{"id":24717,"date":"2021-03-20T22:21:13","date_gmt":"2021-03-20T14:21:13","guid":{"rendered":"https:\/\/swarma.org\/?p=24717"},"modified":"2021-03-20T22:21:13","modified_gmt":"2021-03-20T14:21:13","slug":"%e5%9b%be%e7%bd%91%e7%bb%9c%e7%a9%b6%e7%ab%9f%e5%9c%a8%e7%a0%94%e7%a9%b6%e4%bb%80%e4%b9%88%ef%bc%9f%e4%bb%8e15%e7%af%87%e7%a0%94%e7%a9%b6%e7%bb%bc%e8%bf%b0%e7%9c%8b%e5%9b%be%e7%a5%9e%e7%bb%8f%e7%bd%91","status":"publish","type":"post","link":"https:\/\/swarma.org\/?p=24717","title":{"rendered":"\u56fe\u7f51\u7edc\u7a76\u7adf\u5728\u7814\u7a76\u4ec0\u4e48\uff1f\u4ece15\u7bc7\u7814\u7a76\u7efc\u8ff0\u770b\u56fe\u795e\u7ecf\u7f51\u7edcGNN\u7684\u6700\u65b0\u7814\u7a76\u8fdb\u5c55"},"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

\u8fd1\u5e74\u6765\uff0c\u7531\u4e8e\u56fe\u7ed3\u6784\u7684\u5f3a\u5927\u8868\u73b0\u529b\uff0c\u7528\u673a\u5668\u5b66\u4e60\u65b9\u6cd5\u5206\u6790\u56fe\u7684\u7814\u7a76\u8d8a\u6765\u8d8a\u53d7\u5230\u91cd\u89c6\u3002\u56fe\u795e\u7ecf\u7f51\u7edc<\/strong>\uff08GNN\uff09<\/strong>\u662f\u4e00\u7c7b\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u5904\u7406\u56fe\u57df\u4fe1\u606f\u7684\u65b9\u6cd5\u3002\u56fe\u795e\u7ecf\u7f51\u7edc\u7a76\u7adf\u5728\u7814\u7a76\u4ec0\u4e48\uff1f\u672a\u6765\u8fd8\u6709\u4ec0\u4e48\u6837\u7684\u7814\u7a76\u65b9\u5411\uff1f<\/strong>\u672c\u6587\u6574\u7406\u4e8615\u7bc7\u56fe\u795e\u7ecf\u7f51\u7edc\uff08GNN\uff09\u76f8\u5173\u7684\u6587\u732e\u7efc\u8ff0\uff0c\u8fd915\u7bc7\u76f8\u5173\u7684\u6587\u732e\u7efc\u8ff0\u4ece\u4e0d\u540c\u7684\u89d2\u5ea6\u5bf9\u56fe\u795e\u7ecf\u7f51\u7edc\u7684\u53d1\u5c55\u8fdb\u884c\u4e86\u8be6\u7ec6\u7684\u603b\u7ed3\u548c\u6982\u8ff0\u3002<\/strong><\/span><\/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
<\/section>\n<\/section>\n
<\/p>\n

 Mark<\/span><\/span> | \u4f5c\u8005<\/span><\/strong><\/h3>\n

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

\u6df1\u5ea6\u5b66\u4e60\u4e0e\u56fe\u7f51\u7edc | \u6765\u6e90<\/strong><\/span><\/p>\n<\/section>\n


<\/p>\n


<\/p>\n

\u8fd1\u5e74\u6765\uff0c\u7531\u4e8e\u56fe\u7ed3\u6784\u7684\u5f3a\u5927\u8868\u73b0\u529b\uff0c\u7528\u673a\u5668\u5b66\u4e60\u65b9\u6cd5\u5206\u6790\u56fe\u7684\u7814\u7a76\u8d8a\u6765\u8d8a\u53d7\u5230\u91cd\u89c6\u3002\u56fe\u795e\u7ecf\u7f51\u7edc<\/span>\uff08GNN\uff09<\/span>\u662f\u4e00\u7c7b\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u5904\u7406\u56fe\u57df\u4fe1\u606f\u7684\u65b9\u6cd5\u3002<\/span><\/section>\n

<\/section>\n
\u5230\u76ee\u524d\uff0c\u76f8\u5173\u7814\u7a76\u7684\u5df2\u7ecf\u975e\u5e38\u591a\u4e86\uff0c\u4e0d\u8fc7\u6211\u4eec\u56de\u8fc7\u5934\u6765\u770b\u601d\u8003\u548c\u56de\u987e\u4e00\u4e0b\uff1a\u89e3\u51b3\u56fe<\/span>(Graph)<\/span>\u6570\u636e\u7684\u65b9\u6cd5\u548c\u601d\u8def\u6709\u90a3\u4e9b\uff0c\u8fd9\u4e9b\u65b9\u6cd5\u548c\u601d\u8def\u8ddf\u56fe\u50cf\u6587\u672c\u6709\u4ec0\u4e48\u5f02\u540c\uff1f\u5404\u5927\u9876\u4f1a\u63d0\u51fa\u7684\u56fe\u7f51\u7edc\u6a21\u578b\u6709\u4ec0\u4e48\u6837\u7684\u57fa\u672c\u5047\u8bbe\u548c\u4e3b\u8981\u529f\u80fd\uff1f\u5e38\u7528\u6570\u636e\u96c6Cora, Citeseer, Pubmed\u7b49\u6709\u6ca1\u6709\u7edf\u4e00\u4e00\u81f4\u7684\u5bf9\u6bd4\uff1f\u56fe\u795e\u7ecf\u7f51\u7edc\u548c\u56fe\u5d4c\u5165\u7684\u533a\u522b\u7684\u662f\u4ec0\u4e48\uff1f\u56fe\u7f51\u7edc\u80fd\u505a\u4ec0\u4e48\uff0c\u4e0d\u80fd\u505a\u4ec0\u4e48\uff0c\u56fe\u7f51\u7edc\u8868\u793a\u7684\u4e0a\u754c\u5728\u54ea\u91cc\uff1f\u8282\u70b9\u5206\u7c7b\uff0c\u94fe\u63a5\u9884\u6d4b\uff0c\u56fe\u5206\u7c7b\u7684\u7b97\u6cd5\u548c\u6a21\u578b\u90fd\u6709\u54ea\u4e9b\uff1f\u53e6\u5916\uff0c\u6709\u7684\u56fe\u6570\u636e\u662f\u975e\u5e38\u5c0f\u7684\uff0c\u6709\u7684\u56fe\u6570\u636e<\/span>\uff08\u5982\u5fae\u4fe1\u793e\u4ea4\u7f51\u7edc\uff0c\u5fae\u535a\u793e\u4ea4\u7f51\u7edc\uff09<\/span>\u662f\u975e\u5e38\u5de8\u5927\u7684\uff0c\u90a3\u4e48\u9488\u5bf9\u5c0f\u578b\u56fe\u7f51\u7edc\u4e0e\u5de8\u578b\u7f51\u7edc\u5206\u522b\u90fd\u6709\u4ec0\u4e48\u6837\u7684\u6a21\u578b\uff0c\u7814\u7a76\u7684\u601d\u8def\u8ddf\u4e00\u822c\u7684\u56fe\u6570\u636e\u6709\u4ec0\u4e48\u533a\u522b\uff1f\u6700\u540e\uff0c\u672a\u6765\u8fd8\u6709\u4ec0\u4e48\u6837\u7684\u7814\u7a76\u65b9\u5411\uff1f\u672c\u6587\u6574\u7406\u4e8615\u7bc7\u56fe\u795e\u7ecf\u7f51\u7edc<\/span>(GNN)<\/span>\u76f8\u5173\u7684\u6587\u732e\u7efc\u8ff0\uff0c\u8fd915\u7bc7\u76f8\u5173\u7684\u6587\u732e\u7efc\u8ff0\u4ece\u4e0d\u540c\u7684\u89d2\u5ea6\u5bf9\u56fe\u795e\u7ecf\u7f51\u7edc\u7684\u53d1\u5c55\u8fdb\u884c\u4e86\u8be6\u7ec6\u7684\u603b\u7ed3\u548c\u6982\u8ff0\u3002\u4e0b\u9762\u662f\u672c\u6587\u6574\u7406\u7684\u5217\u8868:<\/span><\/section>\n

<\/span><\/section>\n
1. Deep Learning on Graphs: A Survey<\/span><\/section>\n
\u6e05\u534e\u5927\u5b66 Ziwei Zhang, Peng Cui, Wenwu Zhu<\/span><\/section>\n
2. Graph Neural Networks: A Review of Methods and Applications<\/span><\/section>\n
\u6e05\u534e\u5927\u5b66Jie Zhou, Ganqu Cui, Zhengyan Zhang, Cheng Yang, Zhiyuan Liu, Lifeng Wang, Changcheng Li, Maosong Sun<\/span><\/section>\n
3. A Comprehensive Survey on Graph Neural Networks<\/span><\/section>\n
UTS\u7b49 Zonghan Wu, Shirui Pan, Fengwen Chen, Guodong Long, Chengqi Zhang, Philip S. Yu<\/span><\/section>\n
4. A Gentle Introduction to Deep Learning for Graphs<\/span><\/section>\n
University of Pisa Davide Bacciua , Federico Erricaa , Alessio Michelia , Marco Poddaa<\/span><\/section>\n
5. LEARNING REPRESENTATIONS OF GRAPH DATA: A SURVEY<\/span><\/section>\n
UCL Mital Kinderkhedia<\/span><\/section>\n
6. Graph Neural Networks for Small Graph and Giant Network Representation Learning: An Overview<\/span><\/section>\n
Information Fusion and Mining Laboratory Jiawei Zhang<\/span><\/section>\n
7.  A Survey on The Expressive Power of Graph Neural Networks<\/span><\/section>\n
\u4e1c\u4eac\u5927\u5b66 Ryoma Sato<\/span><\/section>\n
8. Relational inductive biases, deep learning, and graph networks<\/span><\/section>\n
DeepMind;\u7b49<\/span><\/section>\n
9. Adversarial Attack and Defense on Graph Data: A Survey <\/span><\/section>\n
UIUC\u7b49Lichao Sun, Yingtong Dou, Carl Yang\u7b49<\/span><\/section>\n
10. Heterogeneous Network Representation Learning: Survey, Benchmark, Evaluation, and Beyond<\/span><\/section>\n
UIUC Carl Yang, Yuxin Xiao, Yu Zhang, Yizhou Sun, and Jiawei Han,<\/span><\/section>\n
11  Automated Machine Learning on Graphs: A Survey<\/span><\/section>\n
\u6e05\u534e\u5927\u5b66 Ziwei Zhang, Xin Wang, Wenwu Zhu<\/span><\/section>\n
12. Graph Self-Supervised Learning: A Survey<\/span><\/section>\n
Monash University\u7b49 Yixin Liu1 , Shirui Pan1 , Ming Jin1 , Chuan Zhou2 , Feng Xia3 , Philip S. Yu<\/span><\/section>\n
13.  Self-Supervised Learning of Graph Neural Networks: A Unified Review<\/span><\/section>\n
Texas A&M University\u7b49 Yaochen Xie, Zhao Xu, Zhengyang Wang, Shuiwang Ji<\/span><\/section>\n
14. Meta-Learning with Graph Neural Networks: Methods and Applications<\/span><\/section>\n
Columbia University\u7b49<\/span><\/section>\n
Debmalya Mandal, Sourav Medya, Brian Uzzi, Charu Aggarwal<\/span><\/section>\n
15. Deep Graph Structure Learning for Robust Representations: A Survey<\/span><\/section>\n
\u4e2d\u79d1\u9662\u81ea\u52a8\u5316\u7814\u7a76\u6240\uff1aYanqiao Zhu\u2217 , Weizhi Xu\u2217 , Jinghao Zhang\u2217 , Qiang Liu , Shu Wu\u2020 and Liang Wang<\/span><\/section>\n

