672a\u77e5\u7684\u5f62\u6001\u3002<\/span><\/p>\n


<\/section>\n
\u9664\u4e86\u9002\u5e94\u4e0d\u65ad\u53d8\u5316\u7684\u5f62\u6001\u548c\u73af\u5883\u5916\uff0c\u81ea\u7ec4\u7ec7\u7cfb\u7edf\u8fd8\u80fd\u9002\u5e94\u5176\u611f\u5b98\u8f93\u5165\u7684\u53d8\u5316\u3002\u611f\u77e5\u66ff\u4ee3\u662f\u6307\u5927\u8111\u4f7f\u7528\u4e00\u79cd\u611f\u77e5\u65b9\u5f0f<\/span>\uff08\u5982\u89e6\u6478\uff09<\/span>\u6765\u63d0\u4f9b\u901a\u5e38\u7531\u53e6\u4e00\u79cd\u611f\u77e5<\/span>\uff08\u5982\u89c6\u89c9\uff09<\/span>\u6536\u96c6\u7684\u73af\u5883\u4fe1\u606f\u7684\u80fd\u529b\u3002\u7136\u800c\uff0c\u5927\u591a\u6570\u795e\u7ecf\u7f51\u7edc\u5e76\u4e0d\u80fd\u9002\u5e94\u611f\u77e5\u66ff\u4ee3\u3002\u4f8b\u5982\uff0c\u5927\u591a\u6570\u5f3a\u5316\u5b66\u4e60\u4e3b\u4f53\u8981\u6c42\u5b83\u4eec\u7684\u8f93\u5165\u5fc5\u987b\u662f\u7cbe\u786e\u7684\u3001\u9884\u5148\u6307\u5b9a\u7684\u683c\u5f0f\uff0c\u5426\u5219\u5b83\u4eec\u5c31\u4f1a\u5931\u8d25\u3002\u5728\u6700\u8fd1\u7684\u4e00\u9879\u5de5\u4f5c\u4e2d\uff0cTang\u548cHa<\/span>\uff082021\uff09<\/span>\u63a2\u7d22\u4e86\u6392\u5217\u4e0d\u53d8\u7684\u795e\u7ecf\u7f51\u7edc\u4e3b\u4f53\uff0c\u8981\u6c42\u5176\u6bcf\u4e2a\u611f\u77e5\u795e\u7ecf\u5143<\/span>\uff08\u4ece\u73af\u5883\u4e2d\u63a5\u6536\u611f\u89c9\u8f93\u5165\u7684\u611f\u53d7\u5668\uff09<\/span>\u63a8\u65ad\u5176\u8f93\u5165\u4fe1\u53f7\u7684\u610f\u4e49\u548c\u80cc\u666f\uff0c\u800c\u4e0d\u662f\u660e\u786e\u5730\u5047\u8bbe\u4e00\u4e2a\u56fa\u5b9a\u7684\u610f\u4e49\u3002\u4ed6\u4eec\u8bc1\u660e\uff0c\u8fd9\u4e9b\u611f\u77e5\u7f51\u7edc\u53ef\u4ee5\u88ab\u8bad\u7ec3\u6765\u6574\u5408\u5c40\u90e8\u6536\u5230\u7684\u4fe1\u606f\uff0c\u5e76\u4e14\u4f7f\u7528\u6ce8\u610f\u529b\u673a\u5236\u7684\u4ea4\u6d41\uff0c\u53ef\u4ee5\u96c6\u4f53\u4ea7\u751f\u4e00\u4e2a\u5168\u5c40\u4e00\u81f4\u7684\u653f\u7b56\u3002\u6b64\u5916\uff0c\u5373\u4f7f\u5176\u611f\u77e5\u8f93\u5165<\/span>\uff08\u4ee5\u5b9e\u503c\u6570\u5b57\u8868\u793a\uff09<\/span>\u7684\u987a\u5e8f\u5728\u4e00\u4e2a\u4e8b\u4ef6\u4e2d\u88ab\u968f\u673a\u5730\u6539\u53d8\u4e86\u51e0\u6b21\uff0c\u7cfb\u7edf\u4ecd\u7136\u53ef\u4ee5\u6267\u884c\u5176\u4efb\u52a1\u3002\u4ed6\u4eec\u7684\u5b9e\u9a8c\u8868\u660e\uff0c\u8fd9\u79cd\u4e3b\u4f53\u5bf9\u5305\u542b\u8bb8\u591a\u989d\u5916\u7684\u5197\u4f59\u6216\u566a\u58f0\u4fe1\u606f\u7684\u89c2\u5bdf\uff0c\u6216\u5bf9\u635f\u574f\u548c\u4e0d\u5b8c\u6574\u7684\u89c2\u5bdf\u7ed3\u679c\u5177\u6709\u7a33\u5065\u6027\u3002<\/span><\/section>\n

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

\u56fe10. \u5229\u7528\u81ea\u7ec4\u7ec7\u548c\u6ce8\u610f\u529b\u7684\u7279\u6027\uff0c\uff08Tang and Ha, 2021\uff09\u4e2d\u7684\u4f5c\u8005\u7814\u7a76\u4e86\u5f3a\u5316\u5b66\u4e60\u4e3b\u4f53\uff0c\u5b83\u4eec\u5c06\u5176\u89c2\u5bdf\u7ed3\u679c\u89c6\u4e3a\u4efb\u610f\u6709\u5e8f\u3001\u957f\u5ea6\u53ef\u53d8\u7684\u611f\u5b98\u8f93\u5165\u5217\u8868\u3002\u4ed6\u4eec\u5c06CarRacing \u548c Atari Pong\uff08Brockman et al. 2016; Tang et al. 2020\uff09\u7b49\u89c6\u89c9\u4efb\u52a1\u4e2d\u7684\u8f93\u5165\u5212\u5206\u4e3a\u5c0f\u6591\u5757\u7684\u4e8c\u7ef4\u7f51\u683c\uff0c\u5e76\u6253\u4e71\u4e86\u5b83\u4eec\u7684\u987a\u5e8f\uff08\u5de6\uff09\u3002\u4ed6\u4eec\u8fd8\u5728\u8fde\u7eed\u63a7\u5236\u4efb\u52a1\u4e2d\u589e\u52a0\u4e86\u8bb8\u591a\u989d\u5916\u7684\u5197\u4f59\u566a\u58f0\u8f93\u5165\u901a\u9053\uff08Freeman et al. 2019\uff09\uff0c\u4ee5\u6253\u4e71\u987a\u5e8f\uff08\u53f3\uff09\uff0c\u4e3b\u4f53\u5fc5\u987b\u5b66\u4e60\u8bc6\u522b\u54ea\u4e9b\u8f93\u5165\u662f\u6709\u7528\u7684\u3002\u7cfb\u7edf\u4e2d\u7684\u6bcf\u4e2a\u611f\u89c9\u795e\u7ecf\u5143\u90fd\u6536\u5230\u4e00\u4e2a\u7279\u5b9a\u7684\u8f93\u5165\u6d41\uff0c\u901a\u8fc7\u534f\u8c03\u5b8c\u6210\u624b\u5934\u7684\u4efb\u52a1\u3002<\/span><\/p>\n


<\/section>\n
4.3 \u591a\u4e3b\u4f53\u5b66\u4e60<\/span><\/strong><\/span><\/section>\n

<\/section>\n
\u96c6\u4f53\u667a\u80fd\u53ef\u4ee5\u4ece\u51e0\u4e2a\u4e0d\u540c\u7684\u5c3a\u5ea6\u6765\u770b\u5f85\u3002\u5927\u8111\u53ef\u4ee5\u88ab\u770b\u4f5c\u662f\u4e00\u4e2a\u7531\u5355\u4e2a\u795e\u7ecf\u5143\u96c6\u4f53\u8fd0\u4f5c\u7684\u7f51\u7edc\u3002\u6bcf\u4e2a\u5668\u5b98\u53ef\u4ee5\u88ab\u770b\u4f5c\u662f\u6267\u884c\u96c6\u4f53\u529f\u80fd\u7684\u7ec6\u80de\u7684\u96c6\u5408\u3002\u5355\u4e2a\u52a8\u7269\u53ef\u4ee5\u88ab\u770b\u4f5c\u662f\u5171\u540c\u5de5\u4f5c\u7684\u5668\u5b98\u7684\u96c6\u5408\u3002\u5f53\u6211\u4eec\u8fdb\u4e00\u6b65\u653e\u5927\u65f6\uff0c\u6211\u4eec\u4e5f\u53ef\u4ee5\u8d85\u8d8a\u751f\u7269\u5b66\u6765\u770b\u5f85\u4eba\u7c7b\u7684\u667a\u6167\uff0c\u5e76\u5c06\u4eba\u7c7b\u6587\u660e\u89c6\u4e3a\u89e3\u51b3<\/span>\uff08\u548c\u4ea7\u751f\uff09<\/span>\u8d85\u51fa\u5355\u4e2a\u4e2a\u4f53\u80fd\u529b\u7684\u95ee\u9898\u7684\u96c6\u4f53\u667a\u80fd\u3002\u56e0\u6b64\uff0c\u867d\u7136\u5728\u4e0a\u4e00\u8282\u4e2d\uff0c\u6211\u4eec\u8ba8\u8bba\u4e86\u51e0\u9879\u5de5\u4f5c\uff0c\u5b83\u4eec\u5229\u7528\u96c6\u4f53\u667a\u80fd\u7684\u529b\u91cf\uff0c\u5c06\u4e00\u4e2a\u5355\u4e00\u7684\u5f3a\u5316\u5b66\u4e60\u4e3b\u4f53\u57fa\u672c\u4e0a\u5206\u89e3\u6210\u4e00\u4e2a\u8f83\u5c0f\u7684\u5f3a\u5316\u5b66\u4e60\u4e3b\u4f53\u96c6\u5408\uff0c\u5171\u540c\u4e3a\u4e00\u4e2a\u96c6\u4f53\u76ee\u6807\u5de5\u4f5c\uff0c\u7c7b\u4f3c\u4e8e\u751f\u7269\u5c42\u9762\u7684\u96c6\u4f53\u667a\u80fd\u6a21\u578b\uff0c\u4f46\u6211\u4eec\u4e5f\u53ef\u4ee5\u5c06\u591a\u4e3b\u4f53\u95ee\u9898\u89c6\u4e3a\u793e\u4f1a\u5c42\u9762\u4e0a\u7684\u96c6\u4f53\u667a\u80fd\u6a21\u578b\u3002<\/span><\/section>\n

