{"id":14099,"date":"2023-07-29T02:06:20","date_gmt":"2023-07-28T18:06:20","guid":{"rendered":"http:\/\/www.brofive.org\/?p=14099"},"modified":"2023-08-09T21:56:34","modified_gmt":"2023-08-09T13:56:34","slug":"%e5%8d%8e%e4%b8%ba%e7%9b%98%e5%8f%a4%e5%a4%a7%e6%a8%a1%e5%9e%8b%e5%8f%91%e5%b1%95%e7%ae%80%e5%8f%b2%e6%b0%94%e8%b1%a1%e5%a4%a7%e6%a8%a1%e5%9e%8b","status":"publish","type":"post","link":"http:\/\/www.brofive.net\/?p=14099","title":{"rendered":"\u534e\u4e3a\u76d8\u53e4\u5927\u6a21\u578b\u53d1\u5c55\u7b80\u53f2&\u6c14\u8c61\u5927\u6a21\u578b"},"content":{"rendered":"

\u6700\u8fd1\u4e00\u76f4\u5728\u8ddf\u8e2a\u5927\u6a21\u578b\u7684\u4e8b\u60c5\uff0c\u70ed\u5ea6\u5f88\u9ad8\uff0c\u6bcf\u6b21\u5f00\u4f1a\u4e0d\u63d0\u5927\u6a21\u578b\u90fd\u663e\u5f97\u4e0d\u63a5\u5730\u6c14\uff0c\u4f46\u662f\u5927\u6a21\u578b\u770b\u6765\u770b\u53bb\u4f9d\u7136\u5b58\u5728AI\u7684\u901a\u75c5\uff1a\u4e0d\u662f100%\u51c6\u786e\uff1b\u66f4\u65e0\u6cd5\u89e3\u91ca<\/font><\/strong>\u3002\u5927\u6a21\u578b\u77ed\u671f\u5185\u8fd8\u662f\u65e0\u6cd5\u66ff\u4ee3\u4eba\u7c7b\uff0c\u4f46\u662f\u5df2\u7ecf\u53ef\u4ee5\u66f4\u597d\u7684\u8f85\u52a9\u8f85\u52a9\u4eba\u7c7b\u5f00\u5c55\u5404\u9879\u5de5\u4f5c\u4e86\u3002\u5728\u70ed\u5ea6\u4e4b\u4e0b\uff0c\u5c31\u8981\u6709\u4e2a\u6e05\u9192\u7684\u5b9a\u4f4d\u3002\u534e\u4e3a\u57287\u67087\u65e5\u53d1\u5e03\u7684\u76d8\u53e4\u5927\u6a21\u578b\uff0c\u8003\u8651\u5230\u534e\u4e3a\u662f\u5168\u6808\u6280\u672f\uff0c\u53ef\u80fd\u548c\u5176\u4ed6\u516c\u53f8\u53d1\u5e03\u7684\u5927\u6a21\u578b\u6709\u4e9b\u4e0d\u4e00\u6837\uff0c\u7b80\u5355\u8c03\u7814\u4e86\u4e00\u4e0b\u76d8\u53e4\u7684\u5386\u7a0b\uff0c\u4ee5\u53ca\u51e0\u4e2a\u91cd\u8981\u7684\u57fa\u7840\u6a21\u578b\uff0c\u7279\u522b\u662f\u88ab\u300aNature\u300b\u6536\u5f55\u7684\u300a<\/strong><\/em><\/font>Accurate medium-range global weather forecasting with 3D neural networks<\/font><\/strong><\/em><\/a>\u300b<\/strong><\/em>\u3002<\/font><\/p>\n

\u4e00\u3001\u76d8\u53e4\u5927\u6a21\u578b\u53d1\u5c55\u7b80\u53f2<\/font><\/strong><\/p>\n

\u534e\u4e3a\u76d8\u53e4\u5927\u6a21\u578b\u7684\u521b\u59cb\u4eba\u662f\u7530\u9f50\uff08\u73b0\u4efb\u534e\u4e3a\u4e91\u4eba\u5de5\u667a\u80fd\u9886\u57df\u9996\u5e2d\u79d1\u5b66\u5bb6\uff09\uff0c\u7530\u9f50\u56e2\u961f\u4e00\u5f00\u59cb\u9009\u62e9\u548cNLP\u548cCV\u4e24\u4e2a\u8d5b\u9053\uff1b\u786e\u7acb\u4e86\u4e09\u9879\u6838\u5fc3\u8bbe\u8ba1\u539f\u5219\uff1a\u4e00\u662f\u6a21\u578b\u8981\u5927<\/font><\/strong>\uff0c\u53ef\u5438\u6536\u6d77\u91cf\u7684\u6570\u636e\uff1b\u4e8c\u662f\u7f51\u7edc\u7ed3\u6784\u8981\u5f3a<\/font><\/strong>\uff0c\u80fd\u591f\u771f\u6b63\u53d1\u6325\u51fa\u6a21\u578b\u7684\u6027\u80fd\uff1b\u4e09\u662f\u8981\u5177\u6709\u4f18\u79c0\u7684\u6cdb\u5316\u80fd\u529b<\/font><\/strong>\uff0c\u53ef\u4ee5\u771f\u6b63\u843d\u5730\u5230\u5404\u884c\u5404\u4e1a\u7684\u5de5\u4f5c\u573a\u666f\u3002\u8fd9\u4e9b\u9009\u62e9\u4e00\u65b9\u9762\u662f\u534e\u4e3a\u5728AI\u9886\u57df\u5df2\u7ecf\u6df1\u8015\u591a\u5e74\uff0c\u5168\u6808\u6280\u672f\uff0c\u79ef\u7d2f\u591a\uff0c\u8d44\u6e90\u591a\uff0c\u8d77\u6b65\u9ad8\uff1b\u53e6\u5916\u4e00\u65b9\u9762\uff0c\u5bf9\u4e8e\u6280\u672f\u548c\u884c\u4e1a\u7684\u9884\u5224\u76f8\u5f53\u51c6\u786e\uff0c\u5728\u51e0\u5e74\u524d\u5c31\u5f00\u5c55\u4e86\u52a1\u5b9e\u7684\u884c\u4e1a\u5e03\u5c40\uff08\u505a\u4e8b\uff09\u3002<\/font><\/p>\n

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

\u534e\u4e3a\u76d8\u53e4\u5927\u6a21\u578b\u6784\u6210\uff08\u4e92\u8054\u7f51\u56fe\u7247\uff09<\/font><\/p>\n

2020\u5e743\u6708<\/strong>\uff0c\u7530\u5947\uff08\u534e\u4e3a\u4e91\u4eba\u5de5\u667a\u80fd\u9886\u57df\u9996\u5e2d\u79d1\u5b66\u5bb6\u3001IEEE Fellow\uff09<\/font><\/strong>\u52a0\u5165\u534e\u4e3a\u4e91\u540e\u4fbf\u5f00\u59cb\u7ec4\u5efa\u56e2\u961f\uff0c\u5e76\u4e14\u8fdb\u884c\u4e86\u65b9\u5411\u68b3\u7406\uff1b8\u6708\uff0c\u56e2\u961f\u8fce\u6765\u65b0\u7684\u6838\u5fc3\u4e13\u5bb6\uff1b9\u6708\uff0c\u56e2\u961f\u5f00\u59cb\u63a8\u52a8\u76d8\u53e4\u5927\u6a21\u578b\u7684\u7acb\u9879\uff0c\u5e0c\u671b\u80fd\u591f\u5728\u534e\u4e3a\u4e91\u7684\u4ea7\u4e1a\u57fa\u5ea7\u4e0a\uff0c\u5b8c\u6210\u9002\u914d\u5404\u4e2a\u4ea7\u4e1aAI\u5f00\u53d1\u7684\u5927\u6a21\u578b\uff1b11\u6708\uff0c\u76d8\u53e4\u5927\u6a21\u578b\u5728\u534e\u4e3a\u4e91\u5185\u90e8\u7acb\u9879\u6210\u529f\uff0c\u5e76\u5b8c\u6210\u4e86\u4e0e\u5408\u4f5c\u4f19\u4f34\u3001\u9ad8\u6821\u7684\u5408\u4f5c\u642d\u5efa\u3002<\/font><\/p>\n

2021\u5e744\u670825\u65e5<\/strong>\uff0c\u5728\u534e\u4e3a\u5f00\u53d1\u8005\u5927\u4f1a\u4e0a\uff0c\u4f59\u627f\u4e1c<\/font><\/strong>\u53d1\u5e03\u4e86\u76d8\u53e4\u7cfb\u5217\u8d85\u5927\u89c4\u6a21\u9884\u8bad\u7ec3\u6a21\u578b\uff0c\u5305\u62ec 30 \u4ebf\u53c2\u6570\u7684\u89c6\u89c9\uff08CV\uff09\u9884\u8bad\u7ec3\u6a21\u578b\uff0c\u4ee5\u53ca\u4e0e\u5faa\u73af\u667a\u80fd\u3001\u9e4f\u57ce\u5b9e\u9a8c\u5ba4\u8054\u5408\u5f00\u53d1\u7684\u5343\u4ebf\u53c2\u6570\u300140TB \u8bad\u7ec3\u6570\u636e\u7684\u4e2d\u6587\u8bed\u8a00\uff08NLP\uff09\u9884\u8bad\u7ec3\u6a21\u578b\u3002\u534e\u4e3a\u4e91\u76d8\u53e4NLP\u5927\u6a21\u578b\u5728\u4e2d\u6587\u8bed\u8a00\u7406\u89e3\u6d4b\u8bc4\u4e2d\uff0c\u5728\u603b\u699c\u3001\u5206\u7c7b\u699c\u3001\u9605\u8bfb\u7406\u89e3\u699c\u83b7\u5f97\u4e09\u9879\u7b2c\u4e00\uff0c\u5927\u5e45\u5237\u65b0\u4e1a\u754c\u7eaa\u5f55\u3002\u603b\u6392\u884c\u699c\u9886\u5148\u7b2c\u4e8c\u540d1\u5206\u3002\u76d8\u53e4\u56e2\u961f\uff1a1000\u4ebf\u53c2\u6570NLP+CV\uff0c20\u591a\u540d\u535a\u58eb\u300130\u591a\u540d\u5de5\u7a0b\u5e08\u30013\u540d\u5e7f\u53d7\u5173\u6ce8\u7684\u201c\u534e\u4e3a\u5929\u624d\u5c11\u5e74\u201d\uff0c\u8fd8\u670950\u591a\u540d\u6765\u81ea\u5168\u56fdC9\u9ad8\u6821\u7684\u4e13\u5bb6\u3002<\/strong><\/font><\/p>\n