<\/span><\/section>\n
\u6574\u7406\u65f6\u95f4: 2021-3-18\uff0c\u4ee5\u4e0apdf\u90fd\u53ef\u4ee5\u516c\u5f00\u83b7\u5f97<\/span><\/section>\n

<\/section>\n
\u540e\u7eed\u5c06\u5728\u8fd9\u91cc\u66f4\u65b0\uff1a<\/span><\/section>\n
https:\/\/zhuanlan.zhihu.com\/p\/120152289<\/span><\/section>\n
\n
\n
\n

\n
\n


<\/p>\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

1. Deep Learning on Graphs: A Survey<\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

<\/section>\n<\/section>\n<\/h3>\n<\/section>\n<\/section>\n<\/section>\n
\u65f6\u95f4\uff1a2018\u5e7412\u6708\u9996\u53d1\uff0c2020\u5e743\u6708\u66f4\u65b0\u81f3\u7b2c\u4e09\u7248<\/strong><\/span><\/section>\n
\u4f5c\u8005\uff1aZiwei Zhang\u7b49<\/span><\/section>\n
\u5355\u4f4d\uff1a\u6e05\u534e\u5927\u5b66<\/span><\/section>\n
\u7bc7\u5e45\uff1a24\u9875<\/span><\/section>\n
\u94fe\u63a5\uff1a<\/span><\/section>\n
https:\/\/arxiv.org\/pdf\/1812.04202.pdf<\/span><\/section>\n

<\/section>\n
\u4ece\u58f0\u5b66\u3001\u56fe\u50cf\u5230\u81ea\u7136\u8bed\u8a00\u5904\u7406\uff0c\u6df1\u5ea6\u5b66\u4e60\u5728\u8bb8\u591a\u9886\u57df\u90fd\u53d6\u5f97\u4e86\u6210\u529f\u3002\u7136\u800c\uff0c\u5c06\u6df1\u5ea6\u5b66\u4e60\u5e94\u7528\u4e8e\u65e0\u6240\u4e0d\u5728\u7684\u56fe\u6570\u636e\u5e76\u975e\u6613\u4e8b\uff0c\u56e0\u4e3a\u56fe\u5f62\u662f\u4e00\u79cd\u975e\u6b27\u7684\u6570\u636e\u7ed3\u6784\uff0c\u5177\u6709\u72ec\u7279\u7684\u7279\u6027\u3002\u8fd1\u5e74\u6765\uff0c\u8fd9\u4e00\u9886\u57df\u7684\u7814\u7a76\u53d6\u5f97\u4e86\u5f88\u5927\u7684\u8fdb\u5c55\uff0c\u6781\u5927\u5730\u63a8\u52a8\u4e86\u56fe\u5206\u6790\u6280\u672f\u7684\u53d1\u5c55\u3002\u5728\u672c\u7814\u7a76\u4e2d\uff0c\u4f5c\u8005\u5168\u9762\u56de\u987e\u4e86\u5e94\u7528\u4e8e\u56fe\u6df1\u5ea6\u5b66\u4e60\u5404\u79cd\u65b9\u6cd5\u3002\u603b\u7ed3\u4e86\u73b0\u6709\u7684\u56fe\u7f51\u7edc\u7c7b\u578b\uff0c\u5c06\u73b0\u6709\u7684\u65b9\u6cd5\u5206\u4e3a\u4e94\u7c7b:  \u56fe\u9012\u5f52\u795e\u7ecf\u7f51\u7edc\u3001\u56fe\u5377\u79ef\u7f51\u7edc\u3001\u56fe\u81ea\u7f16\u7801\u5668\u3001\u56fe\u589e\u5f3a\u5b66\u4e60\u548c\u56fe\u5bf9\u6297\u65b9\u6cd5\u3002<\/strong><\/span><\/section>\n

<\/section>\n
\u76f8\u5173\u62a5\u9053\u4e0e\u89e3\u8bfb\uff1a<\/span><\/strong><\/section>\n
\u56fe\u7f51\u7edc|\u6f2b\u8c08\u56fe\u795e\u7ecf\u7f51\u7edc\u6a21\u578b<\/span><\/a>\uff1b <\/span><\/section>\n
\u56fe\u7f51\u7edc\u7684\u4e94\u7c7b\u6a21\u578b<\/span>(http:\/\/suo.im\/5G6MdF)<\/span><\/section>\n

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

<\/section>\n<\/section>\n
\u4ee5\u4e0a\u4e94\u7c7b\u6a21\u578b\u7684\u57fa\u672c\u5047\u8bbe\u548c\u4e3b\u8981\u529f\u80fd<\/span><\/section>\n
<\/section>\n

<\/section>\n
\u8fd1\u5e74\u6765\u4e0d\u540c\u795e\u7ecf\u7f51\u7edc\u7684\u5bf9\u6bd4\u4e0e\u603b\u7ed3<\/span><\/section>\n
\n
<\/section>\n<\/section>\n

<\/section>\n
\u56fe\u81ea\u7f16\u7801\u5668\u603b\u7ed3\u4e0e\u5bf9\u6bd4<\/span><\/section>\n
\n
<\/section>\n

<\/section>\n<\/section>\n
\u56fe\u5f3a\u5316\u5b66\u4e60\u7279\u6027\u5bf9\u6bd4<\/span><\/section>\n
<\/section>\n

<\/section>\n
\u56fe\u5bf9\u6297\u5b66\u4e60\u7279\u6027\u5bf9\u6bd4<\/span><\/section>\n
\n

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


<\/section>\n
\u76f8\u5173\u7684\u5b66\u4e60\u6846\u67b6\u603b\u7ed3\uff1aPyG; DGL; AliGraph; Euler<\/span><\/section>\n
<\/section>\n

<\/section>\n
Cora, Citeseer, Pubmed, Reddit, PPI\u6570\u636e\u96c6\u6a21\u578b\u6027\u80fd\u5168\u9762\u603b\u7ed3<\/span><\/section>\n
\n

<\/p>\n


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

\n
\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

2. Graph Neural Networks: A Review of Methods and Applications<\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

<\/section>\n<\/section>\n<\/h3>\n<\/section>\n<\/section>\n<\/section>\n
\u65f6\u95f4\uff1a2018\u5e7412\u6708\uff1b2019\u5e746\u6708\u7b2c\u56db\u7248<\/strong><\/span><\/section>\n
\u4f5c\u8005\uff1aJie Zhou\u7b49<\/span><\/section>\n
\u5355\u4f4d\uff1a\u6e05\u534e\u5927\u5b66<\/span><\/section>\n
\u7bc7\u5e45\uff1a22\u9875\uff0c153\u7bc7\u53c2\u8003\u6587\u732e<\/span><\/section>\n
\u94fe\u63a5\uff1a<\/span><\/section>\n
https:\/\/arxiv.org\/abs\/1812.08434<\/span><\/section>\n

<\/section>\n
\u8be5\u6587\u603b\u7ed3\u4e86\u56fe\u795e\u7ecf\u7f51\u7edc\u9886\u57df\u7684\u7ecf\u5178\u6a21\u578b\u4e0e\u5178\u578b\u5e94\u7528<\/span>\uff08\u6587\u672c\uff0c\u56fe\u50cf\uff0c\u7269\u7406\u5316\u5b66\u836f\u5b66\uff0c\u77e5\u8bc6\u56fe\u8c31\uff09<\/span><\/strong>\uff0c\u5e76\u63d0\u51fa\u4e86\u56db\u4e2a\u5f00\u653e\u6027\u95ee\u9898\u3002\u5bf9\u4e8e\u5e0c\u671b\u5feb\u901f\u4e86\u89e3\u8fd9\u4e00\u9886\u57df\u7684\u8bfb\u8005\uff0c\u4e0d\u59a8\u5148\u4ece\u8fd9\u7bc7\u6587\u7ae0\u770b\u8d77\u3002<\/span><\/section>\n