<\/section>\n
\u96c6\u4f53\u667a\u80fd\u9886\u57df\u7684\u4e00\u4e2a\u4e3b\u8981\u7126\u70b9\u662f\u7814\u7a76\u5927\u91cf\u4e2a\u4f53\u96c6\u5408\u6240\u4ea7\u751f\u7684\u7fa4\u4f53\u667a\u80fd\u548c\u884c\u4e3a\uff0c\u65e0\u8bba\u662f\u4eba\u7c7b<\/span>\uff08Tapscott and Williams\uff0c2008\uff09<\/span>\u3001\u52a8\u7269<\/span>\uff08Sumpter\uff0c2010\uff09<\/span>\u3001\u6606\u866b<\/span>\uff08Dorigo et al. 2000; Seeley, 2010\uff09<\/span>\uff0c\u8fd8\u662f\u4eba\u5de5\u8702\u7fa4\u673a\u5668\u4eba<\/span>\uff08Hamann\uff0c2018\uff1bRubenstein et al. 2014\uff09<\/span>\u3002\u5728\u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60\u9886\u57df\uff0c\u663e\u7136\u7f3a\u5c11\u8fd9\u79cd\u5173\u6ce8\u3002\u867d\u7136\u591a\u4e3b\u4f53\u5f3a\u5316\u5b66\u4e60<\/span>\uff08multi-agent reinforcement learning, MARL\uff09<\/span>\u662f\u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60\u7684\u4e00\u4e2a\u6210\u719f\u7684\u5206\u652f\uff0c\u4f46\u63d0\u51fa\u7684\u5927\u591a\u6570\u5b66\u4e60\u7b97\u6cd5\u548c\u73af\u5883\u90fd\u9488\u5bf9\u76f8\u5bf9\u8f83\u5c11\u7684\u4e3b\u4f53<\/span>\uff08Foerster et al. 2016; OroojlooyJadid and Hajinezhad, 2019\uff09<\/span>\uff0c\u56e0\u6b64\u4e0d\u8db3\u4ee5\u7814\u7a76\u89c4\u6a21\u5e9e\u5927\u7fa4\u4f53\u7684\u6d8c\u73b0\u7279\u6027\u3002\u5728\u6700\u5e38\u89c1\u7684\u591a\u4e3b\u4f53\u5f3a\u5316\u5b66\u4e60\u73af\u5883\u4e2d<\/span>\uff08Resnick et al. 2018; Baker et al. 2019; Jaderberg et al. 2019; Terry et al. 2020\uff09<\/span>\uff0c\u201c\u591a\u4e3b\u4f53\u201d\u4ec5\u4ec5\u610f\u5473\u7740\u4e24\u4e2a\u6216\u56db\u4e2a\u4e3b\u4f53\u7ecf\u8fc7\u8bad\u7ec3\uff0c\u901a\u8fc7\u81ea\u6211\u6e38\u620f\u7684\u65b9\u5f0f\u6267\u884c\u4efb\u52a1<\/span>\uff08Bansal et al. 2017; Liu et al. 2019; Ha, 2020\uff09<\/span>\u3002\u7136\u800c\uff0c\u5728\u81ea\u7136\u754c\u6216\u793e\u4f1a\u4e2d\u89c2\u5bdf\u5230\u7684\u96c6\u4f53\u667a\u80fd\u4f9d\u8d56\u4e8e\u6bd4\u591a\u4e3b\u4f53\u5f3a\u5316\u5b66\u4e60\u4e2d\u901a\u5e38\u7814\u7a76\u7684\u66f4\u591a\u6570\u91cf\u7684\u4e2a\u4f53\uff0c\u6d89\u53ca\u4ece\u6570\u5343\u5230\u767e\u4e07\u7684\u79cd\u7fa4\u89c4\u6a21\u3002\u5728\u672c\u8282\u4e2d\uff0c\u6211\u4eec\u5c06\u8ba8\u8bba\u6700\u8fd1\u6765\u81ea\u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60\u7684\u591a\u4e3b\u4f53\u5f3a\u5316\u5b66\u4e60\u5b50\u9886\u57df\u7684\u5de5\u4f5c\uff0c\u8fd9\u4e9b\u5de5\u4f5c\u53d7\u5230\u4e86\u96c6\u4f53\u667a\u80fd\u7684\u542f\u53d1<\/span>\uff08\u6b63\u5982\u4ed6\u4eec\u7684\u4f5c\u8005\u751a\u81f3\u5728\u4ed6\u4eec\u7684\u51fa\u7248\u7269\u4e2d\u6307\u51fa\u7684\uff09<\/span>\u3002\u4e0e\u5927\u591a\u6570\u591a\u4e3b\u4f53\u5f3a\u5316\u5b66\u4e60\u5de5\u4f5c\u4e0d\u540c\u7684\u662f\uff0c\u8fd9\u4e9b\u5de5\u4f5c\u5f00\u59cb\u91c7\u7528\u5927\u578b\u7684\u4e3b\u4f53\u7fa4\u4f53<\/span>\uff08\u6bcf\u4e2a\u90fd\u7531\u795e\u7ecf\u7f51\u7edc\u5b9e\u73b0\uff09<\/span>\uff0c\u4ece\u6570\u5343\u5230\u6570\u767e\u4e07\uff0c\u4ee5\u4fbf\u771f\u6b63\u5728\u5b8f\u89c2\u5c42\u9762<\/span>\uff081000\u591a\u4e2a\u4e3b\u4f53\uff09<\/span>\uff0c\u800c\u4e0d\u662f\u5728\u5fae\u89c2\u5c42\u9762<\/span>\uff082-4\u4e2a\u4e3b\u4f53\uff09<\/span>\u7814\u7a76\u5b83\u4eec\u7684\u6d8c\u73b0\u7279\u6027\u3002<\/span><\/section>\n

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

\u56fe11. MAgent\uff08Zheng et al. 2018\uff09\u662f\u4e00\u5957\u73af\u5883\u6846\u67b6\uff0c\u5728\u7f51\u683c\u4e16\u754c\u4e2d\u5927\u91cf\u7684\u50cf\u7d20\u4e3b\u4f53\u5728\u6218\u6597\u6216\u5176\u4ed6\u7ade\u4e89\u573a\u666f\u4e2d\u8fdb\u884c\u4e92\u52a8\u3002\u4e0e\u5927\u591a\u6570\u4e13\u6ce8\u4e8e\u5355\u4e00\u4e3b\u4f53\u6216\u53ea\u6709\u5c11\u6570\u4e3b\u4f53\u7684\u5f3a\u5316\u5b66\u4e60\u7814\u7a76\u7684\u5e73\u53f0\u4e0d\u540c\uff0c\u4ed6\u4eec\u7684\u76ee\u6807\u662f\u652f\u6301\u6269\u5c55\u5230\u6570\u767e\u4e07\u4e3b\u4f53\u7684\u5f3a\u5316\u5b66\u4e60\u7814\u7a76\u3002\u8fd9\u4e2a\u5e73\u53f0\u7684\u73af\u5883\u73b0\u5728\u4f5c\u4e3aPettingZoo\uff08Terry et al. 2020\uff09\u5f00\u6e90\u5e93\u7684\u4e00\u90e8\u5206\u8fdb\u884c\u7ef4\u62a4\uff0c\u7528\u4e8e\u591a\u4e3b\u4f53\u5f3a\u5316\u5b66\u4e60\u7814\u7a76\u3002<\/span><\/p>\n


<\/section>\n
\u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60\u7684\u6700\u65b0\u8fdb\u5c55\u5df2\u7ecf\u8bc1\u660e\u4e86\u5728\u590d\u6742\u7684\u4e09\u7ef4\u6a21\u62df\u73af\u5883\u4e2d\u53ea\u7528\u4e00\u4e2aGPU\u5c31\u80fd\u6a21\u62df\u6210\u5343\u4e0a\u4e07\u7684\u4e3b\u4f53<\/span>\uff08Heiden et al. 2021; Rudin et al. 2021\uff09<\/span>\u3002\u4e00\u4e2a\u5173\u952e\u7684\u6311\u6218\u662f\u5728\u66f4\u5927\u7684\u89c4\u6a21\u4e0a\u63a5\u8fd1\u591a\u4e3b\u4f53\u5b66\u4e60\u7684\u95ee\u9898\uff0c\u5229\u7528\u5e76\u884c\u8ba1\u7b97\u786c\u4ef6\u548c\u5206\u5e03\u5f0f\u8ba1\u7b97\u7684\u8fdb\u5c55\uff0c\u4ee5\u8bad\u7ec3\u6570\u767e\u4e07\u4e3b\u4f53\u4e3a\u76ee\u6807\u3002\u5728\u672c\u8282\u4e2d\uff0c\u6211\u4eec\u5c06\u4e3e\u4f8b\u8bf4\u660e\u6700\u8fd1\u5728\u8bad\u7ec3\u5927\u89c4\u6a21\u5728\u96c6\u4f53\u73af\u5883\u4e2d\u4e92\u52a8\u7684\u4e3b\u4f53\u65b9\u9762\u7684\u5c1d\u8bd5\u3002<\/span><\/section>\n

<\/section>\n
Zheng\u7b49\u4eba<\/span>\uff082018\uff09<\/span>\u6ca1\u6709\u5173\u6ce8\u73b0\u5b9e\u7684\u7269\u7406\u5b66\u6216\u73af\u5883\u7684\u771f\u5b9e\u6027\uff0c\u800c\u662f\u5f00\u53d1\u4e86\u4e00\u4e2a\u540d\u4e3aMAgent\u7684\u5e73\u53f0\uff0c\u4e00\u4e2a\u7b80\u5355\u7684\u7f51\u683c\u4e16\u754c\u73af\u5883\uff0c\u53ef\u4ee5\u5141\u8bb8\u6570\u767e\u4e07\u7684\u795e\u7ecf\u7f51\u7edc\u4e3b\u4f53\u3002\u4ed6\u4eec\u7684\u91cd\u70b9\u662f\u53ef\u6269\u5c55\u6027\uff0c\u4ed6\u4eec\u8bc1\u660eMAgent\u53ef\u4ee5\u5728\u4e00\u4e2aGPU\u4e0a\u627f\u8f7d\u591a\u8fbe\u4e00\u767e\u4e07\u4e2a\u4e3b\u4f53<\/span>\uff082017\uff09<\/span>\u3002\u4ed6\u4eec\u7684\u5e73\u53f0\u652f\u6301\u4e3b\u4f53\u7fa4\u4f53\u4e4b\u95f4\u7684\u4e92\u52a8\uff0c\u4e0d\u4ec5\u6709\u5229\u4e8e\u7814\u7a76\u653f\u7b56\u4f18\u5316\u7684\u5b66\u4e60\u7b97\u6cd5\uff0c\u66f4\u5173\u952e\u7684\u662f\uff0c\u80fd\u591f\u7814\u7a76\u4eba\u5de5\u667a\u80fd\u793e\u4f1a\u4e2d\u6570\u767e\u4e07\u4e3b\u4f53\u6d8c\u73b0\u7684\u793e\u4f1a\u73b0\u8c61\uff0c\u5305\u62ec\u53ef\u80fd\u51fa\u73b0\u7684\u8bed\u8a00\u548c\u793e\u4f1a\u7b49\u7ea7\u7ed3\u6784\u7684\u51fa\u73b0\u3002\u73af\u5883\u53ef\u4ee5\u7528\u811a\u672c\u5efa\u7acb\uff0c\u5b83\u4eec\u63d0\u4f9b\u4e86\u4e00\u4e9b\u4f8b\u5b50\uff0c\u5982\u6355\u98df\u8005-\u730e\u7269\u6a21\u62df\u3001\u6218\u573a\u3001\u5bf9\u6297\u6027\u8ffd\u51fb\uff0c\u652f\u6301\u4e0d\u540c\u79cd\u7c7b\u7684\u4e0d\u540c\u4e3b\u4f53\uff0c\u53ef\u80fd\u8868\u73b0\u51fa\u4e0d\u540c\u7684\u884c\u4e3a\u3002<\/span><\/section>\n