2021\u5e744\u670826\u65e5<\/strong>\uff0c\u9e4f\u57ce\u5b9e\u9a8c\u5ba4<\/font><\/strong>\u8054\u5408\u6709\u5173\u5355\u4f4d\u53d1\u5e03\u300c\u76d8\u53e4\u03b1\u300d\uff0c\u8fd9\u662f\u4e1a\u754c\u9996\u4e2a2000\u4ebf\u53c2\u6570\u4e2d\u6587\u9884\u8bad\u7ec3\u8bed\u8a00\u6a21\u578b\uff01\u6a21\u578b\u4f7f\u7528\u56fd\u4ea7\u5168\u6808\u5f0f\u8f6f\u786c\u4ef6\u534f\u540c\u751f\u6001\uff08MindSpore+CANN+\u6607\u817e910+ModelArts\uff09\uff0c\u57fa\u4e8e\u201c\u9e4f\u57ce\u4e91\u8111\u2161\u201d\u548cMindSpore\u6846\u67b6\u7684\u81ea\u52a8\u6df7\u5408\u5e76\u884c\u6a21\u5f0f\u5b9e\u73b0\u57282048\u5361<\/font><\/strong>\u7b97\u529b\u96c6\u7fa4\u4e0a\u7684\u5927\u89c4\u6a21\u5206\u5e03\u5f0f\u8bad\u7ec3\u3002<\/font><\/p>\n

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

\u76d8\u53e4\u03b1\u6280\u672f\u903b\u8f91\u67b6\u6784<\/font><\/p>\n

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

\u76d8\u53e4\u03b1\u7684\u8bad\u7ec3\u548c\u63a8\u7406\u6280\u672f\u6808<\/font><\/p>\n

2021\u5e749\u670823\u65e5<\/strong>\uff0c\u534e\u4e3a\u5168\u8054\u63a52021\u5927\u4f1a\u4e0a\uff0c\u534e\u4e3a\u9ad8\u7ea7\u526f\u603b\u88c1\u3001\u534e\u4e3a\u4e91CEO\u3001\u6d88\u8d39\u8005\u4e91\u670d\u52a1\u603b\u88c1\u5f20\u5e73\u5b89<\/font><\/strong>\u53d1\u8868\u201c\u6df1\u8015\u6570\u5b57\u5316\uff0c\u4e00\u5207\u7686\u670d\u52a1\u201d\u4e3b\u9898\u6f14\u8bb2\uff0c\u5e76\u91cd\u78c5\u53d1\u5e03\u4e86\u534e\u4e3a\u4e91\u76d8\u53e4\u836f\u7269\u5206\u5b50\u5927\u6a21\u578b<\/strong>\u534e\u4e3a\u4e91\u53d1\u5e03\u4e86\u76d8\u53e4\u836f\u7269\u5206\u6790\u5927\u6a21\u578b\u3002\u534e\u4e3a\u4e91\u8054\u5408\u4e2d\u56fd\u79d1\u5b66\u9662\u4e0a\u6d77\u836f\u7269\u7814\u7a76\u6240\uff0c\u5171\u540c\u8bad\u7ec3\u4e86\u534e\u4e3a\u4e91\u76d8\u53e4\u836f\u7269\u5206\u5b50\u5927\u6a21\u578b\u3002\u6b64\u6a21\u578b\u5bf9\u5e02\u9762\u4e0a\u771f\u5b9e\u5b58\u5728\u768417\u4ebf\u4e2a\u836f\u7269\u5206\u5b50\u7684\u5316\u5b66\u7ed3\u6784\u8fdb\u884c\u9884\u8bad\u7ec3\uff0c\u5728\u5316\u5b66\u65e0\u76d1\u7763\u5b66\u4e60\u6a21\u5f0f\u4e0b\uff0c\u5b9e\u73b0\u7ed3\u6784\u91cd\u6784\u7387\u3001\u5408\u6cd5\u6027\u3001\u552f\u4e00\u6027\u7b49\u6307\u6807\u5168\u9762\u4f18\u4e8e\u73b0\u6709\u65b9\u6cd5\u3002\u76d8\u53e4\u836f\u7269\u5206\u5b50\u5927\u6a21\u578b\u9996\u6b21\u91c7\u7528\u201c\u56fe-\u5e8f\u5217\u4e0d\u5bf9\u79f0\u6761\u4ef6\u53d8\u5206\u81ea\u7f16\u7801\u5668<\/strong>\u201d\u67b6\u6784\uff0c \u53ef\u4ee5\u81ea\u52a8\u627e\u51fa\u5316\u5408\u7269\u5173\u952e\u7684\u5206\u5b50\u7279\u5f81\u6307\u7eb9\uff0c\u6781\u5927\u5730\u63d0\u5347\u4e86\u4e0b\u6e38\u4efb\u52a1\u7684\u51c6\u786e\u6027\u3002\u76d8\u53e4\u836f\u7269\u5206\u5b50\u5927\u6a21\u578b\u7684\u5206\u5b50\u751f\u6210\u5668\u751f\u6210\u4e861\u4ebf\u4e2a\u521b\u65b0\u7684\u7c7b\u836f\u7269\u5c0f\u5206\u5b50\u7b5b\u9009\u5e93\uff0c\u5176\u7ed3\u6784\u65b0\u9896\u6027\u4e3a99.68%\u3002\u6b64\u6a21\u578b\u4f9d\u6258\u534e\u4e3a\u4e91\u533b\u7597\u667a\u80fd\u4f53EIHealth\u5e73\u53f0\u3002
\n<\/font><\/p>\n

2022\u5e748\u670816\u65e5<\/strong>\uff0c\u53ef\u4fe1AI\u5cf0\u4f1a\u5728\u5317\u4eac\u4e3e\u529e\u3002\u534e\u4e3a\u4e91\u76d8\u53e4\u9884\u8bad\u7ec3\u5927\u6a21\u578b\u901a\u8fc7\u4e2d\u56fd\u4fe1\u901a\u9662\u9996\u8f6e\u5927\u6a21\u578b\u6d4b\u8bc4\uff0c\u5728\u201c\u6a21\u578b\u5f00\u53d1\u201d\u548c\u201c\u6a21\u578b\u80fd\u529b\u201d\u4e24\u90e8\u5206\u8fbe\u52304+\u7ea7\u6807\u51c6\u3002\u534e\u4e3a\u4e91\u4eba\u5de5\u667a\u80fd\u9886\u57df\u9996\u5e2d\u79d1\u5b66\u5bb6\u3001IEEE Fellow\u3001\u56fd\u9645\u6b27\u4e9a\u79d1\u5b66\u9662\u9662\u58eb\u7530\u5947<\/font><\/strong>\u53d1\u8868\u4e3b\u9898\u6f14\u8bb2\u3002\u4ed6\u8868\u793a\uff0c\u534e\u4e3a\u4e91\u76d8\u53e4\u9884\u8bad\u7ec3\u5927\u6a21\u578b\u4e0d\u65ad\u8fed\u4ee3\uff0c\u5f62\u6210\u4e86\u201cL0\u57fa\u7840\u5927\u6a21\u578b-L1\u884c\u4e1a\u5927\u6a21\u578b-L2\u7ec6\u5206\u573a\u666f\u5927\u6a21\u578b<\/strong>\u201d\u7684\u53d1\u5c55\u8def\u5f84\uff0c\u5b8c\u6210\u4ece\u5b66\u672f\u5927\u6a21\u578b\u5230\u4ea7\u4e1a\u5927\u6a21\u578b\u7684\u8f6c\u53d8\u3002<\/font><\/p>\n

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

\u534e\u4e3a\u4e91\u76d8\u53e4\u9884\u8bad\u7ec3\u5927\u6a21\u578b\u67b6\u6784\uff082022\u5e74\u53ef\u4fe1AI\u5cf0\u4f1a\uff09<\/font><\/p>\n

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

\u300a\u4eba\u5de5\u667a\u80fd\u7814\u53d1\u8fd0\u8425\u4e00\u4f53\u5316\uff08Model\/MLOps\uff09\u80fd\u529b\u6210\u719f\u5ea6\u6a21\u578b\u300b\u6807\u51c6\u4f53\u7cfb\uff08\u4e92\u8054\u7f51\uff09<\/font><\/p>\n

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

\u80fd\u529b\u6210\u719f\u5ea6\u7b49\u7ea7\u5212\u5206\uff08\u4e92\u8054\u7f51\uff09<\/font><\/p>\n

2022\u5e749\u67081\u65e5<\/strong>\uff0c\u534e\u4e3a\u8f6e\u503c\u8463\u4e8b\u957f\u80e1\u539a\u5d11<\/font><\/strong>\u57282022\u4e16\u754c\u4eba\u5de5\u667a\u80fd\u5927\u4f1a\u4e0a\u8868\u793a\uff0c\u4eba\u5de5\u667a\u80fd\u89e6\u53d1\u7684\u4ea7\u4e1a\u53d8\u9769\u6b63\u5728\u6539\u53d8\u6bcf\u4e00\u4e2a\u884c\u4e1a\uff0c\u4eba\u5de5\u667a\u80fd\u4e5f\u5728\u8d8a\u6765\u8d8a\u591a\u7684\u884c\u4e1a\u573a\u666f\u53d1\u6325\u91cd\u8981\u4ef7\u503c\u3002\u57fa\u4e8e\u534e\u4e3a\u4e91\u76d8\u53e4\u9884\u8bad\u7ec3\u5927\u6a21\u578b\u6253\u9020\u7684\u534e\u4e3a\u4e91AI\u8f85\u52a9\u836f\u7269\u8bbe\u8ba1\u5e73\u53f0\uff08Huawei Cloud AI Drug Design Platform\uff09\u83b7\u5f97\u4e86\u201cSAIL\u4e4b\u661f\u201d\u5956\u9879\u3002Drug X\u6709\u671b\u6210\u4e3a\u5168\u7403\u8fd140\u5e74\u6765\u9996\u4e2a\u65b0\u9776\u70b9\u3001\u65b0\u7c7b\u522b\u7684\u6297\u751f\u7d20\uff0c\u5176\u9776\u70b9\u7279\u8d28\u51b3\u5b9a\u4e86\u7ec6\u83cc\u5c06\u96be\u4ee5\u5bf9Drug X\u4ea7\u751f\u8010\u836f\u6027\uff0c\u5bf9\u6297\u759f\uff08\u5373\u759f\u539f\u866b\uff09\u836f\u7269\u7814\u53d1\u7b49\u591a\u4e2a\u9886\u57df\u6709\u7740\u91cd\u8981\u7684\u5f71\u54cd\u3002\u534e\u4e3a\u4e91\u76d8\u53e4\u836f\u7269\u5206\u5b50\u5927\u6a21\u578b\u8ba9\u5148\u5bfc\u836f\u7684\u7814\u53d1\u5468\u671f\u4ece\u6570\u5e74\u7f29\u77ed\u81f3\u4e00\u4e2a\u6708\uff0c\u7814\u53d1\u6210\u672c\u964d\u4f4e70%<\/strong><\/font><\/p>\n