<\/section>\n
\u5f88\u591a\u5b66\u4e60\u4efb\u52a1\u90fd\u9700\u8981\u5904\u7406\u56fe\u6570\u636e\uff0c\u8fd9\u4e9b\u6570\u636e\u5305\u542b\u4e86\u5143\u7d20\u4e4b\u95f4\u4e30\u5bcc\u7684\u5173\u7cfb\u4fe1\u606f\u3002\u5efa\u6a21\u7269\u7406\u7cfb\u7edf\uff0c\u5b66\u4e60\u5206\u5b50\u6307\u7eb9\uff0c\u9884\u6d4b\u86cb\u767d\u8d28\u754c\u9762\uff0c\u4ee5\u53ca\u75be\u75c5\u5206\u7c7b\u90fd\u9700\u8981\u6a21\u578b\u4ece\u56fe\u5f62\u8f93\u5165\u4e2d\u5b66\u4e60\u3002\u5728\u4ece\u6587\u672c\u3001\u56fe\u50cf\u7b49\u975e\u7ed3\u6784\u5316\u6570\u636e\u5b66\u4e60\u7b49\u9886\u57df\uff0c\u5bf9\u63d0\u53d6\u51fa\u7684\u53e5\u5b50\u4f9d\u8d56\u6811\u3001\u56fe\u50cf\u573a\u666f\u56fe\u7b49\u7ed3\u6784\u8fdb\u884c\u63a8\u7406\u662f\u4e00\u4e2a\u91cd\u8981\u7684\u7814\u7a76\u8bfe\u9898\uff0c\u4e5f\u9700\u8981\u56fe\u5f62\u63a8\u7406\u6a21\u578b\u3002\u56fe\u795e\u7ecf\u7f51\u7edc<\/span>(GNNs)<\/span>\u662f\u4e00\u79cd\u8fde\u63a5\u4e3b\u4e49\u6a21\u578b\uff0c\u5b83\u901a\u8fc7\u5728\u56fe\u7684\u8282\u70b9\u4e4b\u95f4\u4f20\u9012\u6d88\u606f\u6765\u83b7\u53d6\u56fe\u7684\u4f9d\u8d56\u6027\u3002\u4e0e\u6807\u51c6\u795e\u7ecf\u7f51\u7edc\u4e0d\u540c\u7684\u662f\uff0c\u56fe\u795e\u7ecf\u7f51\u7edc\u4fdd\u7559\u4e86\u4e00\u79cd\u72b6\u6001\uff0c\u8fd9\u79cd\u72b6\u6001\u53ef\u4ee5\u7528\u4efb\u610f\u6df1\u5ea6\u8868\u793a\u90bb\u5c45\u7684\u4fe1\u606f\u3002\u867d\u7136\u539f\u59cb\u56fe\u795e\u7ecf\u7f51\u7edc\u5f88\u96be\u8bad\u7ec3\u6210\u5b9a\u70b9\uff0c\u4f46\u662f\u6700\u8fd1\u5728\u7f51\u7edc\u7ed3\u6784\u3001\u4f18\u5316\u6280\u672f\u548c\u5e76\u884c\u8ba1\u7b97\u65b9\u9762\u7684\u8fdb\u5c55\u4f7f\u5f97\u5229\u7528\u5b83\u4eec\u8fdb\u884c\u6210\u529f\u7684\u5b66\u4e60\u6210\u4e3a\u53ef\u80fd\u3002<\/span><\/section>\n

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

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


<\/section>\n
\n

<\/p>\n


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

<\/p>\n


<\/section>\n<\/section>\n
\u5404\u7c7b\u7684\u56fe\u795e\u7ecf\u7f51\u7edc\u7684Aggregator\u548cUpdater<\/span><\/section>\n
\n

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


<\/section>\n
\u56fe\u7f51\u7edc\u76f8\u5173\u7684\u5e94\u7528<\/span><\/section>\n
\n

<\/p>\n


<\/section>\n<\/section>\n
\u6587\u7ae0\u6700\u540e\u63d0\u51fa\u4e86\u56fe\u795e\u7ecf\u7f51\u7edc\u9886\u57df\u7684\u56db\u4e2a\u5f00\u653e\u95ee\u9898\uff1a<\/span><\/section>\n

<\/section>\n
1. \u6d45\u5c42\u7ed3\u6784\u3002\u7ecf\u9a8c\u4e0a\u4f7f\u7528\u66f4\u591a\u53c2\u6570\u7684\u795e\u7ecf\u7f51\u7edc\u80fd\u591f\u5f97\u5230\u66f4\u597d\u7684\u5b9e\u9a8c\u6548\u679c\uff0c\u7136\u800c\u5806\u53e0\u591a\u5c42\u7684 GNN \u5374\u4f1a\u4ea7\u751f over-smoothing \u7684\u95ee\u9898\u3002\u5177\u4f53\u6765\u8bf4\uff0c\u5806\u53e0\u5c42\u6570\u8d8a\u591a\uff0c\u8282\u70b9\u8003\u8651\u7684\u90bb\u5c45\u4e2a\u6570\u4e5f\u4f1a\u8d8a\u591a\uff0c\u5bfc\u81f4\u6700\u7ec8\u6240\u6709\u8282\u70b9\u7684\u8868\u793a\u4f1a\u8d8b\u5411\u4e8e\u4e00\u81f4\u3002<\/span><\/section>\n

<\/section>\n
2. \u52a8\u6001\u56fe\u3002\u76ee\u524d\u5927\u90e8\u5206\u65b9\u6cd5\u5173\u6ce8\u4e8e\u5728\u9759\u6001\u56fe\u4e0a\u7684\u5904\u7406\uff0c\u5bf9\u4e8e\u5982\u4f55\u5904\u7406\u8282\u70b9\u4fe1\u606f\u548c\u8fb9\u4fe1\u606f\u968f\u7740\u65f6\u95f4\u6b65\u52a8\u6001\u53d8\u5316\u7684\u56fe\u4ecd\u662f\u4e00\u4e2a\u5f00\u653e\u95ee\u9898\u3002<\/span><\/section>\n

<\/section>\n
3. \u975e\u7ed3\u6784\u5316\u573a\u666f\u3002\u867d\u7136\u5f88\u591a\u5de5\u4f5c\u5e94\u7528\u4e8e\u975e\u7ed3\u6784\u5316\u7684\u573a\u666f<\/span>\uff08\u6bd4\u5982\u6587\u672c\uff09<\/span>\uff0c\u7136\u800c\u5e76\u6ca1\u6709\u901a\u7528\u7684\u65b9\u6cd5\u7528\u4e8e\u5904\u7406\u975e\u7ed3\u6784\u5316\u7684\u6570\u636e\u3002<\/span><\/section>\n

<\/section>\n
4. \u6269\u5c55\u6027\u3002\u867d\u7136\u5df2\u7ecf\u6709\u4e00\u4e9b\u65b9\u6cd5\u5c1d\u8bd5\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\uff0c\u5c06\u56fe\u795e\u7ecf\u7f51\u7edc\u7684\u65b9\u6cd5\u5e94\u7528\u4e8e\u5927\u89c4\u6a21\u6570\u636e\u4e0a\u4ecd\u7136\u662f\u4e00\u4e2a\u5f00\u653e\u6027\u95ee\u9898\u3002<\/span><\/section>\n

<\/section>\n
\u76f8\u5173\u62a5\u9053\u4e0e\u89e3\u8bfb\uff1a<\/span><\/strong><\/section>\n
\u56fe\u795e\u7ecf\u7f51\u7edc\u7efc\u8ff0\uff1a\u6a21\u578b\u4e0e\u5e94\u7528<\/span>(http:\/\/suo.im\/5V9dkJ, \u673a\u5668\u4e4b\u5fc3)<\/span><\/section>\n

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

\n
\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

3. A Comprehensive Survey on Graph Neural Networks<\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

<\/section>\n<\/section>\n<\/h3>\n<\/section>\n<\/section>\n<\/section>\n
\u65f6\u95f4\uff1a2019\u5e741\u6708\u9996\u53d1\uff0c2019\u5e7412\u6708\u7b2c\u56db\u7248<\/strong><\/span><\/section>\n
\u4f5c\u8005\uff1aZonghan Wu\u7b49<\/span><\/section>\n
\u5355\u4f4d\uff1a\u6089\u5c3c\u79d1\u6280\u5927\u5b66\u7b49<\/span><\/section>\n

<\/section>\n
\u672c\u6587\u504f\u5411\u56fe\u6a21\u578b\u7684\u5168\u9762\u603b\u7ed3\uff0c\u4e0e\u7b2c\u4e00\u7bc7\u53ef\u4ee5\u4e92\u8865\u6765\u770b\u3002<\/span><\/strong>\u672c\u6587\u5bf9\u56fe\u7f51\u7edc\u63d0\u51fa\u4e86\u65b0\u7684\u5206\u7c7b\u4f53\u7cfb\uff1a\u56fe\u5377\u79ef\u7f51\u7edc\u3001\u56fe\u6ce8\u610f\u529b\u7f51\u7edc\u3001\u56fe\u81ea\u7f16\u7801\u5668\u3001\u56fe\u751f\u6210\u7f51\u7edc\u548c\u56fe\u65f6\u7a7a\u7f51\u7edc\u3002\u786e\u5b9a\u4e86\u56fe\u795e\u7ecf\u7f51\u7edc\u548c\u7f51\u7edc\u5d4c\u5165\u4e4b\u95f4\u7684\u533a\u522b\uff0c\u5e76\u5728\u4e0d\u540c\u7684\u56fe\u795e\u7ecf\u7f51\u7edc\u67b6\u6784\u4e4b\u95f4\u5efa\u7acb\u4e86\u8054\u7cfb\u3002<\/strong><\/span><\/section>\n

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

<\/p>\n


<\/section>\n<\/section>\n
\u201c\u56fe\u5d4c\u5165\u65e8\u5728\u901a\u8fc7\u4fdd\u7559\u56fe\u7684\u7f51\u7edc\u62d3\u6251\u7ed3\u6784\u548c\u8282\u70b9\u5185\u5bb9\u4fe1\u606f\uff0c\u5c06\u56fe\u4e2d\u9876\u70b9\u8868\u793a\u4e3a\u4f4e\u7ef4\u5411\u91cf\u7a7a\u95f4\uff0c\u4ee5\u4fbf\u4f7f\u7528\u7b80\u5355\u7684\u673a\u5668\u5b66\u4e60\u7b97\u6cd5<\/span>\uff08\u4f8b\u5982\uff0c\u652f\u6301\u5411\u91cf\u673a\u5206\u7c7b\uff09<\/span>\u8fdb\u884c\u5904\u7406\u3002\u8bb8\u591a\u56fe\u5d4c\u5165\u7b97\u6cd5\u901a\u5e38\u662f\u65e0\u76d1\u7763\u7684\u7b97\u6cd5\uff0c\u5b83\u4eec\u53ef\u4ee5\u5927\u81f4\u53ef\u4ee5\u5212\u5206\u4e3a\u4e09\u4e2a\u7c7b\u522b\uff0c\u5373\u77e9\u9635\u5206\u89e3\u3001\u968f\u673a\u6e38\u8d70\u548c\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\u3002\u540c\u65f6\u56fe\u5d4c\u5165\u7684\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\u4e5f\u5c5e\u4e8e\u56fe\u795e\u7ecf\u7f51\u7edc\uff0c\u5305\u62ec\u57fa\u4e8e\u56fe\u81ea\u52a8\u7f16\u7801\u5668\u7684\u7b97\u6cd5<\/span>\uff08\u5982DNGR\u548cSDNE\uff09<\/span>\u548c\u65e0\u76d1\u7763\u8bad\u7ec3\u7684\u56fe\u5377\u79ef\u795e\u7ecf\u7f51\u7edc<\/span>\uff08\u5982GraphSage\uff09<\/span>\u3002\u4e0b\u56fe\u63cf\u8ff0\u4e86\u56fe\u5d4c\u5165\u548c\u56fe\u795e\u7ecf\u7f51\u7edc\u5728\u672c\u6587\u4e2d\u7684\u533a\u522b\u201d<\/span><\/section>\n