<\/section>\n
MAgent\u542f\u53d1\u4e86\u8bb8\u591a\u6700\u8fd1\u7684\u5e94\u7528\uff0c\u5305\u62ec\u591a\u4e3b\u4f53\u9a7e\u9a76<\/span>\uff08Peng et al. 2021\uff09<\/span>\uff0c\u5b83\u7740\u773c\u4e8e\u6574\u4e2a\u9a7e\u9a76\u4e3b\u4f53\u7fa4\u4f53\u7684\u6d8c\u73b0\u884c\u4e3a\uff0c\u4ee5\u4f18\u5316\u9a7e\u9a76\u653f\u7b56\uff0c\u8fd9\u4e0d\u4ec5\u5f71\u54cd\u5230\u4e00\u8f86\u8f66\uff0c\u800c\u4e14\u65e8\u5728\u63d0\u9ad8\u6574\u4e2a\u7fa4\u4f53\u7684\u5b89\u5168\u6027\u3002\u8fd9\u4e9b\u65b9\u5411\u90fd\u662f\u5f88\u597d\u7684\u4f8b\u5b50\uff0c\u8bc1\u660e\u4e86\u6df1\u5ea6\u5b66\u4e60\u7684\u95ee\u9898<\/span>\uff08\u4e3a\u5355\u4e2a\u4e2a\u4f53\u5bfb\u627e\u9a7e\u9a76\u653f\u7b56\uff09<\/span>\u4e0e\u96c6\u4f53\u667a\u80fd\u7684\u95ee\u9898<\/span>\uff08\u4e3a\u6574\u4e2a\u7fa4\u4f53\u5bfb\u627e\u9a7e\u9a76\u653f\u7b56\uff09<\/span>\u4e4b\u95f4\u7684\u533a\u522b<\/span>\uff08\u89c1\u56fe12\uff09<\/span>\u3002<\/span><\/section>\n

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

\u56fe12. Neural MMO\uff08Suarez et al. 2021\uff09\u662f\u4e00\u4e2a\u5728\u7a0b\u5e8f\u5316\u751f\u6210\u7684\u865a\u62df\u4e16\u754c\u4e2d\u6a21\u62df\u4e3b\u4f53\u7fa4\u4f53\u7684\u5e73\u53f0\uff0c\u4ee5\u652f\u6301\u591a\u4e3b\u4f53\u7814\u7a76\uff0c\u540c\u65f6\u4fdd\u6301\u5176\u5728\u8ba1\u7b97\u4e0a\u7684\u8981\u6c42\u3002\u7528\u6237\u4ece\u4e00\u7ec4\u63d0\u4f9b\u7684\u6e38\u620f\u7cfb\u7edf\u4e2d\u9009\u62e9\uff0c\u4e3a\u4ed6\u4eec\u7684\u5177\u4f53\u7814\u7a76\u95ee\u9898\u521b\u9020\u73af\u5883–\u652f\u6301\u591a\u8fbe\u4e00\u5343\u79cd\u4e3b\u4f53\u548c\u51e0\u5343\u4e2a\u65f6\u95f4\u6b65\u957f\u7684\u4e00\u5e73\u65b9\u516c\u91cc\u7684\u5730\u56fe\u3002\u8be5\u9879\u76ee\u6b63\u5728\u79ef\u6781\u5f00\u53d1\u4e2d\uff0c\u6709\u5927\u91cf\u7684\u6587\u6863\u548c\u5de5\u5177\uff0c\u4e3a\u7814\u7a76\u4eba\u5458\u63d0\u4f9b\u8bb0\u5f55\u548c\u53ef\u89c6\u5316\u5de5\u5177\u3002\u622a\u81f3\u53d1\u7a3f\u65f6\uff0c\u8fd9\u4e2a\u5e73\u53f0\u5c06\u57282021\u5e74\u7684Neur IPS\u4f1a\u8bae\u4e0a\u8fdb\u884c\u6f14\u793a\u3002<\/span><\/p>\n


<\/section>\n
\u53d7MMORPG<\/span>\uff08\u5927\u578b\u591a\u4eba\u5728\u7ebf\u89d2\u8272\u626e\u6f14\u6e38\u620f\uff0c\u53c8\u79f0MMO\uff09<\/span>\u6e38\u620f\u7c7b\u578b\u7684\u542f\u53d1\uff0cNeural MMO<\/span>\uff08Suarez et al. 2021\uff09<\/span>\u662f\u4e00\u4e2a\u4eba\u5de5\u667a\u80fd\u7814\u7a76\u73af\u5883\uff0c\u5b83\u652f\u6301\u8bbe\u7f6e\u5927\u91cf\u7684\u4eba\u5de5\u4e3b\u4f53\uff0c\u4e3a\u4e86\u751f\u5b58\u800c\u5fc5\u987b\u7ade\u4e89\u6709\u9650\u7684\u8d44\u6e90\u3002\u56e0\u6b64\uff0c\u4ed6\u4eec\u7684\u73af\u5883\u80fd\u591f\u5927\u89c4\u6a21\u6a21\u62df\u591a\u4e3b\u4f53\u7684\u4e92\u52a8\uff0c\u8981\u6c42\u4e3b\u4f53\u4e0e\u5927\u7fa4\u4e2d\u7684\u5176\u4ed6\u4e3b\u4f53\u4e00\u8d77\u5b66\u4e60\u6218\u6597\u548c\u5bfc\u822a\u653f\u7b56\uff0c\u8fd9\u4e9b\u4e3b\u4f53\u90fd\u8bd5\u56fe\u505a\u540c\u6837\u7684\u4e8b\u60c5\u3002\u4e0e\u5927\u591a\u6570\u591a\u4e3b\u4f53\u5f3a\u5316\u5b66\u4e60\u73af\u5883\u4e0d\u540c\u7684\u662f\uff0c\u5141\u8bb8\u6bcf\u4e2a\u4e3b\u4f53\u62e5\u6709\u81ea\u5df1\u72ec\u7279\u7684\u795e\u7ecf\u7f51\u7edc\u6743\u91cd\u96c6\uff0c\u8fd9\u5728\u5185\u5b58\u6d88\u8017\u65b9\u9762\u662f\u4e00\u4e2a\u6280\u672f\u6311\u6218\u3002\u8be5\u5e73\u53f0\u65e9\u671f\u7248\u672c\u7684\u521d\u6b65\u5b9e\u9a8c\u7ed3\u679c<\/span>\uff08Suarez et al. 2019\u5e74\uff09<\/span>\u8868\u660e\uff0c\u5177\u6709\u4e0d\u540c\u795e\u7ecf\u7f51\u7edc\u6743\u91cd\u53c2\u6570\u7684\u4e3b\u4f53\u5f00\u53d1\u4e86\u586b\u8865\u4e0d\u540c\u9886\u57df\u7684\u6280\u80fd\uff0c\u4ee5\u907f\u514d\u5728\u5927\u91cf\u4e3b\u4f53\u7fa4\u4f53\u4e2d\u7684\u7ade\u4e89\u3002<\/span><\/section>\n

<\/section>\n
\u622a\u81f3\u53d1\u7a3f\u65f6\uff0c\u8be5\u9879\u76ee\u6b63\u5728NeurIPS\u673a\u5668\u5b66\u4e60\u793e\u533a\u4e2d\u79ef\u6781\u5f00\u53d1<\/span>\uff08Suarez et al. 2021\uff09<\/span>\uff0c\u81f4\u529b\u4e8e\u7814\u7a76\u5927\u578b\u4e3b\u4f53\u7fa4\u4f53\u3001\u957f\u65f6\u95f4\u8303\u56f4\u3001\u5f00\u653e\u5f0f\u4efb\u52a1\u548c\u6a21\u5757\u5316\u6e38\u620f\u7cfb\u7edf\u3002\u5f00\u53d1\u8005\u63d0\u4f9b\u4e86\u79ef\u6781\u7684\u652f\u6301\u548c\u6587\u6863\uff0c\u8fd8\u5f00\u53d1\u4e86\u989d\u5916\u7684\u8bad\u7ec3\u3001\u8bb0\u5f55\u548c\u53ef\u89c6\u5316\u5de5\u5177\uff0c\u4ee5\u5b9e\u73b0\u8fd9\u4e00\u5927\u89c4\u6a21\u591a\u4e3b\u4f53\u7814\u7a76\u7684\u8def\u7ebf\u3002\u8fd9\u9879\u5de5\u4f5c\u4ecd\u5904\u4e8e\u65e9\u671f\u9636\u6bb5\uff0c\u53ea\u6709\u65f6\u95f4\u624d\u80fd\u544a\u8bc9\u6211\u4eec\uff0c\u50cfNeural MMO\u6216MAgent\u8fd9\u6837\u80fd\u591f\u7814\u7a76\u5927\u578b\u7fa4\u4f53\u7684\u5e73\u53f0\u662f\u5426\u5728\u6df1\u5ea6\u5b66\u4e60\u793e\u533a\u4e2d\u83b7\u5f97\u8fdb\u4e00\u6b65\u7684\u5438\u5f15\u529b\u3002<\/span><\/section>\n