2022\u5e7411\u67087\u65e5<\/strong>\uff0c\u534e\u4e3a\u5168\u8054\u63a5\u5927\u4f1a 2022 \u4e2d\u56fd\u7ad9\u4e0a\uff0c\u534e\u4e3a\u4e91\u7530\u9f50<\/font><\/strong>\u8fdb\u4e00\u6b65\u8fed\u4ee3\u76d8\u53e4\u5927\u6a21\u578b\u7684\u6280\u672f\u80fd\u529b\uff0c\u6269\u5c55\u76d8\u53e4\u5927\u6a21\u578b\u7684\u670d\u52a1\u8303\u56f4\uff0c\u53d1\u5e03\u76d8\u53e4\u6c14\u8c61\u5927\u6a21\u578b\u3001\u76d8\u53e4\u6d77\u6d6a\u5927\u6a21\u578b\u3001\u76d8\u53e4\u77ff\u5c71\u5927\u6a21\u578b\u3001\u76d8\u53e4OCR\u5927\u6a21\u578b\u7b49\u91cd\u78c5\u670d\u52a1\u3002\u534e\u4e3a\u4e91\u53d1\u5e03\u300a\u9884\u8bad\u7ec3\u5927\u6a21\u578b\u767d\u76ae\u4e66<\/strong>\u300b\u3002<\/font><\/p>\n

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

\u626b\u63cf\u4e8c\u7ef4\u7801\u53ef\u4ee5\u4e0b\u8f7d\u767d\u76ae\u4e66\uff0c\u4f46\u662f\u5728\u534e\u4e3a\u767d\u76ae\u4e66\u7f51\u7ad9\u627e\u4e0d\u5230<\/font><\/p>\n

2023\u5e744\u67088\u65e5<\/strong>\uff0c\u4eba\u5de5\u667a\u80fd\u5927\u6a21\u578b\u6280\u672f\u9ad8\u5cf0\u8bba\u575b\uff0c\u534e\u4e3a\u7530\u9f50<\/font><\/strong>\u5c55\u793a\u53e4\u7cfb\u5217 AI \u5927\u6a21\u578b\uff1aNLP\u5927\u6a21\u578b\u3001CV \u5927\u6a21\u578b\u3001\u79d1\u5b66\u8ba1\u7b97\u5927\u6a21\u578b\uff08\u6c14\u8c61\u5927\u6a21\u578b\uff09\u3001\u591a\u6a21\u6001\u5927\u6a21\u578b\u3001\u8bed\u97f3\u5927\u6a21\u578b\u7b49\u591a\u4e2a\u6a21\u578b<\/font><\/p>\n

2023\u5e747\u67085\u65e5<\/strong>\uff0c\u56fd\u9645\u9876\u520aNature\u767b\u8f7d\u4e86\u534e\u4e3a\u4e91\u76d8\u53e4\u6c14\u8c61\u5927\u6a21\u578b\u7814\u53d1\u56e2\u961f\u7814\u7a76\u6210\u679c\uff0c\u6210\u4e3a\u8fd1\u5e74\u6765\u9996\u7bc7\u4ee5\u4e2d\u56fd\u79d1\u6280\u516c\u53f8\u4e3a\u552f\u4e00\u7f72\u540d\u5355\u4f4d\u53d1\u8868\u7684Nature\u6b63\u520a\u8bba\u6587\u3002\u5ba1\u7a3f\u4eba\u9ad8\u5ea6\u8bc4\u4ef7\u8be5\u6a21\u578b\uff1a\u201c\u534e\u4e3a\u4e91\u76d8\u53e4\u6c14\u8c61\u5927\u6a21\u578b\u8ba9\u4eba\u4eec\u5f00\u59cb\u91cd\u65b0\u5ba1\u89c6\u6c14\u8c61\u9884\u62a5\u6a21\u578b\u7684\u672a\u6765\uff0c\u6a21\u578b\u7684\u5f00\u653e\u5c06\u63a8\u52a8\u8be5\u9886\u57df\u7684\u53d1\u5c55\u3002\u201d\u4f5c\u4e3a\u9996\u4e2a\u7cbe\u5ea6\u8d85\u8fc7\u4f20\u7edf\u6570\u503c\u9884\u62a5\u65b9\u6cd5\u7684AI\u9884\u6d4b\u6a21\u578b<\/font><\/strong>\uff0c\u76d8\u53e4\u6c14\u8c61\u5927\u6a21\u578b\u7684\u9884\u6d4b\u901f\u5ea6\u4e5f\u6709\u5927\u5e45\u63d0\u5347\u3002\u539f\u6765\u9884\u6d4b\u4e00\u4e2a\u53f0\u98ce\u672a\u676510\u5929\u7684\u8def\u5f84\uff0c\u9700\u8981\u57283000\u53f0\u670d\u52a1\u5668\u7684\u9ad8\u6027\u80fd\u8ba1\u7b97\u673a\u96c6\u7fa4\u4e0a\u82b1\u8d395\u5c0f\u65f6\u8fdb\u884c\u4eff\u771f\u3002\u73b0\u5728\u57fa\u4e8e\u9884\u8bad\u7ec3\u7684\u76d8\u53e4\u6c14\u8c61\u5927\u6a21\u578b\uff0c\u901a\u8fc7AI\u63a8\u7406\u7684\u65b9\u5f0f\uff0c\u7814\u7a76\u8005\u53ea\u9700\u5355\u53f0\u670d\u52a1\u5668\u4e0a\u5355\u5361\u914d\u7f6e\uff0c10\u79d2\u5185\u5c31\u53ef\u4ee5\u83b7\u5f97\u66f4\u7cbe\u786e\u7684\u9884\u6d4b\u7ed3\u679c\u3002<\/font><\/p>\n

2023\u5e747\u67087\u65e5<\/strong>\uff0c\u534e\u4e3a\u5f00\u53d1\u8005\u5927\u4f1a\uff08HDC 2023\uff09\uff0c\u534e\u4e3a\u4e91CEO\u5f20\u5e73\u5b89<\/font><\/strong>\u6b63\u5f0f\u5bf9\u5916\u53d1\u5e03\u4e86\u76d8\u53e4\u5927\u6a21\u578b3.0\u3002\u7b2c\u4e00\u662f\u201d\u4e0d\u4f5c\u8bd7\uff0c\u53ea\u505a\u4e8b<\/strong>\u201d\u3002\u7b2c\u4e8c\u662f\u201cALL in AI<\/strong>\u201d\u3002\u76d8\u53e4\u5927\u6a21\u578b 3.0 \u5305\u62ec\u300c5+N+X<\/strong><\/font>\u300d\u4e09\u5c42\u67b6\u6784\uff0c\u4e09\u5c42\u5206\u522b\u6307 L0 \u5c42\u7684 5 \u4e2a\u57fa\u7840\u5927\u6a21\u578b\u3001L1 \u5c42\u7684 N \u4e2a\u884c\u4e1a\u901a\u7528\u5927\u6a21\u578b\u3001\u4ee5\u53ca L2 \u5c42\u53ef\u4ee5\u8ba9\u7528\u6237\u81ea\u4e3b\u8bad\u7ec3\u7684\u66f4\u591a\u7ec6\u5316\u573a\u666f\u6a21\u578b\u3002\u5f20\u5e73\u5b89\u5ba3\u5e03\u5355\u96c6\u7fa4 2000P Flops \u7b97\u529b\u7684\u6607\u817e AI \u4e91\u670d\u52a1\u5728\u534e\u4e3a\u4e91\u7684\u4e4c\u5170\u5bdf\u5e03\u548c\u8d35\u5b89 AI \u7b97\u529b\u4e2d\u5fc3\u540c\u65f6\u4e0a\u7ebf\u3002\u6607\u817e AI \u4e91\u670d\u52a1\u9664\u4e86\u652f\u6301\u534e\u4e3a\u5168\u573a\u666f AI \u6846\u67b6\u6607\u601d MindSpore \u5916\uff0c\u8fd8\u652f\u6301 Pytorch\u3001Tensorflow \u7b49\u4e3b\u6d41 AI \u6846\u67b6\u3002\u6607\u817e AI \u4e91\u670d\u52a1\u53ef\u4ee5\u63d0\u4f9b\u66f4\u957f\u7a33\u7684 AI \u7b97\u529b\u670d\u52a1\uff0c\u5343\u5361\u8bad\u7ec3 30 \u5929\u957f\u7a33\u7387\u8fbe\u5230 90%\uff0c\u65ad\u70b9\u6062\u590d\u65f6\u957f\u4e0d\u8d85\u8fc7 10 \u5206\u949f<\/font><\/p>\n

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

<\/font><\/p>\n

\u76d8\u53e4\u5927\u6a21\u578b 3.0 \u662f\u4e00\u4e2a\u9762\u5411\u884c\u4e1a\u7684\u5927\u6a21\u578b\u7cfb\u5217\uff0c\u5305\u62ec\u300c5+N+X\u300d\u4e09\u5c42\u67b6\u6784\uff08\u534e\u4e3a\u5927\u4f1a\uff09<\/font><\/p>\n

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

\u4e94\u4e2a\u57fa\u7840\u5927\u6a21\u578b<\/strong>\uff1a\u5305\u62ec\u81ea\u7136\u8bed\u8a00\u3001\u89c6\u89c9\u3001\u591a\u6a21\u6001\u3001\u9884\u6d4b\u3001\u79d1\u5b66\u8ba1\u7b97\u5927\u6a21\u578b\u3002\u76d8\u53e4 3.0 \u4e3a\u5ba2\u6237\u63d0\u4f9b 100 \u4ebf\u53c2\u6570\u3001380 \u4ebf\u53c2\u6570\u3001710 \u53c2\u6570\u548c 1000 \u4ebf\u53c2\u6570\u7684\u7cfb\u5217\u5316\u57fa\u7840\u5927\u6a21\u578b\uff08\u534e\u4e3a\u7f51\u7ad9\uff09<\/font><\/p>\n

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

\u76d8\u53e4\u5927\u6a21\u578b\u5168\u6808\u6280\u672f\uff08\u534e\u4e3a\u5927\u4f1a\uff09<\/font><\/p>\n