<\/section>\n
\u8be5\u7efc\u8ff0\u63d0\u4f9b\u4e86\u5728\u56fe\u6570\u636e\u4e0a\u7684\u73b0\u4ee3\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u7684\u5168\u9762\u6982\u8ff0\u3002\u5bf9\u6bcf\u4e00\u79cd\u7c7b\u578b\u7684\u56fe\u795e\u7ecf\u7f51\u7edc\uff0c\u63d0\u4f9b\u4e86\u8868\u5f81\u7b97\u6cd5\u7684\u7ec6\u8282\u63cf\u8ff0\uff0c\u5e76\u505a\u4e86\u5fc5\u8981\u7684\u5bf9\u6bd4\u548c\u5bf9\u5e94\u7b97\u6cd5\u7684\u603b\u7ed3\u3002\u8fd9\u7bc7\u7efc\u8ff0\u63d0\u4f9b\u4e86\u56fe\u795e\u7ecf\u7f51\u7edc\u7684\u4e30\u5bcc\u8d44\u6e90\uff0c\u5176\u4e2d\u5305\u62ec\u5f53\u524d\u6700\u4f73\u7b97\u6cd5\u3001\u57fa\u51c6\u6570\u636e\u96c6\u3001\u5f00\u6e90\u4ee3\u7801\u548c\u5b9e\u8df5\u5e94\u7528\u3002\u8fd9\u7bc7\u7efc\u8ff0\u53ef\u4ee5\u4f5c\u4e3a\u7406\u89e3\u3001\u4f7f\u7528\u548c\u5f00\u53d1\u4e0d\u540c\u5b9e\u9645\u5e94\u7528\u7684\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\u7684\u5b9e\u8df5\u6307\u5357\u3002\u8fd9\u7bc7\u7efc\u8ff0\u8fd8\u5f3a\u8c03\u4e86\u5df2\u6709\u7b97\u6cd5\u7684\u5f53\u524d\u9650\u5236\uff0c\u6307\u51fa\u4e86\u8fd9\u4e2a\u5feb\u901f\u53d1\u5c55\u9886\u57df\u672a\u6765\u7684\u53ef\u80fd\u65b9\u5411\u3002<\/span><\/section>\n

<\/section>\n
RecGNNs and ConvGNNs\u76f8\u5173\u603b\u7ed3<\/span><\/section>\n
\n
<\/section>\n

<\/section>\n<\/section>\n
GAE\u76f8\u5173\u603b\u7ed3<\/span><\/section>\n
\n
<\/section>\n

<\/section>\n<\/section>\n
\u5e38\u7528Benchmark\u6570\u636e\u96c6\u603b\u7ed3<\/span><\/section>\n
\n
<\/section>\n<\/section>\n

<\/section>\n
\u76f8\u5173\u62a5\u9053\u4e0e\u89e3\u8bfb\uff1a<\/span><\/strong><\/section>\n
https:\/\/zhuanlan.zhihu.com\/p\/75307407\uff1b<\/span>
<\/section>\n
\u56fe\u7f51\u7edc|\u6f2b\u8c08\u56fe\u795e\u7ecf\u7f51\u7edc\u6a21\u578b<\/span><\/a>; <\/span><\/section>\n
\n
\n
\n

\n
\n


<\/p>\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

4. A Gentle Introduction to Deep Learning for Graphs<\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

<\/section>\n<\/section>\n<\/h3>\n<\/section>\n<\/section>\n<\/section>\n
\u65f6\u95f4\uff1a2019\u5e7412\u6708<\/span><\/section>\n
\u4f5c\u8005\uff1aDavide Bacciu\u7b49<\/span><\/section>\n
\u5355\u4f4d\uff1aDepartment of Computer Science, University of Pisa<\/span><\/section>\n
\u7bc7\u5e45\uff1a55\u9875\uff0c\u5171\u8ba1131\u7bc7\u53c2\u8003\u6587\u732e<\/span><\/section>\n
\u94fe\u63a5\uff1a<\/span><\/section>\n
https:\/\/arxiv.org\/pdf\/1912.12693.pdf<\/span><\/section>\n

<\/section>\n
\u56fe\u3001\u56fe\u7f51\u7edc\u7684\u7814\u7a76\u4e00\u4e2a\u7ecf\u5178\u7684\u5e76\u4e14\u957f\u671f\u5b58\u5728\u7684\u4e3b\u9898\uff0c\u6700\u8fd1\u5df2\u4e0e\u4e3a\u6df1\u5ea6\u5b66\u4e60\u7ed3\u5408\u5f15\u8d77\u4e86\u76f8\u5173\u7814\u7a76\u4eba\u5458\u7684\u5e7f\u6cdb\u5174\u8da3\u3002\u76f8\u5173\u7814\u7a76\u7684\u6570\u91cf\u548c\u5e7f\u5ea6\u8fc5\u901f\u589e\u52a0\uff0c\u4f46\u662f\u7f3a\u4e4f\u7cfb\u7edf\u5316\u7684\u77e5\u8bc6\u68b3\u7406\u4ee5\u53ca\u7f3a\u5c11\u5bf9\u65e9\u671f\u7814\u7a76\u7684\u5173\u6ce8\u3002\u8fd9\u9879\u6587\u732e\u7efc\u8ff0\u65e8\u5728\u4f5c\u4e3a\u56fe\u6df1\u5ea6\u5b66\u4e60\u9886\u57df\u7684\u6559\u7a0b\u3002
<\/span><\/section>\n

<\/section>\n
\u4e0e\u6700\u65b0\u6587\u732e\u4e0d\u540c\u7684\u662f\uff0c\u672c\u6587\u4e3b\u8981\u4ecb\u7ecd\u4e86\u76f8\u5173\u7684\u6982\u5ff5\u548c\u7ed3\u6784\uff0c\u65b9\u4fbf\u8bfb\u8005\u5b66\u4e60\u548c\u67e5\u770b\u3002\u8bba\u6587\u91c7\u7528\u4e86\u81ea\u4e0a\u800c\u4e0b\u601d\u8def\uff0c\u4ecb\u7ecd\u4e86DNGNs<\/span>(\u6df1\u5ea6\u795e\u7ecf\u56fe\u7f51\u7edc)<\/span>\u3001DBGNs<\/span>(\u6df1\u5ea6\u8d1d\u53f6\u65af\u56fe\u7f51\u7edc)<\/span>\u548cDGGNs<\/span>(\u6df1\u5ea6\u751f\u6210\u56fe\u7f51\u7edc)<\/span>\uff0c\u4e0b\u9762\u662f\u8bba\u6587\u7684Roadmap\u3002<\/span><\/section>\n

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

<\/p>\n


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

\n

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

\n
\n
\n

\n
\n


<\/p>\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

5. LEARNING REPRESENTATIONS OF GRAPH DATA: A SURVEY<\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

<\/section>\n<\/section>\n<\/h3>\n<\/section>\n<\/section>\n<\/section>\n
\u65f6\u95f4\uff1a2019\u5e746\u6708<\/span><\/section>\n
\u4f5c\u8005\uff1aMital Kinderkhedia<\/span><\/section>\n
\u5355\u4f4d\uff1aDepartment of Statistical Science, University College London<\/span><\/section>\n
\u7bc7\u5e45\uff1a22\u9875\uff0c139\u7bc7\u53c2\u8003\u6587\u732e<\/span><\/section>\n
\u94fe\u63a5\uff1a<\/span><\/section>\n
https:\/\/arxiv.org\/pdf\/1906.02989.pdf<\/span><\/section>\n

<\/section>\n
\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u5728\u76ee\u6807\u8bc6\u522b\uff0c\u56fe\u50cf\u5206\u7c7b\u548c\u81ea\u7136\u8bed\u8a00\u5904\u7406\u9886\u57df\u663e\u793a\u51fa\u4e86\u5de8\u5927\u7684\u6210\u529f\u3002\u4f46\u662f\uff0c\u5982\u4f55\u7528\u6df1\u5ea6\u5b66\u4e60\u8bbe\u8ba1\u53ef\u5b66\u4e60\u5e76\u4e14\u80fd\u591f\u6709\u6548\u8f93\u51fa\u7684\u56fe\u7f51\u7edc\u7ed3\u6784\u4ecd\u7136\u662f\u4e00\u4e2a\u9700\u8981\u7814\u7a76\u95ee\u9898\u3002\u672c\u6587\u7efc\u8ff0\u7684\u76ee\u7684\u662f\u603b\u7ed3\u548c\u8ba8\u8bba\u56fe\u6570\u636e\u8868\u793a\u5b66\u4e60\u65b9\u6cd5\u7684\u6700\u65b0\u7814\u7a76\u8fdb\u5c55\u3002<\/span><\/section>\n