<\/section>\n
4.4 \u5143\u5b66\u4e60<\/span><\/strong><\/span><\/section>\n

<\/section>\n
\u5728\u524d\u9762\u7684\u7ae0\u8282\u4e2d\uff0c\u6211\u4eec\u63cf\u8ff0\u4e86\u7528\u72ec\u7acb\u7684\u795e\u7ecf\u7f51\u7edc\u4e3b\u4f53\u7684\u96c6\u5408\u6765\u8868\u8fbe\u5bf9\u95ee\u9898\u7684\u89e3\u51b3\uff0c\u4ee5\u5b9e\u73b0\u4e00\u4e2a\u5171\u540c\u7684\u76ee\u6807\u3002\u8fd9\u4e9b\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\u7684\u53c2\u6570\u88ab\u4f18\u5316\u4e3a\u7fa4\u4f53\u7684\u96c6\u4f53\u8868\u73b0\u3002\u867d\u7136\u8fd9\u4e9b\u7cfb\u7edf\u5df2\u88ab\u8bc1\u660e\u662f\u7a33\u5065\u7684\uff0c\u5e76\u80fd\u9002\u5e94\u5176\u73af\u5883\u7684\u53d8\u5316\uff0c\u4f46\u5b83\u4eec\u6700\u7ec8\u88ab\u786c\u6027\u89c4\u5b9a\u4e3a\u6267\u884c\u67d0\u9879\u4efb\u52a1\uff0c\u9664\u975e\u4ece\u5934\u5f00\u59cb\u91cd\u65b0\u8bad\u7ec3\uff0c\u5426\u5219\u65e0\u6cd5\u6267\u884c\u53e6\u4e00\u9879\u4efb\u52a1\u3002<\/span><\/section>\n

<\/section>\n
\u5143\u5b66\u4e60\u662f\u6df1\u5ea6\u5b66\u4e60\u4e2d\u4e00\u4e2a\u6d3b\u8dc3\u7684\u7814\u7a76\u9886\u57df\uff0c\u5176\u76ee\u6807\u662f\u8bad\u7ec3\u7cfb\u7edf\u7684\u5b66\u4e60\u80fd\u529b\u3002\u5b83\u662f\u673a\u5668\u5b66\u4e60\u7684\u4e00\u4e2a\u5927\u578b\u5b50\u9886\u57df\uff0c\u5305\u62ec\u4ece\u4e00\u4e2a\u8bad\u7ec3\u96c6\u5230\u53e6\u4e00\u4e2a\u8bad\u7ec3\u96c6\u7684\u7b80\u5355\u8fc1\u79fb\u5b66\u4e60\u7b49\u9886\u57df\u3002\u4e3a\u4e86\u6211\u4eec\u7684\u76ee\u7684\uff0c\u6211\u4eec\u9075\u5faaSchmidhuber<\/span>\uff08Schmidhuber, 2020\uff09<\/span>\u7684\u5de5\u4f5c\u601d\u8def\uff0c\u4ed6\u5c06\u5143\u5b66\u4e60\u89c6\u4e3a\u53ef\u4ee5\u5b66\u4e60\u66f4\u597d\u7684\u673a\u5668\u5b66\u4e60\u7b97\u6cd5\u7684\u95ee\u9898\uff0c\u4ed6\u8ba4\u4e3a\u8fd9\u662f\u5efa\u7acb\u771f\u6b63\u7684\u81ea\u6211\u6539\u8fdb\u7684AI\u7cfb\u7edf\u6240\u9700\u8981\u7684\u3002<\/span><\/section>\n

<\/section>\n
\u56e0\u6b64\uff0c\u4e0e\u4f20\u7edf\u4e0a\u8bad\u7ec3\u4e00\u4e2a\u795e\u7ecf\u7f51\u7edc\u6765\u5b8c\u6210\u4e00\u9879\u4efb\u52a1\u4e0d\u540c\uff0c\u4f20\u7edf\u4e0a\u795e\u7ecf\u7f51\u7edc\u7684\u6743\u91cd\u53c2\u6570\u662f\u7528\u68af\u5ea6\u4e0b\u964d\u7b97\u6cd5\u6765\u4f18\u5316\u7684\uff0c\u6216\u8005\u7528\u8fdb\u5316\u7b56\u7565<\/span>\uff08Tang et al. 2022\uff09<\/span>\uff0c\u5143\u5b66\u4e60\u7684\u76ee\u6807\u662f\u8bad\u7ec3\u4e00\u4e2a\u5143\u5b66\u4e60\u8005<\/span>\uff08\u53ef\u4ee5\u662f\u53e6\u4e00\u4e2a\u57fa\u4e8e\u795e\u7ecf\u7f51\u7edc\u7684\u7cfb\u7edf\uff09<\/span>\u6765\u5b66\u4e60\u4e00\u79cd\u5b66\u4e60\u7b97\u6cd5\u3002\u8fd9\u662f\u4e00\u9879\u7279\u522b\u5177\u6709\u6311\u6218\u6027\u7684\u4efb\u52a1\uff0c\u5176\u5386\u53f2\u60a0\u4e45\uff0c\u89c1Schmidhuber\u7684\u8bc4\u8bba<\/span>\uff08Schmidhuber, 2020\uff09<\/span>\u3002\u5728\u8fd9\u4e00\u8282\u4e2d\uff0c\u6211\u4eec\u5c06\u5f3a\u8c03\u6700\u8fd1\u6709\u5e0c\u671b\u7684\u5de5\u4f5c\uff0c\u8fd9\u4e9b\u5de5\u4f5c\u5229\u7528\u96c6\u4f53\u4e3b\u4f53\uff0c\u53ef\u4ee5\u5b66\u4f1a\u5b66\u4e60\uff0c\u800c\u4e0d\u662f\u53ea\u5b66\u4f1a\u6267\u884c\u4e00\u4e2a\u7279\u5b9a\u7684\u4efb\u52a1<\/span>\uff08\u6211\u4eec\u5728\u4e0a\u4e00\u8282\u4e2d\u5df2\u7ecf\u4ecb\u7ecd\u4e86\uff09<\/span>\u3002<\/span><\/section>\n

<\/section>\n
\u81ea\u7ec4\u7ec7\u7684\u6982\u5ff5\u53ef\u4ee5\u81ea\u7136\u5730\u5e94\u7528\u4e8e\u8bad\u7ec3\u795e\u7ecf\u7f51\u7edc\uff0c\u901a\u8fc7\u6269\u5c55\u6784\u6210\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\u7684\u57fa\u672c\u6784\u4ef6\u6765\u8fdb\u884c\u5143\u5b66\u4e60\u3002\u6b63\u5982\u6211\u4eec\u6240\u77e5\uff0c\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\u7531\u76f8\u540c\u7684\u795e\u7ecf\u5143\u7ec4\u6210\uff0c\u8fd9\u4e9b\u795e\u7ecf\u5143\u88ab\u5efa\u6a21\u4e3a\u975e\u7ebf\u6027\u6fc0\u6d3b\u51fd\u6570\u3002\u8fd9\u4e9b\u795e\u7ecf\u5143\u5728\u7f51\u7edc\u4e2d\u901a\u8fc7\u7a81\u89e6\u8fde\u63a5\uff0c\u7a81\u89e6\u662f\u6743\u91cd\u53c2\u6570\uff0c\u901a\u5e38\u7528\u68af\u5ea6\u4e0b\u964d\u7b49\u5b66\u4e60\u7b97\u6cd5\u6765\u8bad\u7ec3\u3002\u4f46\u662f\u53ef\u4ee5\u8003\u8651\u5c06\u795e\u7ecf\u5143\u548c\u7a81\u89e6\u7684\u62bd\u8c61\u6269\u5c55\u5230\u9759\u6001\u6fc0\u6d3b\u51fd\u6570\u548c\u6d6e\u70b9\u53c2\u6570\u4e4b\u5916\u3002\u4e8b\u5b9e\u4e0a\uff0c\u6700\u8fd1\u7684\u5de5\u4f5c<\/span>\uff08Ohsawa et al. 2018; Ott, 2020\uff09<\/span>\u5df2\u7ecf\u63a2\u7d22\u4e86\u5c06\u795e\u7ecf\u7f51\u7edc\u7684\u6bcf\u4e2a\u795e\u7ecf\u5143\u5efa\u6a21\u4e3a\u4e00\u4e2a\u5355\u72ec\u7684\u5f3a\u5316\u5b66\u4e60\u4e3b\u4f53\u3002\u4f7f\u7528\u5f3a\u5316\u5b66\u4e60\u7684\u672f\u8bed\uff0c\u6bcf\u4e2a\u795e\u7ecf\u5143\u7684\u89c2\u5bdf\u662f\u5b83\u7684\u5f53\u524d\u72b6\u6001\uff0c\u5b83\u968f\u7740\u4fe1\u606f\u5728\u7f51\u7edc\u4e2d\u7684\u4f20\u8f93\u800c\u53d8\u5316\uff0c\u6bcf\u4e2a\u795e\u7ecf\u5143\u7684\u884c\u52a8\u4f7f\u6bcf\u4e2a\u795e\u7ecf\u5143\u80fd\u591f\u4fee\u6539\u5b83\u4e0e\u7cfb\u7edf\u4e2d\u5176\u4ed6\u795e\u7ecf\u5143\u7684\u8fde\u63a5\uff1b\u56e0\u6b64\uff0c\u5143\u5b66\u4e60\u7684\u95ee\u9898\u88ab\u89c6\u4e3a\u4e00\u4e2a\u591a\u4e3b\u4f53\u5f3a\u5316\u5b66\u4e60\u95ee\u9898\uff0c\u6bcf\u4e2a\u4e3b\u4f53\u662f\u795e\u7ecf\u7f51\u7edc\u4e2d\u795e\u7ecf\u5143\u96c6\u5408\u7684\u4e00\u90e8\u5206\u3002\u867d\u7136\u8fd9\u79cd\u65b9\u6cd5\u5f88\u4f18\u96c5\uff0c\u4f46\u4e0a\u8ff0\u4f5c\u54c1\u53ea\u80fd\u591f\u5b66\u4e60\u89e3\u51b3\u73a9\u5177\u95ee\u9898<\/span>\uff08toy problem\uff09<\/span>\uff0c\u8fd8\u4e0d\u80fd\u4e0e\u73b0\u6709\u7684\u5b66\u4e60\u7b97\u6cd5\u7ade\u4e89\u3002<\/span><\/section>\n