2023\u5e747\u670818\u65e5<\/strong>\uff0c\u5c71\u4e1c\u80fd\u6e90\u96c6\u56e2\u3001\u534e\u4e3a\u516c\u53f8\u8054\u5408\u53d1\u5e03\u5168\u7403\u9996\u4e2a\u5546\u7528\u4e8e\u80fd\u6e90\u884c\u4e1a\u7684AI\u5927\u6a21\u578b\u2014\u2014\u76d8\u53e4\u77ff\u5c71\u5927\u6a21\u578b\u3002\u8be5\u6a21\u578b\u5c06\u51ed\u501f\u201c\u7ecf\u8425\u7ba1\u7406\u4e0e\u667a\u80fd\u751f\u4ea7\u5206\u79bb\u201d\u201c\u6570\u636e\u4e0d\u51fa\u56ed\u533a\u201d\u201c\u652f\u6301\u89c4\u6a21\u590d\u5236\u201d\u201c\u5b66\u4e60\u5206\u6790\u5c0f\u6837\u672c\u201d\u7b49\u80fd\u529b\u7279\u6027\u3002\u7164\u77ff\u9886\u57df\uff0c\u76d8\u53e4\u77ff\u5c71\u5927\u6a21\u578b\u5df2\u7ecf\u5728\u5168\u56fd8\u4e2a\u77ff\u4e95\u89c4\u6a21\u4f7f\u7528\uff0c\u4e00\u4e2a\u5927\u6a21\u578b\u53ef\u4ee5\u8986\u76d6\u7164\u77ff\u7684\u91c7\u3001\u6398\u3001\u673a\u3001\u8fd0\u3001\u901a\u3001\u6d17\u9009\u7b49\u4e1a\u52a1\u6d41\u7a0b\u4e0b\u76841000\u591a\u4e2a\u7ec6\u5206\u573a\u666f\uff0c\u8ba9\u66f4\u591a\u7684\u7164\u77ff\u5de5\u4eba\u80fd\u591f\u5728\u5730\u9762\u4e0a\u4f5c\u4e1a\uff0c\u4e0d\u4ec5\u80fd\u8ba9\u7164\u77ff\u5de5\u4eba\u7684\u5de5\u4f5c\u73af\u5883\u66f4\u52a0\u8212\u9002\uff0c\u800c\u4e14\u53ef\u4ee5\u6781\u5927\u5730\u51cf\u5c11\u5b89\u5168\u4e8b\u6545<\/font><\/p>\n

<\/font><\/p>\n

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

\u534e\u4e3a\u76d8\u53e4\u77ff\u5c71\u5927\u6a21\u578b\uff0c\u9700\u8981\u6ce8\u610f\u8fd9\u662f\u4e00\u4e2a\u79c1\u6709\u5316\u90e8\u7f72\uff08\u534e\u4e3a\u7f51\u7ad9\uff09<\/font><\/p>\n

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

\u534e\u4e3a\u5927\u6a21\u578b\u57fa\u77f3\uff08\u534e\u4e3a\u7f51\u7edc\u6750\u6599\uff09<\/font><\/p>\n


\n<\/font><\/p>\n

\u4e8c\u3001\u76d8\u53e4\u79d1\u5b66\u8ba1\u7b97\u5927\u6a21\u578b<\/font><\/strong><\/p>\n

\u76d8\u53e43.0\u53d1\u5e03\u524d\uff0c\u534e\u4e3a\u53d1\u5e03\u4e86\u4e00\u4e2a\u300a\u9884\u8bad\u7ec3\u5927\u6a21\u578b\u767d\u76ae\u4e66\u300b\uff0c\u8fd9\u4e2a\u767d\u76ae\u4e66\u4e2d\u4ecb\u7ecd\u4e86\u79d1\u5b66\u8ba1\u7b97\u5927\u6a21\u578b\u7684\u4e00\u4e9b\u5185\u5bb9\uff0c\u6458\u5f55\u90e8\u5206\uff0c\u6709\u52a9\u4e8e\u7406\u89e3\u5176\u5185\u6db5\u3002<\/font><\/p>\n

\u5728\u4eba\u5de5\u667a\u80fd\u4e4b\u524d\uff0c\u79d1\u5b66\u5bb6\u4eec\u901a\u5e38\u901a\u8fc7\u5206\u6790\u5b9e\u9a8c\u6570\u636e\u4e0e\u63a8\u6f14\u673a\u7406\u516c\u5f0f\u7684\u65b9\u5f0f\u63d0\u70bc\u81ea\u7136\u79d1\u5b66\u95ee\u9898\uff08\u6e4d\u6d41\u6a21\u62df\u3001\u5929\u6c14\u9884\u62a5\u3001\u5927\u5f62\u53d8\u5e94\u529b\u7b49\uff09\u7684\u5185\u5728\u89c4\u5f8b\u3002\u8fd9\u4e9b\u4f20\u7edf\u65b9\u6cd5\uff0c\u5bb9\u6613\u5728\u5927\u4f53\u91cf\u3001\u9ad8\u7ef4\u5ea6\u7684\u6570\u636e\u5904\u7406\u4e0a\u9047\u5230\u56f0\u96be\u3002\u8fd1\u5e74\u6765\uff0c\u968f\u7740\u4eba\u5de5\u667a\u80fd\u6280\u672f\u7684\u98de\u901f\u53d1\u5c55\uff0c\u4e1a\u754c\u6d8c\u73b0\u51fa\u4e86 AI+ \u79d1\u5b66\u8ba1\u7b97\u7c7b\u65b9\u6cd5\uff0c\u5373\u4f7f\u7528\u5d4c\u5165\u5404\u7c7b\u79d1\u5b66\u65b9\u7a0b\u7684\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\uff0c\u4ece\u89c2\u6d4b\u6570\u636e\u548c\u4eff\u771f\u6570\u636e\u4e2d\u76f4\u63a5\u5b66\u4e60\u95ee\u9898\u8574\u542b\u7684\u89c4\u5f8b\uff0c\u4ee5\u5bf9\u590d\u6742\u7684\u79d1\u5b66\u6570\u636e\u8fdb\u884c\u5206\u6790\uff0c\u4e86\u89e3\u79d1\u5b66\u8fc7\u7a0b\u7684\u5185\u90e8\u673a\u7406\u3002\u4ece\u9884\u8bad\u7ec3\u5927\u6a21\u578b\u7684\u89d2\u5ea6\u770b\uff0c\u79d1\u5b66\u8ba1\u7b97\u5927\u6a21\u578b\u4e0eNLP\u3001CV\u7b49\u5927\u6a21\u578b\u5b58\u5728\u82e5\u5e72\u76f8\u4f3c\u4e4b\u5904\u3002\u5b83\u4eec\u90fd\u5efa\u7acb\u5728\u5927\u89c4\u6a21\u6570\u636e\u96c6\u4e0a\uff0c\u90fd\u9700\u8981\u8bbe\u8ba1\u5927\u53c2\u6570\u91cf\u7684\u795e\u7ecf\u7f51\u7edc\uff0c\u90fd\u9700\u8981\u590d\u6742\u7684\u4f18\u5316\u8fc7\u7a0b\uff0c\u6700\u540e\u5c06\u77e5\u8bc6\u5b58\u50a8\u5728\u7f51\u7edc\u7684\u53c2\u6570\u4e4b\u4e2d\u3002<\/font><\/em><\/p>\n

1\u3001\u6570\u636e\u6765\u6e90<\/strong>\u3002\u4e0d\u540c\u79d1\u5b66\u8ba1\u7b97\u573a\u666f\u7684\u89c2\u6d4b\u6570\u636e\u5f80\u5f80\u76f8\u5dee\u5de8\u5927\uff0c\u89c2\u6d4b\u6570\u636e\u7684\u6536\u96c6\u5f80\u5f80\u9700\u8981\u7279\u5b9a\u9886\u57df\u7684\u4e13\u4e1a\u4eea\u5668\u4e0e\u7cfb\u7edf\u7684\u5b9e\u9a8c\uff0c\u4f8b\u5982\u86cb\u767d\u8d28\u7ed3\u6784\u9884\u6d4b\u95ee\u9898\u4e2d\u86cb\u767d\u8d28\u7ed3\u6784\u7684\u6d4b\u5b9a\u9700\u8981\u4f9d\u8d56\u4e8e X \u5c04\u7ebf\u884d\u5c04\u65b9\u6cd5\u4e0e\u6838\u78c1\u5171\u632f\u6cd5\u3001\u77ed\u4e34\u964d\u96e8\u9884\u62a5\u95ee\u9898\u4e2d\u9700\u8981\u6c14\u8c61\u96f7\u8fbe\u6536\u96c6\u7684\u96f7\u8fbe\u6ce2\u53cd\u5c04\u7387\u6570\u636e\u3001\u690d\u7269\u8868\u578b\u5206\u6790\u95ee\u9898\u4e2d\u6570\u636e\u5219\u6765\u81ea\u4e8e\u5b9e\u9a8c\u5458\u7684\u6536\u96c6\uff0c\u7b49\u7b49\u3002\u5728\u4e00\u4e9b\u79d1\u5b66\u8ba1\u7b97\u573a\u666f\u4e2d\uff0c\u89c2\u6d4b\u6570\u636e\u7684\u6570\u636e\u91cf\u975e\u5e38\u5e9e\u5927\uff0c\u4f8b\u5982\u6c14\u8c61\u6570\u636e\u4e2d\u7684\u5168\u7403\u6c14\u8c61\u7ad9\u5386\u53f2\u6570\u636e\u3001\u536b\u661f\u6570\u636e\u548c\u96f7\u8fbe\u56de\u6ce2\u6570\u636e\u3002\u4e5f\u6709\u4e00\u4e9b\u573a\u666f\u4e2d\uff0c\u89c2\u6d4b\u6570\u636e\u91cf\u76f8\u5bf9\u8f83\u5c11\uff0c\u4f8b\u5982\u7ed3\u6784\u5e94\u529b\u5206\u6790\u529b\u4f20\u611f\u5668\u6536\u96c6\u7684\u6570\u636e\u3002
\n\u4eff\u771f\u6570\u636e\u6765\u81ea\u4e8e\u6570\u503c\u4eff\u771f\u7b97\u6cd5\u7684\u8f93\u51fa\uff0c\u8574\u542b\u7740\u4e30\u5bcc\u7684\u6570\u5b66\u7269\u7406\u4fe1\u606f\uff0c\u540c\u4e00\u4e2a\u95ee\u9898\u4f7f\u7528\u4e0d\u540c\u7684\u4eff\u771f\u7b97\u6cd5\u53ef\u4ee5\u8f93\u51fa\u4e0d\u540c\u7684\u4eff\u771f\u6570\u636e\u3002\u4eff\u771f\u6570\u636e\u4e0d\u540c\u4e8e\u89c2\u6d4b\u6570\u636e\uff0c\u5176\u7cbe\u5ea6\u53d7\u9650\u4e8e\u4f7f\u7528\u4eff\u771f\u7b97\u6cd5\u7684\u51c6\u786e\u6027\u548c\u4eff\u771f\u8ba1\u7b97\u7684\u7b97\u529b\u591a\u5c11\u3002\u76f8\u5bf9\u4e8e\u89c2\u6d4b\u6570\u636e\uff0c\u4eff\u771f\u6570\u636e\u901a\u5e38\u6570\u636e\u91cf\u66f4\u5927\uff08\u53d6\u51b3\u4e8e\u4eff\u771f\u65f6\u4f7f\u7528\u7684\u7b97\u529b\uff09\uff0c\u540c\u65f6\u7f3a\u7701\u503c\u8f83\u5c11\uff0c\u53ef\u4ee5\u4f5c\u4e3a\u89c2\u6d4b\u6570\u636e\u7684\u6709\u6548\u6269\u5145\u3002\u6709\u4e9b\u573a\u666f\u4e2d\uff0c\u89c2\u6d4b\u6570\u636e\u548c\u4eff\u771f\u6570\u636e\u7531\u7279\u5b9a\u7684\u673a\u7406\u77e5\u8bc6\u7ed3\u5408\u5728\u4e00\u8d77\uff0c\u751f\u6210\u878d\u5408\u6570\u636e\u3002\u5982\u6c14\u8c61\u518d\u5206\u6790\u6570\u636e\uff0c\u518d\u5206\u6790\u6570\u636e\u901a\u5e38\u4f7f\u7528\u540c\u5316\u7b97\u6cd5\u878d\u5408\u4eff\u771f\u6570\u636e\u548c\u5b9e\u9a8c\u6570\u636e\u5f97\u5230\u7ed3\u6784\u5316\u7684\u6570\u636e\uff0c\u6839\u636e\u4e0d\u540c\u540c\u5316\u7b97\u6cd5\u4e0e\u4f7f\u7528\u7684\u4eff\u771f\u6570\u636e\u4e5f\u53ef\u4ee5\u6709\u4e0d\u540c\u7684\u7ed3\u679c\u3002<\/font><\/em><\/p>\n