<\/section>\n
\u672c\u6587\u9996\u5148\u786e\u5b9a\u56fe\u6570\u636e\u7684\u5e38\u7528\u7c7b\u578b\uff0c\u7136\u540e\u56de\u987e\u56fe\u8bba\u57fa\u7840\u3002\u63a5\u4e0b\u6765\u8ba8\u8bba\u4e86\u56fe\u6838\u65b9\u6cd5\u4e0e\u795e\u7ecf\u7f51\u7edc\u4e4b\u95f4\u5173\u7cfb\u3002\u7136\u540e\uff0c\u786e\u5b9a\u7528\u4e8e\u5b66\u4e60\u56fe\u6570\u636e\u8868\u793a\u7684\u4e3b\u8981\u65b9\u6cd5\uff0c\u5373\uff1a\u5185\u6838\u65b9\u6cd5\uff0c\u5377\u79ef\u65b9\u6cd5\uff0c\u56fe\u795e\u7ecf\u7f51\u7edc\u65b9\u6cd5\uff0c\u56fe\u5d4c\u5165\u65b9\u6cd5\u548c\u6982\u7387\u65b9\u6cd5\u3002<\/strong>\u8be5\u90e8\u5206\u8ba8\u8bba\u4e86\u6bcf\u79cd\u65b9\u6cd5\u4e0b\u7684\u5404\u79cd\u65b9\u6cd5\uff0c\u5e76\u901a\u8fc7\u7b80\u8981\u8ba8\u8bba\u56fe\u8868\u793a\u5b66\u4e60\u672a\u6765\u7814\u7a76\u7684\u5185\u5bb9\u3002<\/span><\/section>\n

<\/section>\n
\u4e0b\u9762\u662f\u672c\u6587\u8ba8\u8bba\u7684\u76f8\u5173\u56fe\u6570\u636e<\/span><\/strong>
<\/span><\/section>\n

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

<\/p>\n


<\/section>\n<\/section>\n
\u56fe\u8bba\u4e2d\u90e8\u5206\u57fa\u672c\u7684\u6982\u5ff5<\/span><\/strong>
<\/span><\/section>\n

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

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

\n
\n
\n

\n
\n


<\/p>\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

6. Graph Neural Networks for Small Graph and Giant Network Representation Learning: An Overview<\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

<\/section>\n<\/section>\n<\/h3>\n<\/section>\n<\/section>\n<\/section>\n
\u65f6\u95f4\uff1a2019\u5e748\u6708\u9996\u53d1<\/span><\/section>\n
\u4f5c\u8005\uff1aJiawei Zhang<\/span><\/section>\n
\u5355\u4f4d\uff1aInformation Fusion and Mining Laboratory<\/span><\/section>\n
\u7bc7\u5e45\uff1a30\u9875\uff0c\u53c2\u8003\u6587\u732e9<\/span><\/section>\n
\u94fe\u63a5\uff1a<\/span><\/section>\n
https:\/\/arxiv.org\/abs\/1908.00187<\/span><\/section>\n

<\/section>\n
\u7814\u7a76\u7684\u56fe\u6570\u636e\u901a\u5e38\u53ef\u5206\u4e3a\u4e24\u79cd\u4e3b\u8981\u7c7b\u578b\uff0c\u5373\u5c0f\u578b\u56fe\u7f51\u7edc\u4e0e\u5de8\u578b\u7f51\u7edc\uff0c\u5b83\u4eec\u5728\u5927\u5c0f\uff0c\u7f16\u53f7\u548c\u6807\u7b7e\u6ce8\u91ca\u65b9\u9762\u5dee\u5f02\u5f88\u5927\u3002\u5728\u672c\u6587\u4e2d\uff0c\u4f5c\u8005\u5bf9\u4e8e\u8fd9\u4e24\u79cd\u4e0d\u540c\u7c7b\u578b\u7684\u56fe\u6570\u636e\uff0c\u4ecb\u7ecd\u8fd1\u5e74\u6765\u5f15\u5165\u7684\u56fe\u795e\u7ecf\u7f51\u7edc\u3002<\/span><\/strong>\u66f4\u5177\u4f53\u5730\u8bf4\uff0c\u672c\u6587\u4ecb\u7ecd\u7684\u56fe\u795e\u7ecf\u7f51\u7edc\u5305\u62ecIsoNN [4]\uff0cSDBN [7]\uff0cLF\uff06ER [6]\uff0cGCN [3]\uff0cGAT [5]\uff0cDifNN [8]\uff0cGNL [1]\uff0cGraphSage [2]\u548cseGEN [9]\u3002\u5728\u8fd9\u4e9b\u56fe\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\u4e2d\uff0c\u6700\u521d\u4e3a\u5c0f\u56fe\u63d0\u51fa\u4e86IsoNN\uff0cSDBN\u548cLF\uff06ER\uff0c\u800c\u5176\u4f59\u7684\u5219\u6700\u521d\u4e3a\u5de8\u4eba\u7f51\u7edc\u63d0\u51fa\u4e86\u3002<\/span><\/section>\n

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

<\/p>\n


<\/section>\n<\/section>\n
\u5c0f\u89c4\u6a21\u56fe\u7f51\u7edc\u7684\u65b9\u6cd5\u4e00\u89c8<\/span><\/strong>
<\/span><\/section>\n

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

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

<\/section>\n

<\/section>\n
\u5927\u89c4\u6a21\u56fe\u7f51\u7edc\u5b66\u4e60\u65b9\u6cd5\u4e00\u89c8<\/span><\/strong><\/section>\n

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

<\/p>\n


<\/section>\n<\/section>\n
\u5927\u89c4\u6a21\u56fe\u7f51\u7edc\u4e0b\u91c7\u6837\u65b9\u5f0f\u7684\u56fe\u5d4c\u5165<\/span><\/strong>
<\/span><\/section>\n

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

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

\n
\n
\n

\n
\n


<\/p>\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

7.  A Survey on The Expressive Power of Graph Neural Networks<\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

<\/section>\n<\/section>\n<\/h3>\n<\/section>\n<\/section>\n<\/section>\n
\u65f6\u95f4\uff1a2020\u5e743\u6708<\/span><\/section>\n
\u4f5c\u8005\uff1aRyoma Sato<\/span><\/section>\n
\u4f5c\u8005\u5355\u4f4d\uff1aKyoto University \/ RIKEN AIP<\/span><\/section>\n
\u7bc7\u5e45\uff1a41\u9875<\/span><\/section>\n
\u94fe\u63a5\uff1a<\/span><\/section>\n
https:\/\/arxiv.org\/abs\/2003.04078<\/span><\/section>\n

<\/section>\n
\u56fe\u795e\u7ecf\u7f51\u7edc\u5230\u5e95\u6709\u591a\u4e48\u7684\u5f3a\u5927\u5462\uff1f\u6765\u81ea\u65e5\u672c\u4eac\u90fd\u5927\u5b66\u5b66\u8005\u4ece\u8f83\u4e3a\u7406\u8bba\u7684\u89d2\u5ea6\u5199\u4e86\u4e00\u7bc7\u76f8\u5173\u7684\u7814\u7a76\u7efc\u8ff0\uff0c\u4e3b\u8981\u63a2\u7a76\u56fe\u7f51\u7edc\u4e0eWL test\u7684\u5173\u7cfb\uff0c\u4ee5\u53ca\u56fe\u7f51\u7edc\u80fd\u591f\u89e3\u51b3\u7684\u95ee\u9898\u548c\u56fe\u7f51\u7edc\u4e0d\u80fd\u591f\u89e3\u51b3\u7684\u95ee\u9898\u3002<\/span><\/section>\n

<\/section>\n
\u56fe\u795e\u7ecf\u7f51\u7edc\u662f\u89e3\u51b3\u5404\u79cd\u56fe\u5b66\u4e60\u95ee\u9898\u7684\u6709\u6548\u7684\u673a\u5668\u5b66\u4e60\u6a21\u578b\u3002\u5c3d\u7ba1\u5b83\u4eec\u53d6\u5f97\u4e86\u7ecf\u9a8c\u4e0a\u7684\u6210\u529f\uff0c\u4f46\u662fGNNs\u7684\u7406\u8bba\u5c40\u9650\u6027\u6700\u8fd1\u5df2\u7ecf\u88ab\u63ed\u793a\u51fa\u6765\u3002\u56e0\u6b64\uff0c\u4eba\u4eec\u63d0\u51fa\u4e86\u5404\u5f0f\u5404\u6837\u7684GNN\u6a21\u578b\u6765\u653b\u514b\u8fd9\u4e9b\u9650\u5236\u3002\u5728\u8be5\u7814\u7a76\u7efc\u8ff0\u4e2d\uff0c\u4f5c\u8005\u5168\u9762\u6982\u8ff0\u4e86GNNs\u7684\u8868\u8fbe\u80fd\u529b\u548c\u53ef\u8bc1\u660e\u7684\u5f3a\u5927\u7684GNNs\u53d8\u4f53\u3002\u4e0b\u9762\u662f\u672c\u6587\u7684\u4e3b\u8981\u7ae0\u8282\uff1aSection 2 GNN\u4e0d\u80fd\u5224\u522b\u7684\u4f8b\u5b50; Section 3 GNN\u4e0eWL algorithm\u7684\u8054\u7cfb; Section 4 GNN\u80fd\u591f\u548c\u4e0d\u80fd\u591f\u89e3\u51b3\u7684\u7ec4\u5408\u95ee\u9898; Section 5 GNN, WL algorithm \u4ee5\u53ca distributed local algorithm\u7684\u5173\u7cfb<\/span><\/section>\n

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

\n
\n


<\/p>\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

8. Relational inductive biases, deep learning, and graph networks<\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

<\/section>\n<\/section>\n<\/h3>\n<\/section>\n<\/section>\n<\/section>\n
\u65f6\u95f4\uff1a2018\u5e746\u6708\u9996\u53d1\uff1b2018\u5e7411\u6708\u7b2c\u4e09\u7248<\/span><\/section>\n
\u4f5c\u8005\uff1aPeter W. Battaglia\u7b49<\/span><\/section>\n
\u5355\u4f4d\uff1a1DeepMind; 2Google Brain; 3MIT; 4University of Edinburgh<\/span><\/section>\n
\u7bc7\u5e45\uff1a40\u9875<\/span><\/section>\n
\u94fe\u63a5\uff1a<\/span><\/section>\n
https:\/\/arxiv.org\/pdf\/1806.01261.pdf<\/span><\/section>\n