<\/section>\n
\u6700\u8fd1\u7684\u65b9\u6cd5\u5df2\u7ecf\u8d85\u8d8a\u4e86\u4f7f\u7528\u7b80\u5355\u7684\u6807\u91cf\u6743\u91cd\uff0c\u5728\u795e\u7ecf\u5143\u4e4b\u95f4\u4f20\u8f93\u6807\u91cf\u4fe1\u53f7\u3002Sandler\u7b49\u4eba<\/span>\uff082021\uff09<\/span>\u4ecb\u7ecd\u4e86\u4e00\u79cd\u65b0\u578b\u7684\u5e7f\u4e49\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\uff0c\u5176\u4e2d\u795e\u7ecf\u5143\u548c\u7a81\u89e6\u90fd\u6709\u591a\u79cd\u72b6\u6001\u3002\u4f20\u7edf\u7684\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\u53ef\u4ee5\u88ab\u770b\u4f5c\u662f\u5176\u6846\u67b6\u7684\u4e00\u4e2a\u7279\u4f8b\uff0c\u6709\u4e24\u4e2a\u72b6\u6001\uff0c\u4e00\u4e2a\u7528\u4e8e\u6fc0\u6d3b\uff0c\u53e6\u4e00\u4e2a\u7528\u4e8e\u4f7f\u7528\u53cd\u5411\u4f20\u64ad\u5b66\u4e60\u89c4\u5219\u4ea7\u751f\u7684\u68af\u5ea6\u3002\u5728\u4e00\u822c\u6846\u67b6\u4e2d\uff0c\u4ed6\u4eec\u4e0d\u9700\u8981\u53cd\u5411\u4f20\u64ad\u7a0b\u5e8f\u6765\u8ba1\u7b97\u4efb\u4f55\u68af\u5ea6\uff0c\u800c\u662f\u4f9d\u9760\u4e00\u4e2a\u5171\u4eab\u7684\u5c40\u90e8\u5b66\u4e60\u89c4\u5219\u6765\u66f4\u65b0\u7a81\u89e6\u548c\u795e\u7ecf\u5143\u7684\u72b6\u6001\u3002\u8fd9\u79cdHebbian\u5f0f\u7684\u53cc\u5411\u5c40\u90e8\u66f4\u65b0\u89c4\u5219\u53ea\u9700\u8981\u6bcf\u4e2a\u7a81\u89e6\u548c\u795e\u7ecf\u5143\u6536\u96c6\u81ea\u5176\u90bb\u8fd1\u7684\u7a81\u89e6\u548c\u795e\u7ecf\u5143\u7684\u72b6\u6001\u4fe1\u606f\uff0c\u7c7b\u4f3c\u4e8e\u5143\u80de\u81ea\u52a8\u673a\u3002\u8be5\u89c4\u5219\u88ab\u53c2\u6570\u5316\u4e3a\u4e00\u4e2a\u4f4e\u7ef4\u57fa\u56e0\u7ec4\u5411\u91cf\uff0c\u5e76\u4e14\u5728\u6574\u4e2a\u7cfb\u7edf\u4e2d\u662f\u4e00\u81f4\u7684\u3002\u4ed6\u4eec\u91c7\u7528\u4e86\u8fdb\u5316\u7b56\u7565\u6216\u4f20\u7edf\u7684\u4f18\u5316\u6280\u672f\u6765\u5143\u5b66\u4e60\u8fd9\u4e2a\u57fa\u56e0\u7ec4\u5411\u91cf\u3002\u4ed6\u4eec\u7684\u4e3b\u8981\u7ed3\u679c\u662f\uff0c\u5728\u8bad\u7ec3\u4efb\u52a1\u4e0a\u5143\u5b66\u4e60\u7684\u66f4\u65b0\u89c4\u5219\u53ef\u4ee5\u63a8\u5e7f\u5230\u672a\u77e5\u7684\u65b0\u6d4b\u8bd5\u4efb\u52a1\u3002\u6b64\u5916\uff0c\u5bf9\u4e8e\u51e0\u4e2a\u6807\u51c6\u7684\u5206\u7c7b\u4efb\u52a1\uff0c\u66f4\u65b0\u89c4\u5219\u6bd4\u57fa\u4e8e\u68af\u5ea6\u4e0b\u964d\u7684\u5b66\u4e60\u7b97\u6cd5\u8868\u73b0\u5f97\u66f4\u5feb\u3002<\/span><\/section>\n

<\/section>\n
Kirsch\u548cSchmidhuber<\/span>\uff082020\uff09<\/span>\u4e5f\u91c7\u53d6\u4e86\u7c7b\u4f3c\u7684\u65b9\u5411\uff0c\u795e\u7ecf\u7f51\u7edc\u7684\u795e\u7ecf\u5143\u548c\u7a81\u89e6\u4e5f\u88ab\u6cdb\u5316\u4e3a\u66f4\u9ad8\u7ef4\u5ea6\u7684\u4fe1\u606f\u4f20\u9012\u7cfb\u7edf\uff0c\u4f46\u5728\u4ed6\u4eec\u7684\u6848\u4f8b\u4e2d\uff0c\u6bcf\u4e2a\u7a81\u89e6\u90fd\u88ab\u4e00\u4e2a\u5177\u6709\u76f8\u540c\u5171\u4eab\u53c2\u6570\u7684\u5faa\u73af\u795e\u7ecf\u7f51\u7edc<\/span>\uff08RNN\uff09<\/span>\u53d6\u4ee3\u3002\u8fd9\u4e9b\u5faa\u73af\u795e\u7ecf\u7f51\u7edc\u7a81\u89e6\u662f\u53cc\u5411\u7684\uff0c\u652f\u914d\u7740\u6574\u4e2a\u7f51\u7edc\u7684\u4fe1\u606f\u6d41\u3002\u4e0eSandler\u7b49\u4eba<\/span>\uff082021\uff09<\/span>\u4e00\u6837\uff0c\u53cc\u5411\u5c5e\u6027\u5141\u8bb8\u7f51\u7edc\u5728\u524d\u5411\u4f20\u64ad\u6a21\u5f0f\u4e0b\u8fd0\u884c\u7cfb\u7edf\uff0c\u540c\u65f6\u7528\u4e8e\u63a8\u7406\u548c\u5b66\u4e60\u3002\u8fd9\u4e2a\u7cfb\u7edf\u7684\u6743\u91cd\u57fa\u672c\u4e0a\u5b58\u50a8\u5728\u5faa\u73af\u795e\u7ecf\u7f51\u7edc\u7684\u9690\u85cf\u72b6\u6001\u4e2d\uff0c\u6240\u4ee5\u901a\u8fc7\u7b80\u5355\u5730\u8fd0\u884c\u7cfb\u7edf\uff0c\u5b83\u4eec\u53ef\u4ee5\u4f7f\u7528\u9519\u8bef\u4fe1\u53f7\u4f5c\u4e3a\u53cd\u9988\u6765\u8bad\u7ec3\u81ea\u5df1\u3002\u7531\u4e8e\u5faa\u73af\u795e\u7ecf\u7f51\u7edc\u662f\u901a\u7528\u8ba1\u7b97\u673a\uff0c\u4ed6\u4eec\u80fd\u591f\u8bc1\u660e\u8be5\u7cfb\u7edf\u53ef\u4ee5\u7f16\u7801\u57fa\u4e8e\u68af\u5ea6\u7684\u53cd\u5411\u4f20\u64ad\u7b97\u6cd5\uff0c\u901a\u8fc7\u8bad\u7ec3\u8be5\u7cfb\u7edf\u6765\u7b80\u5355\u5730\u6a21\u62df\u53cd\u5411\u4f20\u64ad\uff0c\u800c\u4e0d\u662f\u901a\u8fc7\u624b\u5de5\u8bbe\u8ba1\u660e\u786e\u5730\u8ba1\u7b97\u68af\u5ea6\u3002\u5f53\u7136\uff0c\u4ed6\u4eec\u7684\u7cfb\u7edf\u6bd4\u53cd\u5411\u4f20\u64ad\u8981\u901a\u7528\u5f97\u591a\uff0c\u56e0\u6b64\u80fd\u591f\u5b66\u4e60\u6bd4\u53cd\u5411\u4f20\u64ad\u66f4\u6709\u6548\u7684\u65b0\u5b66\u4e60\u7b97\u6cd5<\/span>\uff08\u89c1\u56fe13\uff09<\/span>\u3002<\/span><\/section>\n

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

\u56fe13. Sandler\uff082021\uff09\u548cKirsch\u548cSchmidhuber\uff082020\uff09\u6700\u8fd1\u7684\u5de5\u4f5c\u8bd5\u56fe\u6982\u62ec\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\u7684\u516c\u8ba4\u6982\u5ff5\uff0c\u6bcf\u4e2a\u795e\u7ecf\u5143\u53ef\u4ee5\u6301\u6709\u591a\u4e2a\u72b6\u6001\u800c\u4e0d\u662f\u4e00\u4e2a\u6807\u91cf\u503c\uff0c\u6bcf\u4e2a\u7a81\u89e6\u7684\u529f\u80fd\u662f\u53cc\u5411\u7684\uff0c\u4ee5\u4fc3\u8fdb\u5b66\u4e60\u548c\u63a8\u7406\u3002\u5728\u8fd9\u4e2a\u56fe\u4e2d\uff0c\uff08Kirsch and Schmidhuber, 2020\uff09\u4f7f\u7528\u4e00\u4e2a\u76f8\u540c\u7684\u5faa\u73af\u795e\u7ecf\u7f51\u7edc\uff08RNN\uff09\uff08\u5177\u6709\u4e0d\u540c\u7684\u5185\u90e8\u9690\u85cf\u72b6\u6001\uff09\u6765\u6a21\u62df\u6bcf\u4e2a\u7a81\u89e6\uff0c\u5e76\u8868\u660e\u7f51\u7edc\u53ef\u4ee5\u901a\u8fc7\u7b80\u5355\u5730\u5411\u524d\u8fd0\u884c\u5faa\u73af\u795e\u7ecf\u7f51\u7edc\u6765\u8bad\u7ec3\uff0c\u800c\u4e0d\u4f7f\u7528\u53cd\u5411\u4f20\u64ad\u3002<\/span><\/p>\n


<\/section>\n
\u672c\u8282\u63d0\u5230\u7684\u524d\u4e24\u9879\u5de5\u4f5c\u5728\u5199\u4f5c\u65f6\u624d\u521a\u521a\u53d1\u8868\uff0c\u6211\u4eec\u76f8\u4fe1\uff0c\u8fd9\u4e9b\u53bb\u4e2d\u5fc3\u5316\u7684\u5c40\u90e8\u5143\u5b66\u4e60\u65b9\u6cd5\u6709\u53ef\u80fd\u5728\u672a\u6765\u5f7b\u5e95\u6539\u53d8\u795e\u7ecf\u7f51\u7edc\u7684\u4f7f\u7528\u65b9\u5f0f\uff0c\u6311\u6218\u76ee\u524d\u5206\u79bb\u6a21\u578b\u8bad\u7ec3\u548c\u6a21\u578b\u90e8\u7f72\u7684\u8303\u5f0f\u3002\u5728\u8bc1\u660e\u8fd9\u4e9b\u65b9\u6cd5\u53ef\u4ee5\u6269\u5c55\u5230\u66f4\u5927\u7684\u6570\u636e\u96c6\u65b9\u9762\u4ecd\u6709\u8bb8\u591a\u5de5\u4f5c\u8981\u505a\uff0c\u8fd9\u662f\u56e0\u4e3a\u56fa\u6709\u7684\u66f4\u5927\u7684\u5185\u5b58\u9700\u6c42<\/span>\uff08\u7531\u4e8e\u7cfb\u7edf\u7684\u5185\u90e8\u72b6\u6001\u5927\u5f88\u591a\uff09<\/span>\u3002\u6b64\u5916\uff0c\u867d\u7136\u8fd9\u4e9b\u7b97\u6cd5\u80fd\u591f\u4ea7\u751f\u6bd4\u68af\u5ea6\u4e0b\u964d\u66f4\u6709\u6548\u7684\u5b66\u4e60\u7b97\u6cd5\uff0c\u4f46\u8fd9\u79cd\u6548\u7387\u53ea\u5728\u5b66\u4e60\u7684\u65e9\u671f\u9636\u6bb5\u660e\u663e\uff0c\u800c\u4e14\u6027\u80fd\u5f80\u5f80\u5728\u65e9\u671f\u8fbe\u5230\u9876\u5cf0\u3002\u68af\u5ea6\u4e0b\u964d\uff0c\u867d\u7136\u6548\u7387\u8f83\u4f4e\uff0c\u4f46\u5bf9\u5c11\u6837\u672c\u5b66\u4e60\u7684\u504f\u5411\u8f83\u5c0f\uff0c\u800c\u4e14\u53ef\u4ee5\u7ee7\u7eed\u8fd0\u884c\u66f4\u591a\u7684\u5468\u671f\u6765\u5b8c\u5584\u6743\u91cd\u53c2\u6570\uff0c\u6700\u7ec8\u4ea7\u751f\u7684\u7f51\u7edc\u5c06\u53d6\u5f97\u66f4\u9ad8\u7684\u6027\u80fd\u3002<\/span><\/section>\n