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

2\u3001\u6a21\u578b\u6784\u5efa<\/strong>\u3002\u6839\u636e\u8f93\u5165\u6570\u636e\u7684\u6027\u8d28\uff0c\u7b97\u6cd5\u4f1a\u9009\u7528\u4e0d\u540c\u7684\u57fa\u7840\u6a21\u578b\u7528\u4e8e\u8bad\u7ec3\u3002\u4ee5\u6d77\u6d6a\u9884\u6d4b\u4efb\u52a1\u4e3a\u4f8b\uff0c\u5176\u76ee\u6807\u4e3a\u9884\u6d4b\u5168\u7403\u8303\u56f4\u5185\u6d77\u5e73\u9762\u7684\u5b9e\u65f6\u6d6a\u9ad8\uff0c\u8f93\u5165\u548c\u8f93\u51fa\u6570\u636e\u5747\u4e3a\u5e26\u6709\u65f6\u95f4\u6233\u7684\u4e8c\u7ef4\u7403\u9762\u6570\u636e\uff0c\u56e0\u6b64\u9002\u5408\u4f7f\u7528\u4e8c\u7ef4\u7f51\u7edc\u6a21\u578b\u3002\u5982\u679c\u5c06\u9884\u6d4b\u8303\u56f4\u6269\u5c55\u81f3\u4e09\u7ef4\u7a7a\u95f4\uff0c\u5982\u8fdb\u884c\u5168\u7403\u8303\u56f4\u5185\u7684\u6c14\u8c61\u9884\u6d4b\uff0c\u8f93\u51fa\u548c\u8f93\u51fa\u5747\u4e3a\u5e26\u6709\u65f6\u95f4\u6233\u7684\u4e09\u7ef4\u6570\u636e\uff08\u5305\u62ec\u9ad8\u5ea6\uff09\uff0c\u5219\u9002\u5408\u4f7f\u7528\u4e09\u7ef4\u7f51\u7edc\u6a21\u578b\u3002\u4e8c\u7ef4\u7f51\u7edc\u548c\u4e09\u7ef4\u7f51\u7edc\u5747\u53ef\u4ee5\u501f\u9274\u8ba1\u7b97\u673a\u89c6\u89c9\u9886\u57df\u7684\u76f8\u5e94\u6a21\u578b\uff0c\u5982\u4f7f\u7528\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u6216\u8005\u89c6\u89c9 Transformer \u4f5c\u4e3a\u9aa8\u5e72\u67b6\u6784\uff0c\u914d\u5408\u5927\u6570\u636e\u8fdb\u884c\u9884\u8bad\u7ec3\u3002
\n\u79d1\u5b66\u8ba1\u7b97\u7684\u4e00\u4e2a\u663e\u8457\u7279\u70b9\uff0c\u662f\u53ef\u4ee5\u5229\u7528\u4eba\u7c7b\u5728\u76f8\u5e94\u95ee\u9898\u4e0a\u79ef\u7d2f\u7684\u7ecf\u9a8c\uff0c\u800c\u8fd9\u4e9b\u7ecf\u9a8c\u901a\u5e38\u4f1a\u65bd\u52a0\u5728\u8f93\u51fa\u6570\u636e\u4e0a\uff0c\u6210\u4e3a\u67d0\u79cd\u7ea6\u675f\u6027\u8d28\u7684\u504f\u5fae\u5206\u65b9\u7a0b\u7ec4\u3002\u5982\u4e0b\u56fe\u6240\u793a\uff0c\u6211\u4eec\u53ef\u4ee5\u5c06\u6b64\u7c7b\u65b9\u7a0b\u7ec4\u5d4c\u5165\u795e\u7ecf\u7f51\u7edc\u4e2d\uff0c\u8f85\u52a9\u6a21\u578b\u67b6\u6784\u8bbe\u8ba1\u6216\u6210\u4e3a\u989d\u5916\u7684\u7ea6\u675f\u6761\u4ef6\uff0c\u4e0e\u6807\u51c6\u89c2\u6d4b\u6216\u4eff\u771f\u6570\u636e\u4e00\u8d77\uff0c\u8bad\u7ec3\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\u3002\u5728\u826f\u597d\u7684\u5b9e\u73b0\u4e0b\uff0c\u8fd9\u7c7b\u77e5\u8bc6\u901a\u5e38\u80fd\u591f\u589e\u5f3a\u6a21\u578b\u7684\u9c81\u68d2\u6027\uff0c\u964d\u4f4e\u6a21\u578b\u62df\u5408\u8bad\u7ec3\u6570\u636e\u7684\u96be\u5ea6\u548c\u4e0d\u7a33\u5b9a\u6027\u3002<\/font><\/em><\/p>\n

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

\u5de6\u56fe\u4e3a\u5d4c\u5165\u504f\u5fae\u5206\u65b9\u7a0b\u7684\u795e\u7ecf\u7f51\u7edc\u793a\u610f\u56fe\uff0c\u53f3\u56fe\u4e3a\u6d77\u6d6a\u9884\u62a5\u95ee\u9898\u4f7f\u7528\u7684\u504f\u5fae\u5206\u65b9\u7a0b<\/font><\/p>\n

3\u3001\u6d77\u6d6a\u5927\u6a21\u578b<\/strong>\u3002\u6d77\u6d6a\u9884\u6d4b\u95ee\u9898\u7684\u8f93\u5165\u8f93\u51fa\u90fd\u662f\u7ecf\u7eac\u5ea6\u7f51\u683c\u70b9\u4e0a\u7684\u6c14\u8c61\u8981\u7d20\u6570\u636e\uff0c\u5728\u6570\u636e\u5f62\u5f0f\u4e0a\u4e0e\u89c6\u9891\u6570\u636e\u76f8\u4f3c\u3002\u4e0d\u540c\u4e4b\u5904\u5728\u4e8e\uff0c\u89c6\u9891\u6570\u636e\u6bcf\u4e2a\u5143\u6570\u636e\u662f 0-255 \u7684\u50cf\u7d20\u503c\uff0c\u800c\u98ce\u901f\u3001\u5730\u5f62\u3001\u6d77\u6d6a\u9ad8\u5ea6\u7b49\u6570\u636e\u7684\u6bcf\u4e2a\u5143\u6570\u636e\u5747\u4e3a\u6d6e\u70b9\u6570\u3002\u540c\u65f6\uff0c\u6d77\u6d6a\u9884\u6d4b\u7684\u8f93\u51fa\u901a\u5e38\u4e0d\u662f\u67d0\u79cd\u5206\u7c7b\uff0c\u800c\u662f\u8fde\u7eed\u7684\u9884\u6d4b\u503c\uff0c\u56e0\u6b64\u9700\u8981\u7528\u56de\u5f52\u635f\u5931\u66ff\u6362\u6df1\u5ea6\u5b66\u4e60\u4e2d\u5e38\u7528\u7684\u5206\u7c7b\u3001\u5206\u5272\u635f\u5931\u3002\u6b64\u5916\uff0c\u6d77\u6d6a\u6570\u636e\u548c\u89c6\u9891\u6570\u636e\u76f8\u6bd4\uff0c\u5e76\u4e0d\u6ee1\u8db3\u5e73\u79fb\u5bf9\u79f0\u6027\u7b49\uff0c\u4f46\u6ee1\u8db3\u7403\u5750\u6807\u6761\u4ef6\u4e0b\u7684\u4e00\u7cfb\u5217\u4e0d\u53d8\u6027\uff0c\u4f8b\u5982\u7ed5\u5730\u8f74\u65cb\u8f6c\uff0c\u56e0\u6b64\u9700\u8981\u9009\u5b9a\u6ee1\u8db3\u7279\u5b9a\u4e0d\u53d8\u6027\u7684 CNN \u6216\u8005 Transformer \u67b6\u6784\u3002
\n\u76d8\u53e4\u6d77\u6d6a\u9884\u6d4b\u6a21\u578b\u7684\u4e3b\u4f53\u662f\u8003\u8651\u4e86\u65cb\u8f6c\u4e0d\u53d8\u6027\u7684\u89c6\u89c9 Transformer \u67b6\u6784\uff0c\u53c2\u6570\u91cf\u7ea6\u4e3a\u4e94\u4ebf\u3002<\/strong>\u5982\u4e0a\u6240\u8ff0\uff0c\u795e\u7ecf\u7f51\u7edc\u7684\u635f\u5931\u51fd\u6570\u7531\u4e24\u90e8\u5206\u7ec4\u6210\uff0c\u5373\u5b9e\u9645\u6570\u636e\u4e0a\u7684\u9884\u6d4b\u8bef\u5dee\u548c\u6d77\u6d6a\u9884\u6d4b\u672c\u8eab\u9700\u8981\u6ee1\u8db3\u7684\u504f\u5fae\u5206\u65b9\u7a0b\u3002\u901a\u8fc7\u722c\u53d6\u5168\u7403\u8fd110\u5e74\u7684\u5b9e\u65f6\u6d77\u6d6a\u9ad8\u5ea6\u6570\u636e\u8fdb\u884c\u8bad\u7ec3\uff0c\u6a21\u578b\u5728\u9a8c\u8bc1\u96c6\u4e0a\u9884\u6d4b\u7684\u5e73\u5747\u8bef\u5dee\u5c0f\u4e8e5cm\uff0c\u4e0e\u4f20\u7edf\u9884\u6d4b\u65b9\u6cd5\u76f8\u5f53\uff0c\u5b8c\u5168\u53ef\u4ee5\u6ee1\u8db3\u5b9e\u9645\u5e94\u7528\u9700\u6c42\u3002\u66f4\u91cd\u8981\u7684\u662f\uff0c AI \u7b97\u6cd5\u7684\u9884\u6d4b\u65f6\u95f4\u8f83\u4f20\u7edf\u65b9\u6cd5\u5927\u5e45\u51cf\u5c11\uff1a\u5728\u5355\u5f20\u534e\u4e3a\u6607\u817e\u82af\u7247\u4e0a\uff0c1s \u4e4b\u5185\u5373\u53ef\u5f97\u5230\u5168\u7403\u6d77\u6d6a\u9ad8\u5ea6\u9884\u6d4b\uff0c 1 \u5206\u949f\u5185\u80fd\u591f\u5b8c\u6210\u8d85\u8fc7 100 \u6b21\u6d77\u6d6a\u9884\u6d4b\u4efb\u52a1<\/strong>\uff0c\u63a8\u7406\u6548\u7387\u8f83\u4f20\u7edf\u65b9\u6cd5\u63d0\u5347\u4e86 4-5 \u4e2a\u6570\u91cf\u7ea7\u3002\u4f7f\u7528 AI \u7b97\u6cd5\uff0c\u6211\u4eec\u53ef\u4ee5\u8fc5\u901f\u5f97\u5230\u4e0d\u540c\u53ef\u80fd\u7684\u98ce\u901f\u6761\u4ef6\u4e0b\u7684\u6d77\u6d6a\u9ad8\u5ea6\uff0c\u4ece\u800c\u8fdb\u884c\u5b9e\u65f6\u9884\u6d4b\u548c\u672a\u6765\u60c5\u51b5\u6a21\u62df\uff0c\u5bf9\u4e8e\u6e14\u4e1a\u517b\u6b96\u3001\u707e\u5bb3\u9632\u63a7\u7b49\u573a\u666f\u6709\u6781\u5927\u7684\u4ef7\u503c<\/em>\u3002<\/font><\/p>\n