<\/section>\n
\u8fd9\u7bc7\u6587\u7ae0\u8054\u5408\u4e86DeepMind\u3001\u8c37\u6b4c\u5927\u8111\u3001MIT\u548c\u7231\u4e01\u5821\u5927\u5b66\u768427\u540d\u4f5c\u8005<\/span>\uff08\u5176\u4e2d22\u4eba\u6765\u81eaDeepMind\uff09<\/span>\uff0c\u752837\u9875\u7684\u7bc7\u5e45\uff0c\u5bf9\u5173\u7cfb\u5f52\u7eb3\u504f\u7f6e\u548c\u56fe\u7f51\u7edc<\/span>\uff08Graph network<\/span>\uff09\u8fdb\u884c\u4e86\u5168\u9762\u9610\u8ff0\u3002 \u5728\u8bba\u6587\u91cc\uff0c\u4f5c\u8005\u63a2\u8ba8\u4e86\u5982\u4f55\u5728\u6df1\u5ea6\u5b66\u4e60\u7ed3\u6784<\/span>\uff08\u6bd4\u5982\u5168\u8fde\u63a5\u5c42\u3001\u5377\u79ef\u5c42\u548c\u9012\u5f52\u5c42\uff09<\/span>\u4e2d\uff0c\u4f7f\u7528\u5173\u7cfb\u5f52\u7eb3\u504f\u7f6e<\/span>\uff08relational inductive biases\uff09<\/span>\uff0c\u4fc3\u8fdb\u5bf9\u5b9e\u4f53\u3001\u5bf9\u5173\u7cfb\uff0c\u4ee5\u53ca\u5bf9\u7ec4\u6210\u5b83\u4eec\u7684\u89c4\u5219\u8fdb\u884c\u5b66\u4e60\u3002<\/span><\/section>\n

<\/section>\n
\u8bba\u6587\u63d0\u51fa\u4e86\u4e00\u4e2a\u65b0\u7684AI\u6a21\u5757\u2014\u2014\u56fe\u7f51\u7edc<\/span>\uff08graph network\uff09<\/span>\uff0c\u5bf9\u4ee5\u524d\u5404\u79cd\u5bf9\u56fe\u8fdb\u884c\u64cd\u4f5c\u7684\u795e\u7ecf\u7f51\u7edc\u65b9\u6cd5\u7684\u63a8\u5e7f\u548c\u6269\u5c55\u3002\u56fe\u7f51\u7edc\u5177\u6709\u5f3a\u5927\u7684\u5173\u7cfb\u5f52\u7eb3\u504f\u7f6e\uff0c\u4e3a\u64cd\u7eb5\u7ed3\u6784\u5316\u77e5\u8bc6\u548c\u751f\u6210\u7ed3\u6784\u5316\u884c\u4e3a\u63d0\u4f9b\u4e86\u4e00\u4e2a\u76f4\u63a5\u7684\u754c\u9762\u3002\u4f5c\u8005\u8fd8\u8ba8\u8bba\u4e86\u56fe\u7f51\u7edc\u5982\u4f55\u652f\u6301\u5173\u7cfb\u63a8\u7406\u548c\u7ec4\u5408\u6cdb\u5316\uff0c\u4e3a\u66f4\u590d\u6742\u3001\u53ef\u89e3\u91ca\u548c\u7075\u6d3b\u7684\u63a8\u7406\u6a21\u5f0f\u6253\u4e0b\u57fa\u7840\u3002<\/span><\/section>\n

<\/span><\/section>\n
\u76f8\u5173\u62a5\u9053\u4e0e\u89e3\u8bfb\uff1a\u56fe\u7f51\u7edc\u8ba9\u6df1\u5ea6\u5b66\u4e60\u4e5f\u80fd\u56e0\u679c\u63a8\u7406<\/span>\uff08http:\/\/suo.im\/5V9FBj\uff09<\/span><\/section>\n

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

\n
\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

9. Adversarial Attack and Defense on Graph Data: A Survey<\/span><\/strong><\/strong><\/span> <\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

<\/section>\n<\/section>\n<\/h3>\n<\/section>\n<\/section>\n<\/section>\n
\u628a\u5bf9\u6297\u601d\u60f3\u5e94\u7528\u5230\u56fe\u6570\u636e\u662f\u4e00\u4e2a\u81ea\u7136\u7684\u8fc7\u7a0b\uff0c\u800c\u4e14\u5176\u5728\u5177\u4f53\u5e94\u7528\u573a\u666f\u4e2d\u4e5f\u6709\u5e7f\u6cdb\u7684\u6848\u4f8b\uff0c\u4f8b\u5982\u5728\u5728\u7ebf\u793e\u4ea4\u7f51\u7edc\u4e2d\uff0c\u6c34\u519b\u8d26\u6237\u901a\u8fc7\u5173\u6ce8\u6b63\u5e38\u8d26\u6237\uff0c\u53d1\u5e03\u65e5\u5e38\u5185\u5bb9\uff0c\u6765\u964d\u4f4e\u81ea\u5df1\u5728\u793e\u4ea4\u7f51\u7edc\u7684\u4e2d\u53ef\u7591\u5ea6\uff0c\u4ece\u800c\u907f\u514d\u88ab\u68c0\u6d4b\u5230\u800c\u5c01\u53f7\u3002<\/span><\/section>\n

<\/section>\n
\u56fe\u6570\u636e\u5bf9\u6297\u653b\u51fb\u7684\u5b9a\u4e49\u662f\uff1a\u7ed9\u5b9a\u4e00\u4e2a\u56fe<\/span>\uff08\u8282\u70b9\u548c\u8fb9\uff09<\/span>\uff0c\u901a\u8fc7\u4fee\u6539\u8fd9\u4e2a\u56fe\uff0c\u4f7f\u5f97\u8fd9\u79cd\u4fee\u6539\u5728\u4e0d\u88ab\u5bdf\u89c9\u5230\u60c5\u51b5\u4e0b\uff0c\u80fd\u591f\u964d\u4f4e\u7b97\u6cd5<\/span>\uff08\u4f8b\u5982\u8282\u70b9\u5206\u7c7b\u548c\u94fe\u63a5\u9884\u6d4b\uff09<\/span>\u5728\u56fe\u6570\u636e\u4e0a\u9762\u7684\u8868\u73b0\u3002<\/span><\/section>\n

<\/section>\n
\u548c\u4f18\u5316\u95ee\u9898\u76f8\u4f3c\uff0c\u653b\u51fb\u65b9\u7684\u76ee\u6807\u662f\u5728\u4e00\u5b9a\u7ea6\u675f\u7684\u6761\u4ef6\u4e0b\u6700\u5927\u53ef\u80fd\u5730\u5e72\u6270\u7b97\u6cd5\u7684\u8868\u73b0\u3002\u5728KDD 2018 [1]\u7684\u6587\u7ae0\u4e2d\uff0c\u4f5c\u8005\u5c06\u56fe\u88ab\u4fee\u6539\u4e4b\u540e\u5e26\u6765\u7684\u53d8\u5316\u5b9a\u4e49\u4e3aperturbations<\/span>\uff08\u6270\u52a8\uff09<\/span>\uff0c\u5373\u653b\u51fb\u8005\u5bf9\u56fe\u6570\u636e\u7684\u653b\u51fb\u540e\u7684\u6270\u52a8\u5fc5\u987b\u6ee1\u8db3\u4e0e\u4e00\u4e9b\u7ea6\u675f\u3002\u4f8b\u5982\uff0c\u6211\u4eec\u52a0\u8fb9\u6216\u8005\u51cf\u8fb9\u4e0d\u80fd\u592a\u591a\u6761\uff0c\u8981\u4fdd\u6301\u56fe\u7684\u57fa\u672c\u7ed3\u6784\u7b49\u7b49\uff0c\u8fd9\u662f\u5bf9\u6297\u653b\u51fb\u5b9e\u73b0\u7684\u524d\u63d0\u3002<\/span><\/section>\n

<\/section>\n
[1] Z\u00fcgner, Daniel, Amir Akbarnejad, and Stephan G\u00fcnnemann. “Adversarial attacks on neural networks for graph data.”<\/strong> In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp. 2847-2856. ACM, 2018<\/span><\/section>\n

<\/section>\n
\u6839\u636e\u56fe\u6570\u636e\u7684\u7279\u70b9\uff0c\u53ef\u4ee5\u56fe\u6570\u636e\u5bf9\u6297\u653b\u51fb\u4efb\u52a1\u5206\u4e3a\u4e09\u7c7b\uff1a<\/span><\/section>\n

<\/section>\n
\u8282\u70b9\u76f8\u5173\u7684\u4efb\u52a1<\/span>\uff08Node-related Task\uff09<\/span>\uff1a\u5bf9\u8282\u70b9\u5206\u7c7b\u4efb\u52a1\u8fd8\u6709\u5bf9\u8282\u70b9\u5d4c\u5165\uff08node embedding\uff09\u7684\u653b\u51fb\u90fd\u5c5e\u4e8e\u8282\u70b9\u5c42\u9762\u7684\u653b\u51fb\uff0c\u5176\u76ee\u7684\u662f\u8ba9\u5206\u7c7b\u5668\u5206\u9519\uff0c\u964d\u4f4e\u5176\u7cbe\u786e\u7387\u6216\u8005\u53ec\u56de\u7387\u3002\u56e0\u4e3a\u8282\u70b9\u5206\u7c7b\u662f\u76ee\u524d\u56fe\u6570\u636e\u7684\u4e3b\u8981\u4efb\u52a1\uff0c\u56e0\u6b64\u5927\u90e8\u5206\u56fe\u6570\u636e\u5bf9\u6297\u653b\u51fb\u8bba\u6587\u90fd\u6709\u7814\u7a76\u5230\u8282\u70b9\u5206\u7c7b\u4efb\u52a1\uff1b
<\/span><\/section>\n

<\/section>\n
\u94fe\u63a5\u76f8\u5173\u7684\u4efb\u52a1<\/span>\uff08Link-related Task\uff09<\/span>\uff1a\u94fe\u63a5\u9884\u6d4b<\/span>\uff08Link Prediction\uff09<\/span>\u662f\u56fe\u6570\u636e\u4e0a\u9762\u7684\u53e6\u4e00\u4e2a\u4e3b\u8981\u4efb\u52a1\uff0c\u63a8\u8350\u7cfb\u7edf\uff0c\u77e5\u8bc6\u56fe\u8c31\uff0c\u793e\u4ea4\u7f51\u7edc\u90fd\u7528\u5230\u5b83\u3002\u5bf9\u4e8e\u94fe\u63a5\u5c42\u9762\u7684\u653b\u51fb\uff0c\u4e3b\u8981\u76ee\u7684\u8ba9\u7b97\u6cd5\u9884\u6d4b\u5230\u9519\u7684\u94fe\u63a5\u76ee\u6807\uff1b
<\/span><\/section>\n