<\/section>\n

<\/section>\n

\n
\n
\n
\n
\n

<\/section>\n

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

5. \u8ba8\u8bba<\/span><\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n
 <\/span><\/section>\n
\u5728\u8fd9\u7bc7\u7efc\u8ff0\u4e2d\uff0c\u6211\u4eec\u9996\u5148\u7b80\u8981\u4ecb\u7ecd\u4e86\u5386\u53f2\u80cc\u666f\uff0c\u4ee5\u63cf\u8ff0\u6df1\u5ea6\u5b66\u4e60\u548c\u96c6\u4f53\u667a\u80fd\u7814\u7a76\u7684\u4ea4\u7ec7\u53d1\u5c55\u3002\u8fd9\u4e24\u4e2a\u7814\u7a76\u9886\u57df\u7684\u8bde\u751f\u65f6\u95f4\u5927\u81f4\u76f8\u540c\uff0c\u6211\u4eec\u4e5f\u53ef\u4ee5\u53d1\u73b0\u8fd9\u4e24\u4e2a\u9886\u57df\u5728\u6574\u4e2a\u5386\u53f2\u4e0a\u7684\u5174\u8870\u6709\u4e00\u4e9b\u6b63\u76f8\u5173\u7684\u5173\u7cfb\u3002\u8fd9\u5e76\u975e\u5de7\u5408\uff0c\u56e0\u4e3a\u8fd9\u4e24\u4e2a\u9886\u57df\u4e2d\u7684\u4e00\u4e2a\u9886\u57df\u7684\u8fdb\u5c55\u548c\u7a81\u7834\u901a\u5e38\u53ef\u4ee5\u521b\u9020\u65b0\u7684\u60f3\u6cd5\uff0c\u6216\u8865\u5145\u53e6\u4e00\u4e2a\u9886\u57df\u95ee\u9898\u7684\u89e3\u51b3\u65b9\u6848\u3002\u4f8b\u5982\uff0c\u5c06\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u548c\u76f8\u5173\u8bad\u7ec3\u7b97\u6cd5\u5f15\u5165\u5143\u80de\u81ea\u52a8\u673a\uff0c\u4f7f\u6211\u4eec\u80fd\u591f\u5f00\u53d1\u51fa\u6297\u566a\u97f3\u548c\u5177\u6709\u201c\u81ea\u6211\u4fee\u590d\u201d\u7279\u6027\u7684\u56fe\u50cf\u751f\u6210\u7b97\u6cd5\u3002\u8fd9\u9879\u8c03\u67e5\u63a2\u7d22\u4e86\u6df1\u5ea6\u5b66\u4e60\u4e2d\u7684\u51e0\u9879\u5de5\u4f5c\uff0c\u8fd9\u4e9b\u5de5\u4f5c\u4e5f\u53d7\u5230\u4e86\u96c6\u4f53\u667a\u80fd\u4e2d\u7684\u6982\u5ff5\u7684\u542f\u53d1\u3002\u5728\u5b8f\u89c2\u5c42\u9762\u4e0a\uff0c\u591a\u4e3b\u4f53\u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60\u4e2d\u7684\u96c6\u4f53\u667a\u80fd\u63a8\u52a8\u4e86\u6709\u8da3\u7684\u5de5\u4f5c\uff0c\u8fd9\u4e9b\u5de5\u4f5c\u53ef\u4ee5\u901a\u8fc7\u96c6\u4f53\u81ea\u6211\u535a\u5f08\u8d85\u8d8a\u4eba\u7c7b\u7684\u8868\u73b0\uff0c\u5e76\u53bb\u4e2d\u5fc3\u5316\u7684\u81ea\u7ec4\u7ec7\u673a\u5668\u4eba\u63a7\u5236\u5668\u7684\u53d1\u5c55\uff1b\u5728\u5fae\u89c2\u5c42\u9762\u4e0a\uff0c\u96c6\u4f53\u667a\u80fd\u4e5f\u8574\u542b\u5728\u7528\u6df1\u5ea6\u6a21\u578b\u5728\u7cfb\u7edf\u5185\u4ee5\u66f4\u7cbe\u7ec6\u7684\u9897\u7c92\u5ea6\u6a21\u62df\u6bcf\u4e2a\u795e\u7ecf\u5143\u3001\u7a81\u89e6\u6216\u5176\u4ed6\u5bf9\u8c61\u7684\u5148\u8fdb\u65b9\u6cd5\u91cc\u9762\u3002<\/span><\/section>\n

<\/section>\n
\u5c3d\u7ba1\u5728\u672c\u6587\u4e2d\u63cf\u8ff0\u7684\u5de5\u4f5c\u53d6\u5f97\u4e86\u8fdb\u5c55\uff0c\u4f46\u4ecd\u6709\u8bb8\u591a\u6311\u6218\u6446\u5728\u9762\u524d\u3002\u867d\u7136\u795e\u7ecf\u5143\u80de\u81ea\u52a8\u673a\u6280\u672f\u5df2\u7ecf\u88ab\u5e94\u7528\u4e8e\u56fe\u50cf\u5904\u7406\uff0c\u4f46\u5230\u76ee\u524d\u4e3a\u6b62\uff0c\u5b83\u4eec\u7684\u5e94\u7528\u4ec5\u9650\u4e8e\u76f8\u5bf9\u8f83\u5c0f\u548c\u7b80\u5355\u7684\u6570\u636e\u96c6\uff0c\u800c\u4e14\u5b83\u4eec\u7684\u56fe\u50cf\u751f\u6210\u8d28\u91cf\uff0c\u4ecd\u7136\u8fdc\u8fdc\u4f4e\u4e8e\u5982ImageNet\u6216Celebrity Faces<\/span>\uff08Palm et al. 2022\uff09<\/span>\u7b49\u66f4\u590d\u6742\u6570\u636e\u96c6\u4e0a\u7684\u6700\u65b0\u6280\u672f\u6c34\u5e73\u3002\u5bf9\u4e8e\u6df1\u5ea6\u5f3a\u5316\u5b66\u4e60\u6765\u8bf4\uff0c\u867d\u7136\u6240\u7814\u7a76\u7684\u5de5\u4f5c\u5df2\u7ecf\u8bc1\u660e\u4e86\u4e00\u4e2a\u5168\u5c40\u7b56\u7565\u53ef\u4ee5\u88ab\u8f83\u5c0f\u7684\u5355\u4e2a\u7b56\u7565\u7684\u96c6\u5408\u6240\u53d6\u4ee3\uff0c\u4f46\u6211\u4eec\u8fd8\u6ca1\u6709\u5c06\u8fd9\u4e9b\u5b9e\u9a8c\u8f6c\u79fb\u5230\u771f\u6b63\u7684\u7269\u7406\u673a\u5668\u4eba\u4e0a\u3002\u6700\u540e\uff0c\u6211\u4eec\u5df2\u7ecf\u89c1\u8bc1\u4e86\u81ea\u7ec4\u7ec7\u6307\u5bfc\u5143\u5b66\u4e60\u7b97\u6cd5\u3002\u867d\u7136\u8fd9\u4e00\u5de5\u4f5c\u8def\u7ebf\u975e\u5e38\u6709\u524d\u9014\uff0c\u4f46\u7531\u4e8e\u7528\u6574\u4e2a\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\u53d6\u4ee3\u6bcf\u4e00\u4e2a\u795e\u7ecf\u8fde\u63a5\u6240\u5e26\u6765\u7684\u5927\u91cf\u8ba1\u7b97\u8981\u6c42\uff0c\u5b83\u4eec\u76ee\u524d\u53ea\u9650\u4e8e\u5c0f\u89c4\u6a21\u7684\u5b9e\u9a8c\u3002\u6211\u4eec\u76f8\u4fe1\u8bb8\u591a\u6311\u6218\u5c06\u5728\u9002\u5f53\u7684\u65f6\u5019\u5f97\u5230\u89e3\u51b3\uff0c\u56e0\u4e3a\u5b83\u4eec\u7684\u53d1\u5c55\u8f68\u8ff9\u5df2\u7ecf\u5f00\u59cb\u4e86\u3002<\/span><\/section>\n