\n<\/font><\/p>\n

\u4e09\u3001\u300a<\/strong><\/font>Accurate medium-range global weather forecasting with 3D neural networks<\/strong><\/font><\/a>\u300b<\/strong>
\n<\/font><\/p>\n

\u5728\u534e\u4e3a\u7684\u6a21\u578b\u4f53\u7cfb\u4e2d\uff0c\u6c14\u8c61\u5927\u6a21\u578b\u662fL1\u5c42\uff0c\u800c\u53f0\u98ce\u9884\u6d4b\u548c\u6d77\u6d6a\u9884\u6d4b\u5219\u662fL2\u5c42\uff0c\u8fd9\u79cd\u5206\u5c42\u5728\u903b\u8f91\u4e0a\u6709\u5176\u4ef7\u503c\uff0c\u4f46\u662f\u770b\u4e0a\u53bb\u672a\u5fc5\u6709L2\u5927\u6a21\u578b\uff0c\u66f4\u50cf\u662fL1\u6a21\u578b\u7684\u5e94\u7528\u3002\u6458\u5f55\u4e00\u4e9b\u8bba\u6587\u5185\u5bb9\uff0c\u82f1\u6587\u7684\u770b\u7684\u66f4\u660e\u767d\u3002<\/font><\/p>\n

In the past decade, high-performance computing systems have greatly accelerated research in the field of numerical weather prediction (NWP)<\/strong> methods. Conventional NWP methods are primarily concerned with describing the transitions between discretized grids of atmospheric states using partial differential equations (PDEs) and then solving them with numerical simulations. These methods are often slow; a single simulation for a ten-day forecast can take hours of computation in a supercomputer that has hundreds of nodes<\/font><\/strong>. In addition, conventional NWP algorithms largely rely on parameterization, which uses approximate functions to capture unresolved processes, where errors can be introduced by approximation. \uff08\u4f20\u7edf\u65b9\u6cd5\u662f\u5229\u7528\u8d85\u7b97\u89e3\u504f\u5fae\u5206\u65b9\u7a0b\uff09<\/font><\/p>\n

First, we integrated height information into a new dimension so that the input and output of our deep neural networks can be conceptualized in three dimensions. We further designed a three-dimensional (3D) Earth-specific transformer (3DEST)<\/font><\/strong> architecture to inject Earth-specific priors into the deep networks. Our experiments show that 3D models, by formulating height into an individual dimension, have the ability to capture the relationship between atmospheric states in different pressure levels and thus yield significant accuracy<\/font><\/strong> gains, compared with two-dimensional models such as FourCastNet<\/strong>. Second, we applied a hierarchical temporal aggregation algorithm that involves training a series of models with increasing forecast lead times<\/font><\/strong>. Hence, in the testing stage, the number of iterations needed for medium-range weather forecasting was largely reduced, and the cumulative forecast errors were alleviated. Experiments on the fifth generation of ECMWF reanalysis (ERA5) data18 validated that Pangu-Weather is good at deterministic forecast and extreme weather forecast while being more than 10,000-times faster than the operational IFS<\/strong><\/font> ( integrated forecasting system). \uff08\u4e24\u5927\u8d21\u732e\uff0c\u5176\u4e2d\u63d0\u5347\u7cbe\u5ea6\u4f9d\u9760\u4e863D\u5efa\u6a21\u3001\u63d0\u5347\u9884\u6d4b\u80fd\u529b\u91c7\u7528\u4e864\u4e2a\u6a21\u578b\uff09<\/font><\/p>\n

We established our weather forecast system via deep learning. The methodology involves training deep neural networks to take reanalysis weather data at a given point in time as input, and then produce reanalysis weather data at a future point in time as output. We used a single point in time for both input and output. The time resolution of the ERA5 data is 1\u2009h; in the training subset (1979\u20132017), there were as many as 341,880 time points, the amount of training data in one epoch<\/font><\/strong>. To alleviate the risk of over-fitting, we randomly permuted the order of sample from the training data at the start of each epoch. We trained four deep networks with lead times (the time difference between input and output) at 1\u2009h, 3\u2009h, 6\u2009h and 24\u2009h, respectively. Each of the four deep networks was trained for 100 epochs, and each of them takes approximately 16\u2009days on a cluster of 192 NVIDIA Tesla-V100 GPUs<\/font><\/strong>.\uff08\u6ce8\u610f\uff1a\u4f7f\u7528\u4e86Nvidia\u7684V100\u5361\uff0c\u4e0d\u662fA100\u4e5f\u4e0d\u662f\u6607\u817e\uff09<\/font><\/p>\n

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

3DEST architecture. Based on the standard encoder\u2013decoder design of vision transformers, we adjusted the shifted-window mechanism19<\/a><\/sup> and applied an Earth-specific positional bias. b<\/b>, Hierarchical temporal aggregation. Once given a lead time, we used a greedy algorithm to perform forecasting with as few steps as possible. We use FM1, FM3, FM6 and FM24 to indicate the forecast models with lead times being 1\u2009h, 3\u2009h, 6\u2009h or 24\u2009h, respectively. A<\/b>0<\/sub> is the input weather state and A^A^t denotes the predicted weather state at time t<\/i> (in hours).<\/font><\/p>\n


\n<\/font><\/p>\n

This architecture is known as the 3D Earth-specific transformer (3DEST). We fed all included weather variables, including 13 layers of upper-air variables and the surface variables, into a single deep network. We then performed patch embedding to reduce the spatial resolution and combined the down-sampled data into a 3D cube. The 3D data are propagated through an encoder\u2013decoder architecture derived from the Swin transformer19<\/a><\/sup><\/font><\/strong>\uff08\u5fae\u8f6f\u4e9a\u6d32\u9662\u63d0\u51fa\u7684\u6a21\u578b\uff0c\u9488\u5bf9CV\u8bc6\u522b\u505a\u4e86\u7a97\u53e3\u3001\u5e73\u79fb\u548c\u6279\u5904\u7406\u4f18\u5316\uff09, a variant of a vision transformer20<\/a><\/sup>, which has 16 blocks. The output is split into upper-air variables and surface variables and is up-sampled with patch recovery to restore the original resolution. To inject Earth-specific priors into the deep network, we designed an Earth-specific positional bias (a mechanism of encoding the position of each unit; detailed in <\/strong><\/font>Methods<\/font><\/strong><\/a>) to replace the original relative positional bias of Swin<\/strong>. This modification increases the number of bias parameters by a factor of 527, with each 3D deep network containing approximately 64\u2009million parameters<\/strong><\/font>. Compared with the baseline, however, 3DEST has the same computational cost and has a faster convergence speed. \uff08\u6bd5\u7adf\u662f\u4e00\u4e2a\u591a\u5c42\u7684\u6c14\u8c61\u7403\u4f53\uff0c\u91c7\u7528\u56fe\u50cf\u5206\u5757\u5904\u7406\u6ca1\u6bdb\u75c5\uff09<\/font><\/p>\n

The lead time of a medium-range weather forecast is 7\u2009days or longer<\/font><\/strong>. This prompted us to call the base deep networks (lead times being 1\u2009h, 3\u2009h, 6\u2009h or 24\u2009h) iteratively, using each forecasted result as the input of the next step. To reduce the cumulative forecast errors, we introduced hierarchical temporal aggregation, a greedy algorithm that always calls for the deep network with the largest affordable lead time. Mathematically, this greatly reduces the number of iterations. As an example, when the lead time was 56\u2009h, we would execute the 24-hour forecast model 2 times, the 6-hour forecast model 1 time, and the 1-hour forecast model 2 times<\/font><\/strong> (Fig. <\/font>1b<\/font><\/a>). Compared with FourCastNet2<\/a><\/sup>, which uses a fixed 6-hour forecast model, our method is faster and more accurate. The limitation of this strategy is discussed in <\/font>Methods<\/font><\/a>. \uff08\u8fd9\u662f\u5bf9\u6a21\u578b\u7684\u4f7f\u7528\uff09<\/font><\/p>\n