<\/section>\n
\u5168\u56fe\u76f8\u5173\u7684\u4efb\u52a1<\/span>\uff08Graph-related Task\uff09<\/span>\uff1a\u5168\u56fe\u76f8\u5173\u7684\u4efb\u52a1\u4e3b\u8981\u662f\u5bf9\u6574\u4e2a\u56fe\u7684\u5206\u7c7b\uff0c\u5e38\u89c1\u4e8e\u5bf9\u751f\u7269\u7ed3\u6784\u7684\u5206\u7c7b\u4efb\u52a1\u3002\u4e00\u822c\u662f\u5b66\u4e60\u56fe\u6574\u4f53\u7ed3\u6784\u7684\u4f4e\u7ef4\u5d4c\u5165\uff0c\u7136\u540e\u8fdb\u884c\u5206\u7c7b<\/span><\/section>\n

<\/section>\n
\u76f8\u5173\u7684\u7ae0\u8282\u76ee\u5f55<\/span><\/strong><\/span><\/section>\n

<\/section>\n
section 3: \u4ece\u5404\u79cd\u89c6\u89d2\u5bf9\u5f53\u524d\u7684\u653b\u51fb\u65b9\u6cd5\u8fdb\u884c\u5206\u7c7b\u3002<\/span><\/section>\n
section 4\uff1a\u5bf9\u5178\u578b\u7684\u7b97\u6cd5\u8fdb\u884c\u56de\u987e\u3002<\/span><\/section>\n
section 5: \u6839\u636e\u9632\u5fa1\u7b56\u7565\u5bf9\u5f53\u524d\u7684\u9632\u6cbb\u63aa\u65bd\u8fdb\u884c\u5206\u7c7b\uff0c\u5e76\u4e14\u5bf9\u6bcf\u4e00\u7c7b\u4e2d\u7684\u4ee3\u8868\u6027\u7b97\u6cd5\u8fdb\u884c\u56de\u987e\u3002<\/span><\/section>\n
section 6: \u57fa\u4e8e\u6211\u4eec\u5f00\u53d1\u7684\u77e5\u8bc6\u5e93\u8fdb\u884c\u76f8\u5173\u7684\u7814\u7a76\u548c\u63a2\u8ba8\uff0c\u8be5\u77e5\u8bc6\u5e93\u53ef\u4ee5\u5bf9\u57fa\u4e8e\u56fe\u7684\u653b\u51fb\u548c\u9632\u5fa1\u63d0\u4f9b\u4e00\u4e2a\u5168\u9762\u7684\u7406\u89e3\u3002<\/span><\/section>\n
section 7: \u5bf9\u8be5\u9886\u57df\u672a\u6765\u7684\u65b9\u5411\u548c\u524d\u666f\u8fdb\u884c\u63a2\u8ba8\u3002<\/span><\/section>\n

<\/span><\/section>\n
\u76f8\u5173\u62a5\u9053\u4e0e\u89e3\u8bfb\uff1a<\/span><\/strong><\/span><\/section>\n
\u3010\u7efc\u8ff0\u3011\u57fa\u4e8e\u56fe\u7684\u5bf9\u6297\u5f0f\u653b\u51fb\u548c\u9632\u5fa1\uff0c\u964422\u9875\u8bba\u6587\u4e0b\u8f7d<\/span><\/a><\/section>\n

\n
\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

10. \u5f02\u8d28\u56fe<\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

Heterogeneous Network Representation Learning: Survey, Benchmark, Evaluation, and Beyond<\/span><\/strong><\/section>\n<\/section>\n<\/h3>\n
\u65f6\u95f4\uff1a2020\u5e744\u67081\u65e5<\/span><\/section>\n
\u4f5c\u8005\uff1aUIUC Carl Yang, Yuxin Xiao, Yu Zhang, Yizhou Sun, and Jiawei Han,<\/span><\/section>\n
\u5355\u4f4d\uff1aUIUC\u7b49<\/span><\/section>\n


<\/p>\n

\u8fd9\u7bc7\u8bba\u6587\u6587\u7ae0\u7ed3\u6784\u5982\u4e0b\uff0c\u7b2c2\u8282\u9996\u5148\u4ecb\u7ecd\u4f5c\u8005\u63d0\u51fa\u7684\u901a\u7528HNE\u8303\u5f0f\u3002\u968f\u540e\uff0c\u7b2c3\u8282\u5bf9\u4ee3\u8868\u6027\u6a21\u578b\u8fdb\u884c\u4e86\u6982\u5ff5\u4e0a\u7684\u5206\u7c7b\u548c\u5206\u6790\u3002\u7136\u540e\uff0c\u5728\u7b2c4\u8282\u4e2d\u63d0\u4f9b\u4e86\u57fa\u51c6\u6570\u636e\u96c6\uff0c\u5e76\u8fdb\u884c\u4e86\u6df1\u5165\u7684\u5206\u6790\u3002\u5728\u7b2c5\u8282\u4e2d\uff0c\u5bf910\u79cd\u5e38\u7528\u7684HNE\u7b97\u6cd5\u8fdb\u884c\u4e86\u7cfb\u7edf\u800c\u5168\u9762\u7684\u5b9e\u8bc1\u7814\u7a76\uff0c\u5bf9HNE\u7684\u53d1\u5c55\u73b0\u72b6\u8fdb\u884c\u4e86\u8bc4\u4ef7\u3002\u7b2c\u516d\u90e8\u5206\u662f\u5bf9\u672a\u6765HNE\u5e73\u53f0\u4f7f\u7528\u548c\u7814\u7a76\u7684\u5c55\u671b\u3002<\/span><\/section>\n
<\/p>\n
\u5f02\u6784\u7f51\u7edc\u793a\u4f8b<\/span><\/section>\n<\/figure>\n


<\/h3>\n

\u7b97\u6cd5\u5206\u7c7b<\/span><\/strong><\/h3>\n


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

Proximity-Preserving Methods<\/span><\/strong><\/section>\n

<\/span><\/strong><\/section>\n
\u5982\u524d\u6240\u8ff0\uff0c\u7f51\u7edc\u5d4c\u5165\u7684\u4e00\u4e2a\u57fa\u672c\u76ee\u6807\u662f\u6355\u83b7\u7f51\u7edc\u62d3\u6251\u4fe1\u606f\u3002\u8fd9\u53ef\u4ee5\u901a\u8fc7\u5728\u8282\u70b9\u4e4b\u95f4\u4fdd\u7559\u4e0d\u540c\u7c7b\u578b\u7684\u90bb\u8fd1\u6027\u6765\u5b9e\u73b0\u3002\u5728HNE\u4e2d\uff0c\u6709\u4e24\u7c7b\u4e3b\u8981\u7684\u63a5\u8fd1\u6027\u4fdd\u62a4\u65b9\u6cd5:\u57fa\u4e8e\u968f\u673a\u6b65\u6cd5\u7684\u65b9\u6cd5<\/span>(\u7075\u611f\u6765\u81eaDeepWalk[49])\u548c\u57fa\u4e8e\u4e00\u9636\/\u4e8c\u9636\u63a5\u8fd1\u6027\u7684\u65b9\u6cd5(\u7075\u611f\u6765\u81eaLINE[72])<\/span>\u3002<\/span><\/section>\n
<\/figure>\n


<\/p>\n

Message-Passing Methods<\/span><\/strong><\/section>\n
\u7f51\u7edc\u4e2d\u7684\u6bcf\u4e2a\u8282\u70b9\u90fd\u53ef\u4ee5\u5c06\u5c5e\u6027\u4fe1\u606f\u8868\u793a\u4e3a\u7279\u5f81\u5411\u91cfxu\u3002\u6d88\u606f\u4f20\u9012\u65b9\u6cd5\u7684\u76ee\u6807\u662f\u901a\u8fc7\u805a\u5408\u6765\u81eau\u90bb\u5c45\u7684\u4fe1\u606f\u6765\u5b66\u4e60\u57fa\u4e8exu\u7684\u8282\u70b9\u5d4c\u5165eu\u3002\u5728\u6700\u8fd1\u7684\u7814\u7a76\u4e2d\uff0c\u56fe\u795e\u7ecf\u7f51\u7edc<\/span>(GNNs)<\/span>[37]\u88ab\u5e7f\u6cdb\u7528\u4e8e\u4fc3\u8fdb\u8fd9\u79cd\u805a\u5408\/\u6d88\u606f\u4f20\u9012\u8fc7\u7a0b\u3002<\/span><\/section>\n
<\/figure>\n

<\/section>\n
Relation-Learning\u65b9\u6cd5<\/span><\/strong><\/section>\n
\u5f02\u7c7b\u7f51\u7edc\u4e2d\u7684\u6bcf\u6761\u8fb9\u90fd\u53ef\u4ee5\u770b\u4f5c\u662f\u4e00\u4e2a\u4e09\u5143\u7ec4<\/span>(u, l, v)<\/span>\uff0c\u7531\u4e24\u4e2a\u8282\u70b9u, v\u2208v\u548c\u4e00\u4e2a\u8fb9\u7f18\u7c7b\u578bl\u2208TE<\/span>(\u5373\u3002\uff0c\u5b9e\u4f53\u548c\u5173\u7cfb\uff0c\u7528KG\u8868\u793a)<\/span>\u3002\u5173\u7cfb\u5b66\u4e60\u65b9\u6cd5\u7684\u76ee\u6807\u662f\u5b66\u4e60\u4e00\u4e2a\u8bc4\u5206\u51fd\u6570sl<\/span>(u, v)<\/span>\uff0c\u8be5\u51fd\u6570\u5bf9\u4efb\u610f\u4e09\u5143\u7ec4\u6c42\u503c\u5e76\u8f93\u51fa\u4e00\u4e2a\u6807\u91cf\u6765\u5ea6\u91cf\u8be5\u4e09\u5143\u7ec4\u7684\u53ef\u63a5\u53d7\u6027\u3002\u8fd9\u79cd\u601d\u60f3\u5728KB\u5d4c\u5165\u4e2d\u88ab\u5e7f\u6cdb\u91c7\u7528\u3002\u7531\u4e8e\u5df2\u7ecf\u6709\u5173\u4e8eKB\u5d4c\u5165\u7b97\u6cd5\u7684\u8c03\u67e5[81]\uff0c\u6211\u4eec\u5728\u8fd9\u91cc\u53ea\u8ba8\u8bba\u6700\u6d41\u884c\u7684\u65b9\u6cd5\uff0c\u5e76\u5f3a\u8c03\u5b83\u4eec\u4e0eHNE\u7684\u8054\u7cfb\u3002<\/span><\/section>\n
<\/figure>\n