<\/section>\n
\u7eb5\u89c2\u5b83\u4eec\u5404\u81ea\u7684\u53d1\u5c55\u8f68\u8ff9\uff0c\u6df1\u5ea6\u5b66\u4e60\u5728\u5f00\u53d1\u65b0\u578b\u67b6\u6784\u548c\u8bad\u7ec3\u7b97\u6cd5\u65b9\u9762\u53d6\u5f97\u4e86\u663e\u8457\u7684\u6210\u5c31\uff0c\u4ece\u800c\u5b9e\u73b0\u4e86\u9ad8\u6548\u5b66\u4e60\u548c\u66f4\u597d\u7684\u6027\u80fd\u3002\u6df1\u5ea6\u5b66\u4e60\u7684\u7814\u7a76\u548c\u5f00\u53d1\u5468\u671f\u66f4\u6ce8\u91cd\u5de5\u7a0b\u6784\u5efa\uff0c\u56e0\u6b64\u770b\u5230\u7684\u8fdb\u5c55\u66f4\u591a\u7684\u662f\u57fa\u4e8e\u57fa\u51c6<\/span>\uff08\u5982\u56fe\u50cf\u8bc6\u522b\u95ee\u9898\u7684\u5206\u7c7b\u51c6\u786e\u5ea6\u6216\u8bed\u8a00\u5efa\u6a21\u548c\u673a\u5668\u7ffb\u8bd1\u95ee\u9898\u7684\u76f8\u5173\u91cf\u5316\u6307\u6807\uff09<\/span>\u3002\u6df1\u5ea6\u5b66\u4e60\u7684\u8fdb\u5c55\u901a\u5e38\u662f\u66f4\u591a\u7684\u589e\u91cf\u548c\u53ef\u9884\u6d4b\u7684\uff0c\u800c\u96c6\u4f53\u667a\u80fd\u5219\u66f4\u591a\u5730\u5173\u6ce8\u95ee\u9898\u7684\u63d0\u51fa\u548c\u73af\u5883\u673a\u5236\uff0c\u4ee5\u6fc0\u53d1\u65b0\u7684\u6d8c\u73b0\u7684\u7fa4\u4f53\u884c\u4e3a\u3002\u6b63\u5982\u6211\u4eec\u5728\u8fd9\u6b21\u8c03\u67e5\u4e2d\u6240\u663e\u793a\u7684\uff0c\u57fa\u4e8e\u96c6\u4f53\u667a\u80fd\u7684\u6280\u672f\u80fd\u591f\u5b9e\u73b0\u4ee5\u524d\u6839\u672c\u4e0d\u53ef\u80fd\u5b9e\u73b0\u7684\u65b0\u80fd\u529b\u3002\u4f8b\u5982\uff0c\u4e0d\u53ef\u80fd\u5c06\u4e00\u4e2a\u56fa\u5b9a\u7684\u673a\u5668\u4eba\u9010\u6b65\u6539\u8fdb\u6210\u4e00\u4e2a\u80fd\u591f\u81ea\u7ec4\u88c5\u7684\u673a\u5668\u4eba\uff0c\u5e76\u4ece\u8fd9\u79cd\u6a21\u5757\u5316\u4e2d\u83b7\u5f97\u6240\u6709\u7684\u597d\u5904\u3002\u81ea\u7136\uff0c\u8fd9\u4e24\u4e2a\u9886\u57df\u53ef\u4ee5\u76f8\u4e92\u8865\u5145\u3002\u6211\u4eec\u76f8\u4fe1\uff0c\u8fd9\u79cd\u643a\u624b\u5e76\u8fdb\u7684\u5171\u540c\u5f00\u53d1\u65b9\u5f0f\u5c06\u7ee7\u7eed\u4e0b\u53bb\u3002<\/span><\/section>\n