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

To fairly compare Pangu-Weather against FourCastNet2<\/a><\/sup>, we trained our 3D deep networks on 39\u2009years of data (from 1979 to 2017), validated them on 2019 data and tested them on 2018 data. We studied 69 factors, including 5 upper-air variables at 13 pressure levels (50\u2009hPa, 100\u2009hPa, 150\u2009hPa, 200\u2009hPa, 250\u2009hPa, 300\u2009hPa, 400\u2009hPa, 500\u2009hPa, 600\u2009hPa, 700\u2009hPa, 850\u2009hPa, 925\u2009hPa and 1,000\u2009hPa) and 4 surface variables. When tested against reanalysis data, for each tested variable, Pangu-Weather produces a lower RMSE<\/font><\/strong> and a higher ano<\/font>maly correlation coefficient (ACC) <\/font><\/strong>than the operational IFS and FourCastNet, the best NWP and AI-based methods, respectively. \uff08\u8fd9\u90e8\u5206\u9700\u8981\u6c14\u8c61\u5b66\u77e5\u8bc6\u4e86\uff09<\/font><\/p>\n

\u3010\u53f0\u98ce\u8f68\u8ff9\u3011<\/strong>Next, we used Pangu-Weather to track tropical cyclones. Given an initial time point, we set the lead time to be multiples of 6\u2009h (ref. 23<\/a><\/sup>) and initiated Pangu-Weather to forecast future weather states. We looked for the local minimum of MSLP that satisfied certain conditions, such as the cyclone eye. The tracking algorithm is described in the supplementary material for this paper. We used the International Best Track Archive for Climate Stewardship (IBTrACS) project24<\/a>,25<\/a><\/sup>, which contains the best available estimations for tropical cyclones.<\/font><\/p>\n

We compared Pangu-Weather with ECMWF-HRES, a strong cyclone tracking method based on high-resolution (9\u2009km\u2009\u00d7\u20099\u2009km) operational weather forecasting<\/strong>. We chose 88 named tropical cyclones in 2018 that appear in both IBTrACS and ECMWF-HRES. As shown in Fig. <\/font>4<\/font><\/a>, Pangu-Weather statistically produced more accurate tracking results than ECMWF-HRES for these cyclones. The 3-day and 5-day mean direct position errors for cyclone eyes were reported at 120.29\u2009km and 195.65\u2009km for Pangu-Weather, which are smaller than 162.28\u2009km and 272.10\u2009km for ECMWF-HRES, respectively<\/font><\/strong>. The breakdowns of tracking errors with respect to regions and intensities are provided in Extended Data Fig. <\/font>5<\/font><\/a>. The advantage of Pangu-Weather becomes more significant as the lead time increases. We also show the tracking results of the two strongest cyclones in the western Pacific, Kong-rey and Yutu, in Fig. <\/font>4<\/font><\/a>. See the supplementary material for a detailed analysis.<\/font><\/p>\n

\u3010\u63a8\u7406\u901f\u5ea6\u3011<\/font>The inference speed of Pangu-Weather is comparable to that of FourCastNet2<\/a><\/sup>. In a system-level comparison, FourCastNet requires 0.28\u2009s for inferring a 24-hour forecast on a Tesla-A100 GPU (312\u2009teraFLOPS), whereas Pangu-Weather needs 1.4\u2009s on a Tesla-V100 GPU (120\u2009teraFLOPS). Taking GPU performance into consideration, Pangu-Weather is about 50% slower than FourCastNet. Pangu-Weather is more than 10,000-times faster than the operational IFS, which requires several hours in a supercomputer with hundreds of nodes.<\/font><\/h5>\n
\u3010\u76d8\u53e4\u7531\u6765\u3011<\/font>Pangu is a primordial being and creation figure in Chinese mythology who separated heaven and earth and became geographic features such as mountains and rivers (see <\/font>https:\/\/en.wikipedia.org\/wiki\/Pangu<\/font><\/a>). Pangu is also a series of pre-trained AI models developed by Huawei Cloud that covers computer vision, natural language processing, multimodal understanding, scientific computing (including weather forecasting) and so on.<\/font><\/h5>\n

\u3010\u9884\u8bad\u7ec3\u6a21\u578b\u3011<\/strong>We released the trained models, inference code and the pseudocode of details to the public at a GitHub repository: <\/font>https:\/\/github.com\/198808xc\/Pangu-Weather<\/font><\/a> (<\/font>https:\/\/doi.org\/10.5281\/zenodo.7678849<\/font><\/a>).<\/font><\/p>\n

\u3010\u6f14\u8bb2\u6750\u6599\u3011<\/strong> https:\/\/pan.baidu.com\/s\/14ZGywcr4XAK5dk75-8PUqA?pwd=9sco<\/a><\/font><\/font><\/p>\n

\u6a21\u578b\u7684\u8be6\u7ec6\u4ecb\u7ecd\u53ef\u4ee5\u770b\u539f\u6587\uff1a<\/font>https:\/\/www.nature.com\/articles\/s41586-023-06185-3<\/font><\/a><\/p>\n

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

<\/font><\/p>\n


\n<\/font><\/strong><\/p>\n

\u5c0f\u7ed3\uff1a<\/font><\/strong><\/p>\n

\u770b\u770b\u8bba\u6587\u539f\u6587\u6709\u52a9\u4e8e\u8fa8\u522b\u4e00\u4e9b\u7f51\u4e0a\u81ea\u5a92\u4f53\u8f6f\u6587\u7684\u8bef\u5bfc\uff0c3DEST\u662f\u4e00\u4e2a\u57fa\u4e8e\u524d\u4eba\u5de5\u4f5c\u57fa\u7840\u7684\u521b\u65b0\uff0c\u628a\u6570\u636e\u5206\u5c42\u63d0\u5347\u4e86\u9884\u6d4b\u7cbe\u5ea6\uff0c\u800c\u8bad\u7ec3\u4e0d\u540c\u65f6\u95f4\u6bb5\u7684\u6a21\u578b\u867d\u7136\u4e0d\u591f\u4f18\u96c5\uff0c\u4f46\u662f\u89e3\u51b3\u4e2d\u671f\u6c14\u8c61\u9884\u6d4b\u7684\u4e2d\u53ef\u80fd\u9020\u6210\u7684\u7d2f\u8ba1\u8bef\u5dee\uff0c\u63d0\u5347\u4e86\u7ed3\u679c\u7684\u51c6\u786e\u6027\u3002\u80fd\u591f\u767b\u4e0aNature\u7684\u539f\u56e0\u4e0d\u662f\u56e0\u4e3a\u6a21\u578b\u6709\u591a\u5148\u8fdb\uff0c\u662f\u5de5\u7a0b\u5316\u548c\u7b97\u6cd5\u7684\u534f\u540c\u521b\u65b0\uff0c\u89e3\u51b3\u4e86\u9884\u6d4b\u9886\u57df\u7684\u5b9e\u9645\u95ee\u9898\uff0c\u7ed3\u679c\u8fd8\u7279\u522b\u597d\u3002\u534e\u4e3a\u5728\u4e00\u4e9b\u5ba3\u4f20\u4e2d\u628a\u6c14\u8c61\u9884\u6d4b\u5927\u6a21\u578b\u548c\u6607\u817e\u5361\u4e00\u8d77\u4ecb\u7ecd\uff0c\u591a\u5c11\u6709\u4e9b\u8bef\u5bfc\uff0c\u8fd9\u8fd8\u662f\u7528V100\u8bad\u7ec3\u7684\u3002\u540c\u65f6\u628a\u5404\u79cd\u5927\u6a21\u578b\u5206\u6210L0-L1-L2\u4e5f\u6709\u4e9b\u8fc7\u62df\u5408\uff0c\u611f\u89c9L0\u5e94\u8be5\u662fTransformer\uff0c\u7136\u540e\u624d\u662f\u57fa\u7840\u5927\u6a21\u578b\u548c\u5fae\u8c03\u5927\u6a21\u578b\u3002\u5f53\u7136\u8fd9\u91cc\u9762\u5b58\u5728\u5546\u4e1a\u5ba3\u4f20\u65b9\u9762\u7684\u8bc9\u6c42\uff0c\u6bd5\u7adf\u8ba9\u666e\u901a\u5ba2\u6237\u770b\u61c2\u6700\u4e3a\u91cd\u8981\u3002<\/font><\/p>\n

\u5927\u6a21\u578b\u80af\u5b9a\u8fd8\u4f1a\u70ed\u5f88\u4e45\uff0c\u5bfb\u627e\u81ea\u5df1\u7684\u4f4d\u7f6e\u5e76\u575a\u6301\u521b\u65b0\u662f\u534e\u4e3a\u76d8\u53e4\u5927\u6a21\u578b\u53d6\u5f97\u6210\u529f\u7684\u7ecf\u9a8c\uff01<\/font><\/p>\n


\n<\/font><\/strong><\/p>\n

\u76f8\u5173\u4fe1\u606f\uff1a<\/font><\/strong><\/p>\n

<\/font><\/p>\n

\u76d8\u53e4\u5927\u6a21\u578b\uff08\u534e\u4e3a\u4e91\uff09<\/font><\/a><\/p>\n

Accurate medium-range global weather forecasting with 3D neural networks<\/font><\/strong><\/a><\/p>\n

https:\/\/openi.pcl.ac.cn\/PCL-Platform.Intelligence\/PanGu-Alpha<\/font><\/strong><\/a><\/p>\n

\u4e13\u5bb6\u89e3\u60d1 | \u5173\u4e8e\u534e\u4e3a\u4e91\u76d8\u53e4\u5927\u6a21\u578b\uff0c\u4f60\u60f3\u95ee\u7684\u90fd\u5728\u8fd9\u91cc<\/font><\/a><\/p>\n

6000 \u5b57\u8be6\u89e3\u534e\u4e3a\u76d8\u53e4\u5927\u6a21\u578b\uff1a\u80fd\u5426\u6491\u8d77\u4e16\u754c AI \u53e6\u4e00\u6781\uff1f<\/font><\/strong><\/a><\/p>\n

\u534e\u4e3a\u76d8\u53e4\u5927\u6a21\u578b\u65b0\u8fdb\u5c55\uff0c\u534e\u4e3a\u4e91AI\u9996\u5e2d\u79d1\u5b66\u5bb67000\u5b57\u6f14\u8bb2\u7cbe\u534e<\/font><\/strong><\/a><\/p>\n

\u4f59\u627f\u4e1c\u65b0\u8eab\u4efd\u9996\u79c0\uff1a\u534e\u4e3a\u4e91\u53d1\u5e03\u5168\u7403\u6700\u5927\u76d8\u53e4\u5927\u6a21\u578b\uff0c2.2\u4ebf\u7f8e\u5143\u57f9\u80b2\u6c83\u571f<\/font><\/a><\/p>\n

\u534e\u4e3a\u201c\u76d8\u53e4\u6c14\u8c61\u201d\u767b\u4e0aNature\uff01\u6bd4\u5168\u7403\u6700\u597d\u6570\u503c\u5929\u6c14\u9884\u62a5\u7cfb\u7edf\u66f4\u51c6\u786e\uff0c\u901f\u5ea6\u8fd8\u5feb10000\u500d<\/font><\/a><\/p>\n

2022 \u5e74\u4e2d\u56de\u987e \uff5c \u5927\u6a21\u578b\u6280\u672f\u6700\u65b0\u8fdb\u5c55<\/font><\/a><\/p>\n

\u8de8\u8d8aAI\u5f00\u53d1\u5929\u5811\uff1a\u534e\u4e3a\u4e91\u76d8\u53e4\u5927\u6a21\u578b\u7684\u6545\u4e8b<\/font><\/a><\/p>\n

\u534e\u4e3a\u76d8\u53e4\u5927\u6a21\u578b\u751f\u6001\u94fe\u53ca\u76f8\u5173\u6982\u5ff5\u80a1<\/font><\/a><\/p>\n

\u4e2d\u56fdAI\u7814\u7a76\u65b0\u7a81\u7834\uff1a\u300c\u76d8\u53e4\u03b1\u300d\u53d1\u5e03<\/font><\/a><\/p>\n

\u201c\u76d8\u53e4\u201d\u5f00\u5929\u8bb0\uff0cAI\u843d\u5730\u65f6\uff1a\u4e13\u8bbf\u7530\u9f50<\/font><\/a><\/p>\n

\u534e\u4e3a\u4e91\u76d8\u53e4\u5927\u6a21\u578b\u901a\u8fc7\u4e2d\u56fd\u4fe1\u901a\u9662\u5927\u89c4\u6a21\u9884\u8bad\u7ec3\u6a21\u578b\u9996\u8f6e\u8bc4\u6d4b<\/font><\/a><\/p>\n

\u76f4\u51fbWAIC 2022 \u4e28 \u534e\u4e3a\u4e91\u76d8\u53e4\u5927\u6a21\u578b\u52a0\u901fAI\u5728\u5343\u884c\u767e\u4e1a\u7684\u5b75\u5316\u4e0e\u521b\u65b0<\/font><\/a><\/p>\n

\u56fd\u5185\u5404\u5927\u5382ChatGPT\u6280\u672f\u5e03\u5c40\u53ca\u5e94\u7528\u573a\u666f<\/font><\/a><\/p>\n

\u76d8\u53e4\u5927\u6a21\u578b\u7684\u63a8\u7406\u89e3\u51b3\u65b9\u6848\uff1a\u589e\u91cf\u63a8\u7406+\u5206\u5e03\u5f0f\u63a8\u7406<\/font><\/a><\/p>\n

\u3010\u5927\u5382\u5185\u53c2\u3011\u7b2c16\u671f\uff1a\u534e\u4e3a\u4e91AI\u5f00\u53d1\u751f\u4ea7\u7ebf\uff0c\u7834\u89e3AI\u5168\u6d41\u7a0b\u5f00\u53d1\u96be\u9898<\/font><\/strong><\/a><\/p>\n

\u76d8\u53e4\u5927\u6a21\u578b\u884c\u4e1a\u6df1\u5ea6\uff1a\u53d1\u5c55\u8def\u5f84\u3001\u4ea7\u4e1a\u94fe\u53ca\u76f8\u5173\u516c\u53f8\u6df1\u5ea6\u68b3\u7406<\/font><\/a><\/p>\n

\u89c6\u89c9\u5927\u6a21\u578b\u8c03\u7814\uff08Survey of Visual Foundation Model\uff09<\/font><\/a><\/p>\n

\u5e72\u8d27\u5206\u4eab|HDC 2022 \u5f00\u53d1\u8005\u4e3b\u9898\u6f14\u8bb2\u4e0e\u6280\u672f\u5206\u8bba\u575bPPT\u6c47\u603b\uff08\u5efa\u8bae\u6536\u85cf\uff09<\/font><\/strong><\/a><\/p>\n

\u4fc3\u8fdb\u884c\u4e1a\u6807\u51c6\u5316\u8fdb\u7a0b \u534e\u4e3a\u4e91\u53d1\u5e03\u9884\u8bad\u7ec3\u5927\u6a21\u578b\u767d\u76ae\u4e66<\/font><\/a><\/p>\n

\u534e\u4e3a\u76d8\u53e4\u5927\u6a21\u578b\u7814\u7a76\uff1a\u76d8\u53e4\u5f00\u5929\uff0cAI\u843d\u5730\uff08\u9644\u4e0b\u8f7d\uff09<\/font><\/a><\/p>\n

\u5168\u7403\u9996\u4e2a\uff01\u5c71\u4e1c\u80fd\u6e90\u548c\u534e\u4e3a\u8054\u5408\u53d1\u5e03\u77ff\u5c71\u9886\u57df\u5546\u7528\u4eba\u5de5\u667a\u80fd\u5927\u6a21\u578b<\/font><\/a><\/p>\n

\u77ff\u5c71AI\u5927\u6a21\u578b\uff08\u534e\u4e3a\uff09<\/font><\/a><\/p>\n

\u534e\u4e3a\u4e91\u63d0\u51fa\u76d8\u53e4\u6c14\u8c61\u5927\u6a21\u578b\uff1a\u4e2d\u957f\u671f\u6c14\u8c61\u9884\u62a5\u7cbe\u5ea6\u9996\u6b21\u8d85\u8fc7\u4f20\u7edf\u6570\u503c\u65b9\u6cd5\uff0c\u901f\u5ea6\u63d0\u534710000\u500d\u4ee5\u4e0a<\/font><\/strong><\/a><\/p>\n

HDC.Cloud 2021 | \u534e\u4e3a\u4e91\u53d1\u5e03\u5168\u7403\u6700\u5927\u9884\u8bad\u7ec3\u6a21\u578b\uff0c\u5f00\u542f\u5de5\u4e1a\u5316AI\u5f00\u53d1\u65b0\u6a21\u5f0f<\/font><\/a><\/p>\n

\u767bNature\u6b63\u520a\u540e\uff0c\u534e\u4e3a\u4e91\u76d8\u53e4\u5927\u6a21\u578b3.0\u6765\u4e86<\/font><\/a><\/p>\n

\u534e\u4e3a\u4e91\u76d8\u53e4\u6c14\u8c61\u5927\u6a21\u578b\u7814\u7a76\u6210\u679c\u5728\u300aNature\u300b\u6b63\u520a\u53d1\u8868<\/font><\/a><\/p>\n

CLiB\u4e2d\u6587\u5927\u6a21\u578b\u80fd\u529b\u8bc4\u6d4b\u699c\u5355\uff08\u6301\u7eed\u66f4\u65b0\uff09<\/font><\/a><\/p>\n

\u4e2d\u56fd\u4fe1\u901a\u9662\u53d1\u5e03\u300a\u4eba\u5de5\u667a\u80fd\u7814\u53d1\u8fd0\u8425\u4e00\u4f53\u5316\uff08Model\/MLOps\uff09\u80fd\u529b\u6210\u719f\u5ea6\u6a21\u578b\u300b\u6807\u51c6\u4f53\u7cfb<\/font><\/a><\/p>\n

\u4eba\u5de5\u667a\u80fd\u7814\u53d1\u8fd0\u8425\u4e00\u4f53\u5316\uff08Model\/MLOps\uff09\u80fd\u529b\u6210\u719f\u5ea6\u6a21\u578b\u89e3\u8bfb \u7b2c\u4e00\u90e8\u5206\uff1a\u5f00\u53d1\u7ba1\u7406\uff08\u4e0b\uff09<\/font><\/a><\/p>\n

\u4e2d\u56fd\u4fe1\u901a\u9662 – MLOps\u5de5\u7a0b\u80fd\u529b\u8bc4\u4f30\u6807\u51c6\uff08\u8fde\u8f7d\uff09<\/font><\/strong><\/a><\/p>\n

\u9884\u8bad\u7ec3\u6a21\u578b\u53d1\u5c55\u5386\u53f2<\/font><\/a><\/p>\n

\u9884\u8bad\u7ec3\u8bed\u97f3\u6a21\u578b\u8c03\u7814\u5c0f\u7ed3<\/font><\/a><\/p>\n

Swin Transformer\uff1a\u57fa\u4e8eShifted Windows\u7684\u5c42\u6b21\u5316\u89c6\u89c9\u8bbe\u8ba1<\/font><\/a><\/p>\n

<\/font><\/p>\n

<\/font><\/p>\n

<\/font><\/p>\n

<\/font><\/p>\n

<\/font><\/p>\n

<\/font><\/p>\n

<\/font><\/p>\n

<\/font><\/p>\n

<\/font><\/p>\n

<\/font><\/p>\n

<\/font><\/p>\n

<\/font><\/p>\n

<\/font><\/p>\n","protected":false},"excerpt":{"rendered":"

\u6700\u8fd1\u4e00\u76f4\u5728\u8ddf\u8e2a\u5927\u6a21\u578b\u7684\u4e8b\u60c5\uff0c\u70ed\u5ea6\u5f88\u9ad8\uff0c\u6bcf\u6b21\u5f00\u4f1a\u4e0d\u63d0\u5927\u6a21\u578b\u90fd\u663e\u5f97\u4e0d\u63a5\u5730\u6c14\uff0c\u4f46\u662f\u5927\u6a21\u578b\u770b\u6765\u770b\u53bb\u4f9d\u7136\u5b58\u5728AI\u7684\u901a\u75c5\uff1a…<\/p>\n","protected":false},"author":1,"featured_media":14065,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[39,22,173,1684],"tags":[1691,1686,1690,1687,382,1685,735,1689,1688],"views":4656,"_links":{"self":[{"href":"http:\/\/www.brofive.net\/index.php?rest_route=\/wp\/v2\/posts\/14099"}],"collection":[{"href":"http:\/\/www.brofive.net\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.brofive.net\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.brofive.net\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/www.brofive.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=14099"}],"version-history":[{"count":2,"href":"http:\/\/www.brofive.net\/index.php?rest_route=\/wp\/v2\/posts\/14099\/revisions"}],"predecessor-version":[{"id":14101,"href":"http:\/\/www.brofive.net\/index.php?rest_route=\/wp\/v2\/posts\/14099\/revisions\/14101"}],"wp:featuredmedia":[{"embeddable":true,"href":"http:\/\/www.brofive.net\/index.php?rest_route=\/wp\/v2\/media\/14065"}],"wp:attachment":[{"href":"http:\/\/www.brofive.net\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=14099"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.brofive.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=14099"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.brofive.net\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=14099"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}