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

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

\n
\n
\n
\n

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

    \n
  1. \n
    Alam M, Samad MD, Vidyaratne L, et al. (2020) Survey on deep neural networks in speech and vision systems. Neurocomputing 417: 302\u2013321.<\/section>\n<\/li>\n
  2. \n
    Baker B, Kanitscheider I, Markov T, et al. (2019) Emergent Tool Use from Multi-Agent Autocurricula. arXiv preprint arXiv:1909<\/em>. 07528.<\/section>\n<\/li>\n
  3. \n
    Bansal T, Pachocki J, Sidor S, et al. (2017) Emergent Complexity via Multi-Agent Competition. arXiv preprint arXiv:1710<\/em>. 03748.<\/section>\n<\/li>\n
  4. \n
    Bhatia J, Jackson H, Tian Y, et al. (2021) Evolution gym: A large-scale benchmark for evolving soft robots. In: Advances in Neural Information Processing Systems. Curran Associates, Inc.<\/section>\n<\/li>\n
  5. \n
    Brockman G, Cheung V, Pettersson L, et al. (2016) Openai Gym. arXiv preprint arXiv:1606<\/em>. 01540.<\/section>\n<\/li>\n
  6. \n
    Brown TB, Mann B, Ryder N, et al. (2020) Language Models Are Few-Shot Learners. arXiv preprint arXiv:2005<\/em>. 14165.<\/section>\n<\/li>\n
  7. \n
    Cheney N, MacCurdy R, Clune J, et al. (2014) Unshackling evolution: evolving soft robots with multiple materials and a powerful generative encoding. ACM SIGEVOlution 7(1): 11\u201323.<\/section>\n<\/li>\n
  8. \n
    Chollet F, et al. (2015) Keras.<\/section>\n<\/li>\n
  9. \n
    Chua LO, Roska T (2002) Cellular Neural Networks and Visual Computing: Foundations and Applications. Cambridge University Press.<\/section>\n<\/li>\n
  10. \n
    Chua LO, Yang L (1988a) Cellular neural networks: Applications. IEEE Transactions on Circuits and Systems 35(10): 1273\u20131290.<\/section>\n<\/li>\n
  11. \n
    Chua LO, Yang L (1988b) Cellular neural networks: Theory. IEEE Transactions on Circuits and Systems 35(10): 1257\u20131272.<\/section>\n<\/li>\n
  12. \n
    Conway J, et al. (1970) The game of life. Scientific American 223(4): 4.<\/section>\n<\/li>\n
  13. \n
    Daigavane A, Ravindran B, Aggarwal G (2021) Understanding Convolutions on Graphs. Distill. https:\/\/distill.pub\/2021\/understanding-gnns<\/section>\n<\/li>\n
  14. \n
    Deneubourg J-L, Goss S (1989) Collective patterns and decision-making. Ethology Ecology & Evolution 1(4): 295\u2013311.<\/section>\n<\/li>\n
  15. \n
    Deng J, Dong W, Socher R, et al. (2009) Imagenet: A large-scale hierarchical image database. In: 2009 IEEE conference on computer vision and pattern recognition, pp. 248\u2013255.<\/section>\n<\/li>\n
  16. \n
    Dorigo M, Bonabeau E, Theraulaz G (2000) Ant algorithms and stigmergy. Future Generation Computer Systems 16(8): 851\u2013871.<\/section>\n<\/li>\n
  17. \n
    Foerster JN, Assael YM, De Freitas N, et al. (2016) Learning to Communicate with Deep Multi-Agent Reinforcement Learning. arXiv preprint arXiv:1605.06676<\/em>.<\/section>\n<\/li>\n
  18. \n
    Freeman CD, Metz L, Ha D (2019) Learning to Predict without Looking Ahead: World Models without Forward Prediction.<\/section>\n<\/li>\n
  19. \n
    Gilpin W (2019) Cellular automata as convolutional neural networks. Physical Review E 100(3): 032402.<\/section>\n<\/li>\n
  20. \n
    GoraS L, Chua LO, Leenaerts D (1995) Turing patterns in cnns. i. once over lightly. IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications 42(10): 602\u2013611.<\/section>\n<\/li>\n
  21. \n
    Grattarola D, Livi L, Alippi C (2021) Learning Graph Cellular Automata.<\/section>\n<\/li>\n
  22. \n
    Ha D (2018) Reinforcement Learning for Improving Agent Design.<\/section>\n<\/li>\n
  23. \n
    Ha D (2020) Slime Volleyball Gym Environment. https:\/\/github.com\/hardmaru\/slimevolleygym<\/section>\n<\/li>\n
  24. \n
    Ha D, Schmidhuber J (2018) Recurrent world models facilitate policy evolution. Advances in Neural Information Processing Systems 31: 2451\u20132463. https:\/\/worldmodels.github.io<\/section>\n<\/li>\n
  25. \n
    Hamann H (2018) Swarm Robotics: A Formal Approach. Springer.<\/section>\n<\/li>\n
  26. \n
    He K, Zhang X, Ren S, et al. (2016) Deep residual learning for image recognition. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770\u2013778.<\/section>\n<\/li>\n
  27. \n
    Heiden E, Millard D, Coumans E, et al. (2021) NeuralSim: Augmenting differentiable simulators with neural networks. In: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA).<\/section>\n<\/li>\n
  28. \n
    Hill A, Raffin A, Ernestus M, et al. (2018) Stable Baselines.<\/section>\n<\/li>\n
  29. \n
    Hooker S (2020) The Hardware Lottery. arXiv preprint arXiv:2009.06489<\/em>.<\/section>\n<\/li>\n
  30. \n
    Horibe K, Walker K, Risi S (2021) Regenerating soft robots through neural cellular automata. In: EuroGP, pp. 36\u201350.<\/section>\n<\/li>\n
  31. \n
    Huang W, Mordatch I, Pathak D (2020) One policy to control them all: Shared modular policies for agent-agnostic control. In: International Conference on Machine Learning, pp. 4455\u20134464.<\/section>\n<\/li>\n
  32. \n
    Jabbar A, Li X, Omar B (2021) A survey on generative adversarial networks: Variants, applications, and training. ACM Computing Surveys (CSUR) 54(8): 1\u201349.<\/section>\n<\/li>\n
  33. \n
    Jaderberg M, Czarnecki WM, Dunning I, et al. (2019) Human-level performance in 3d multiplayer games with population-based reinforcement learning. Science 364(6443): 859\u2013865.<\/section>\n<\/li>\n
  34. \n
    Jenal M (2011) What Ants Can Teach Us about the Market.<\/section>\n<\/li>\n
  35. \n
    Joachimczak M, Suzuki R, Arita T (2016) Artificial metamorphosis: Evolutionary design of transforming, soft-bodied robots. Artificial Life 22(3): 271\u2013298.<\/section>\n<\/li>\n
  36. \n
    Kirsch L, Schmidhuber J (2020) Meta Learning Backpropagation and Improving it. arXiv preprint arXiv:2012<\/em>. 14905.<\/section>\n<\/li>\n
  37. \n
    Kozek T, Roska T, Chua LO (1993) Genetic algorithm for cnn template learning. IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications 40(6): 392\u2013402.<\/section>\n<\/li>\n
  38. \n
    Krizhevsky A, Sutskever I, Hinton GE (2012) Imagenet classification with deep convolutional neural networks. Advances in Neural Information Processing Systems 25: 1097\u20131105.<\/section>\n<\/li>\n
  39. \n
    Lajad R, Moreno E, Arenas A (2021) Young honeybees show learned preferences after experiencing adulterated pollen. Scientific Reports 11(1): 1\u201311.<\/section>\n<\/li>\n
  40. \n
    Leimeister JM (2010) Collective intelligence. Business & Information Systems Engineering 2(4): 245\u2013248.<\/section>\n<\/li>\n
  41. \n
    L\u00e9vy P (1997) Collective Intelligence.<\/section>\n<\/li>\n
  42. \n
    Liu J-B, Raza Z, Javaid M (2020) Zagreb connection numbers for cellular neural networks. Discrete Dynamics in Nature and Society 2020: 1\u20138.<\/section>\n<\/li>\n
  43. \n
    Liu S, Lever G, Merel J, et al. (2019) Emergent Coordination through Competition. arXiv preprint arXiv:1902<\/em>.07151<\/section>\n<\/li>\n
  44. \n
    Mataric MJ (1993) Designing emergent behaviors: From local interactions to collective intelligence. In: Proceedings of the Second International Conference on Simulation of Adaptive Behavior, pp. 432\u2013441.<\/section>\n<\/li>\n
  45. \n
    Mnih V, Kavukcuoglu K, Silver D, et al. (2015) Human-level control through deep reinforcement learning. Nature 518(7540): 529\u2013533.<\/section>\n<\/li>\n
  46. \n
    Mordvintsev A, Randazzo E, Niklasson E, et al. (2020) Growing Neural Cellular Automata. Distill.<\/section>\n<\/li>\n
  47. \n
    Ohsawa S, Akuzawa K, Matsushima T, et al. (2018) Neuron as an Agent.<\/section>\n<\/li>\n
  48. \n
    OroojlooyJadid A, Hajinezhad D (2019) A Review of Cooperative Multi-Agent Deep Reinforcement Learning. arXiv preprint arXiv:1908.03963<\/em>.<\/section>\n<\/li>\n
  49. \n
    Ott J (2020) Giving up Control: Neurons as Reinforcement Learning Agents. arXiv preprint arXiv:2003.11642<\/em>.<\/section>\n<\/li>\n
  50. \n
    Palm RB, Duque MG, Sudhakaran S, et al. (2022) Variational neural cellular automata. In: International Conference on Learning Representations.<\/section>\n<\/li>\n
  51. \n
    Pathak D, Lu C, Darrell T, et al. (2019) Learning to Control Self-Assembling Morphologies: A Study of Generalization via Modularity. arXiv preprint arXiv:1902.05546<\/em>.<\/section>\n<\/li>\n
  52. \n
    Peng Z, Hui KM, Liu C, et al. (2021) Learning to simulate self-driven particles system with coordinated policy optimization. Advances in Neural Information Processing Systems 34.<\/section>\n<\/li>\n
  53. \n
    Pickering A (2010) The Cybernetic Brain. University of Chicago Press.<\/section>\n<\/li>\n
  54. \n
    Qin Y, Feng M, Lu H, et al. (2018) Hierarchical cellular automata for visual saliency. International Journal of Computer Vision 126(7): 751\u2013770.<\/section>\n<\/li>\n
  55. \n
    Qu X, Sun Z, Ong YS, et al. (2020) Minimalistic attacks: How little it takes to fool deep reinforcement learning policies. IEEE Transactions on Cognitive and Developmental Systems 13: 806\u2013817.<\/section>\n<\/li>\n
  56. \n
    Radford A, Kim JW, Hallacy C, et al. (2021) Learning Transferable Visual Models from Natural Language Supervision. arXiv preprint arXiv:2103.00020<\/em>.<\/section>\n<\/li>\n
  57. \n
    Radford A, Narasimhan K, Salimans T, et al. (2018) Improving Language Understanding by Generative Pre-training.<\/section>\n<\/li>\n
  58. \n
    Radford A, Wu J, Child R, et al. (2019) Language models are unsupervised multitask learners. OpenAI blog 1(8): 9.<\/section>\n<\/li>\n
  59. \n
    Randazzo E, Mordvintsev A, Niklasson E, et al. (2020) Self-classifying Mnist Digits. Distill.<\/section>\n<\/li>\n
  60. \n
    Resnick C, Eldridge W, Ha D, et al. (2018) Pommerman: A Multi-Agent Playground. arXiv preprint arXiv:1809.07124<\/em>.<\/section>\n<\/li>\n
  61. \n
    Rubenstein M, Cornejo A, Nagpal R (2014) Programmable self-assembly in a thousand-robot swarm. Science 345(6198): 795\u2013799.<\/section>\n<\/li>\n
  62. \n
    Rudin N, Hoeller D, Reist P, et al. (2021) Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv preprint arXiv:2109.11978<\/em>.<\/section>\n<\/li>\n
  63. \n
    Sanchez-Lengeling B, Reif E, Pearce A, et al. (2021) A gentle introduction to graph neural networks. Distill 6(9): e33.<\/section>\n<\/li>\n
  64. \n
    Sandler M, Vladymyrov M, Zhmoginov A, et al. (2021) Meta-learning bidirectional update rules. In: International Conference on Machine Learning, pp. 9288\u20139300.<\/section>\n<\/li>\n
  65. \n
    Sandler M, Zhmoginov A, Luo L, et al. (2020) Image Segmentation via Cellular Automata. arXiv preprint arXiv:2008.04965<\/em>.<\/section>\n<\/li>\n
  66. \n
    Schilling MA (2000) Toward a general modular systems theory and its application to interfirm product modularity. Academy of Management Review 25(2): 312\u2013334.<\/section>\n<\/li>\n
  67. \n
    Schilling MA, Steensma HK (2001) The use of modular organizational forms: An industry-level analysis. Academy of Management Journal 44(6): 1149\u20131168.<\/section>\n<\/li>\n
  68. \n
    Schmidhuber J (2014) Who Invented Backpropagation? More[DL2]<\/em>.<\/section>\n<\/li>\n
  69. \n
    Schmidhuber J (2020) Metalearning Machines Learn to Learn (1987). https:\/\/people.idsia.ch\/juergen\/metalearning.html<\/section>\n<\/li>\n
  70. \n
    Schoenholz S, Cubuk ED (2020) Jax md: a framework for differentiable physics. Advances in Neural Information Processing Systems 33.<\/section>\n<\/li>\n
  71. \n
    Schweitzer F, Farmer JD (2003) Brownian Agents and Active Particles: Collective Dynamics in the Natural and Social Sciences. Springer. Volume 1.<\/section>\n<\/li>\n
  72. \n
    Seeley TD (2010) Honeybee Democracy. Princeton University Press.<\/section>\n<\/li>\n
  73. \n
    Silver D, Huang A, Maddison CJ, et al. (2016) Mastering the game of go with deep neural networks and tree search. Nature 529(7587): 484\u2013489.<\/section>\n<\/li>\n
  74. \n
    Simonyan K, Zisserman A (2014) Very Deep Convolutional Networks for Large-Scale Image Recognition. arXiv preprint arXiv:1409.1556<\/em>.<\/section>\n<\/li>\n
  75. \n
    Stahlberg F (2020) Neural machine translation: A review. Journal of Artificial Intelligence Research 69: 343\u2013418.<\/section>\n<\/li>\n
  76. \n
    Stoy K, Brandt D, Christensen DJ, et al. (2010) Self-reconfigurable Robots: An Introduction.<\/section>\n<\/li>\n
  77. \n
    Suarez J, Du Y, Isola P, et al. (2019) Neural Mmo: A Massively Multiagent Game Environment for Training and Evaluating Intelligent Agents. arXiv preprint arXiv:1903.00784<\/em>.<\/section>\n<\/li>\n
  78. \n
    Suarez J, Du Y, Zhu C, et al. (2021) The neural mmo platform for massively multiagent research. In: Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track.<\/section>\n<\/li>\n
  79. \n
    Sudhakaran S, Grbic D, Li S, et al. (2021) Growing 3d Artefacts and Functional Machines with Neural Cellular Automata. arXiv preprint arXiv:2103.08737<\/em>.<\/section>\n<\/li>\n
  80. \n
    Sumpter DJ (2010) Collective Animal Behavior. Princeton University Press.<\/section>\n<\/li>\n
  81. \n
    Surowiecki J (2005) The Wisdom of Crowds. Anchor.<\/section>\n<\/li>\n
  82. \n
    Tan X, Qin T, Soong F, et al. (2021) A Survey on Neural Speech Synthesis. arXiv preprint arXiv:2106.15561<\/em>.<\/section>\n<\/li>\n
  83. \n
    Tang Y, Ha D (2021) The sensory neuron as a transformer: Permutation-invariant neural networks for reinforcement learning. In: Thirty-Fifth Conference on Neural Information Processing Systems. https:\/\/attentionneuron.github.io<\/section>\n<\/li>\n
  84. \n
    Tang Y, Nguyen D, Ha D (2020) Neuroevolution of self-interpretable agents. In: Proceedings of the Genetic and Evolutionary Computation Conference.<\/section>\n<\/li>\n
  85. \n
    Tang Y, Tian Y, Ha D (2022) Evojax: Hardware-Accelerated Neuroevolution. arXiv preprint arXiv:2202.05008<\/em>.<\/section>\n<\/li>\n
  86. \n
    Tapscott D, Williams AD (2008) Wikinomics: How Mass Collaboration Changes Everything. Penguin.<\/section>\n<\/li>\n
  87. \n
    Terry JK, Black B, Jayakumar M, et al. (2020) Pettingzoo: Gym for Multi-Agent Reinforcement Learning. arXiv preprint arXiv:2009.14471<\/em>.<\/section>\n<\/li>\n
  88. \n
    Toner J, Tu Y, Ramaswamy S (2005) Hydrodynamics and phases of flocks. Annals of Physics 318(1): 170\u2013244.<\/section>\n<\/li>\n
  89. \n
    Vinyals O, Babuschkin I, Czarnecki WM, et al. (2019) Grandmaster level in starcraft ii using multi-agent reinforcement learning. Nature 575(7782): 350\u2013354.<\/section>\n<\/li>\n
  90. \n
    Wang T, Liao R, Ba J, et al. (2018) Nervenet: Learning structured policy with graph neural networks. In: International Conference on Learning Representations.<\/section>\n<\/li>\n
  91. \n
    Wang Z, She Q, Ward TE (2021) Generative adversarial networks in computer vision: A survey and taxonomy. ACM Computing Surveys (CSUR) 54(2): 1\u201338.<\/section>\n<\/li>\n
  92. \n
    Wikipedia (2022) Trajan\u2019s Bridge at Alcantara. Wikipedia.<\/section>\n<\/li>\n
  93. \n
    Wolfram S (2002) A New Kind of Science. Champaign, IL: Wolfram media. Volume 5.<\/section>\n<\/li>\n
  94. \n
    Wu Z, Pan S, Chen F, et al. (2020) A comprehensive survey on graph neural networks. IEEE Transactions on Neural Networks and Learning Systems 32(1): 4\u201324.<\/section>\n<\/li>\n
  95. \n
    Zhang D, Choi C, Kim J, et al. (2021) Learning to generate 3d shapes with generative cellular automata. In: International Conference on Learning Representations.<\/section>\n<\/li>\n
  96. \n
    Zheng L, Yang J, Cai H, et al. (2018) Magent: A many-agent reinforcement learning platform for artificial collective intelligence. Proceedings of the AAAI Conference on Artificial Intelligence 32.<\/section>\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

    <\/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\u6211\u7684\u96c6\u667a\u300d\u63a8\u9001\u8bba\u6587\u4fe1\u606f\u3002\u626b\u63cf\u4e0b\u65b9\u4e8c\u7ef4\u7801\u5373\u53ef\u4e00\u952e\u8ba2\u9605\uff1a<\/span>
    <\/section>\n


    <\/p>\n

    <\/p>\n


    <\/section>\n

    <\/section>\n
    \n

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