\n
\n


<\/p>\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

11. \u7efc\u8ff0\uff1a\u56fe\u4e0a\u7684AutoML<\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

<\/section>\n<\/section>\n<\/h3>\n
Automated Machine Learning on Graphs: A Survey<\/span><\/strong><\/section>\n
\u4f5c\u8005\uff1aZiwei Zhang, Xin Wang, Wenwu Zhu<\/span><\/section>\n
\u5355\u4f4d\uff1a\u6e05\u534e\u5927\u5b66 <\/span><\/section>\n
\u94fe\u63a5\uff1a<\/span><\/section>\n
https:\/\/arxiv.org\/abs\/2103.00742<\/span><\/section>\n

<\/span><\/section>\n
\u56fe\u673a\u5668\u5b66\u4e60\u5728\u5b66\u672f\u754c\u548c\u5de5\u4e1a\u754c\u90fd\u5f97\u5230\u4e86\u5e7f\u6cdb\u7684\u7814\u7a76\u3002\u7136\u800c\uff0c\u968f\u7740\u56fe\u5b66\u4e60\u6587\u732e\u7684\u5927\u91cf\u6d8c\u73b0\uff0c\u6d8c\u73b0\u51fa\u5927\u91cf\u7684\u65b9\u6cd5\u548c\u6280\u672f\uff0c\u5bf9\u4e8e\u4e0d\u540c\u7684\u56fe\u76f8\u5173\u4efb\u52a1\uff0c\u4eba\u5de5\u8bbe\u8ba1\u6700\u4f18\u7684\u673a\u5668\u5b66\u4e60\u7b97\u6cd5\u53d8\u5f97\u8d8a\u6765\u8d8a\u56f0\u96be\u3002\u4e3a\u4e86\u89e3\u51b3\u8fd9\u4e00\u5173\u952e\u6311\u6218\uff0c\u56fe\u4e0a\u7684\u81ea\u52a8\u673a\u5668\u5b66\u4e60<\/span>(AutoML)<\/span>\u7ed3\u5408\u4e86\u56fe\u673a\u5668\u5b66\u4e60\u548c\u81ea\u52a8\u5b66\u4e60\u7684\u4f18\u70b9\uff0c\u6b63\u53d7\u5230\u5b66\u672f\u754c\u7684\u5173\u6ce8\u3002\u56e0\u6b64\uff0c\u672c\u6587\u5bf9\u56fe\u81ea\u52a8\u5b66\u4e60\u65b9\u6cd5\u8fdb\u884c\u4e86\u5168\u9762\u7684\u7efc\u8ff0\uff0c\u91cd\u70b9\u7814\u7a76\u4e86\u7528\u4e8e\u56fe\u673a\u5668\u5b66\u4e60\u7684\u8d85\u53c2\u6570\u4f18\u5316<\/span>(HPO)<\/span>\u548c\u795e\u7ecf\u7ed3\u6784\u641c\u7d22<\/span>(NAS)<\/span>\u3002\u672c\u6587\u8fdb\u4e00\u6b65\u6982\u8ff0\u4e86\u4e0e\u81ea\u52a8\u56fe\u673a\u5668\u5b66\u4e60\u76f8\u5173\u7684\u5e93\uff0c\u5e76\u6df1\u5165\u8ba8\u8bba\u4e86AutoGL\uff0c\u8fd9\u662f\u7b2c\u4e00\u4e2a\u7528\u4e8eAutoML on graphs\u7684\u4e13\u7528\u5f00\u6e90\u5e93\u3002\u6700\u540e\uff0c\u4f5c\u8005\u5206\u4eab\u4e86\u5bf9\u81ea\u52a8\u56fe\u673a\u5b66\u4e60\u672a\u6765\u7814\u7a76\u65b9\u5411\u7684\u770b\u6cd5\u3002\u636e\u4f5c\u8005\u6240\u77e5\uff0c\u8fd9\u7bc7\u8bba\u6587\u662f\u7b2c\u4e00\u4e2a\u5173\u4e8e\u56fe\u4e0a\u7684\u81ea\u52a8\u673a\u5668\u5b66\u4e60\u7684\u7cfb\u7edf\u548c\u5168\u9762\u7684\u7efc\u8ff0\u3002<\/strong>
<\/span><\/section>\n

<\/section>\n
\n

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


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

\n
\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

12. \u7efc\u8ff0\uff1a\u56fe\u81ea\u76d1\u7763\u5b66\u4e60<\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

<\/section>\n<\/section>\n<\/h3>\n<\/section>\n<\/section>\n<\/section>\n
Graph Self-Supervised Learning: A Survey<\/span><\/strong><\/section>\n
\u4f5c\u8005\uff1aYixin Liu, Shirui Pan, Ming Jin, Chuan Zhou, Feng Xia, Philip S. Yu<\/span><\/section>\n
\u5355\u4f4d\uff1aMonash University\u7b49 <\/span><\/section>\n

<\/section>\n
\u6700\u8fd1\uff0c\u57fa\u4e8e\u56fe\u7684\u6df1\u5ea6\u5b66\u4e60\u5f15\u8d77\u4e86\u4eba\u4eec\u7684\u6781\u5927\u5174\u8da3\u3002\u4f46\u662f\uff0c\u5927\u591a\u6570\u5de5\u4f5c\u90fd\u96c6\u4e2d\u5728<\/span>\uff08\u534a\uff09<\/span>\u76d1\u7763\u5b66\u4e60\u4e0a\uff0c\u5bfc\u81f4\u5bf9\u6807\u7b7e\u7684\u4f9d\u8d56\u5ea6\u8fc7\u9ad8\uff0c\u6cdb\u5316\u6027\u5dee\u548c\u9c81\u68d2\u6027\u5dee\u3002\u4e3a\u4e86\u89e3\u51b3\u8fd9\u4e9b\u95ee\u9898\uff0c\u81ea\u76d1\u7763\u5b66\u4e60<\/span>\uff08SSL\uff09<\/span>\u53ef\u4ee5\u901a\u8fc7\u7cbe\u5fc3\u8bbe\u8ba1\u7684\u4efb\u52a1\u63d0\u53d6\u4fe1\u606f\u77e5\u8bc6\uff0c\u800c\u65e0\u9700\u4f9d\u8d56\u4eba\u5de5\u6807\u7b7e\uff0c\u8fd9\u5df2\u7ecf\u6210\u4e3a\u4e00\u79cdpromising\u7684\u8d8b\u52bf\u56fe\u5b66\u4e60\u8303\u4f8b\u3002\u4e0e\u8ba1\u7b97\u673a\u89c6\u89c9\/\u81ea\u7136\u8bed\u8a00\u5904\u7406\u7b49\u5176\u4ed6\u9886\u57df\u4e0d\u540c\uff0c\u56fe\u4e0a\u7684SSL\u5177\u6709\u4e13\u6709\u7684\u80cc\u666f\uff0c\u8bbe\u8ba1\u601d\u60f3\u548c\u5206\u7c7b\u6cd5\u3002\u5728\u56fe\u81ea\u76d1\u7763\u5b66\u4e60\u7684\u4fdd\u62a4\u4e0b\uff0c\u672c\u6587\u5bf9\u4f7f\u7528SSL\u6280\u672f\u5904\u7406\u56fe\u6570\u636e\u7684\u73b0\u6709\u65b9\u6cd5\u8fdb\u884c\u4e86\u53ca\u65f6\u800c\u5168\u9762\u7684\u56de\u987e\u3002\u6839\u636e\u5176\u81ea\u76d1\u7763\u4efb\u52a1\u7684\u8bbe\u8ba1\uff0c\u672c\u6587\u5c06\u5b83\u4eec\u5206\u4e3a\u56db\u7c7b\u3002\u8fdb\u4e00\u6b65\u8ba8\u8bba\u8be5\u7814\u7a76\u9886\u57df\u4e2d\u5c1a\u5b58\u7684\u6311\u6218\u548c\u6f5c\u5728\u7684\u672a\u6765\u65b9\u5411\u3002<\/span><\/section>\n

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

<\/p>\n


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

\n
\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

13. A unified Review: \u56fe\u795e\u7ecf\u7f51\u7edc\u81ea\u76d1\u7763\u5b66\u4e60<\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

<\/section>\n<\/section>\n<\/h3>\n<\/section>\n<\/section>\n<\/section>\n
Self-Supervised Learning of Graph Neural Networks: A Unified Review<\/span><\/strong><\/section>\n
\u4f5c\u8005\uff1aYaochen Xie, Zhao Xu, Zhengyang Wang, Shuiwang Ji<\/span>
<\/section>\n
\u5355\u4f4d\uff1aTexas A&M University\u7b49 <\/span><\/section>\n
\u94fe\u63a5\uff1a<\/span><\/section>\n
https:\/\/arxiv.org\/abs\/2102.10757<\/span><\/section>\n

<\/section>\n
\u5728\u76d1\u7763\u6a21\u5f0f\u4e0b\u8bad\u7ec3\u7684\u6df1\u5ea6\u6a21\u578b\u5728\u5404\u79cd\u4efb\u52a1\u4e0a\u90fd\u53d6\u5f97\u4e86\u663e\u8457\u7684\u6210\u529f\u3002\u5728\u6807\u8bb0\u6837\u672c\u6709\u9650\u7684\u60c5\u51b5\u4e0b\uff0c\u81ea\u76d1\u7763\u5b66\u4e60<\/span>(self-supervised learning, SSL)<\/span>\u6210\u4e3a\u5229\u7528\u5927\u91cf\u672a\u6807\u8bb0\u6837\u672c\u7684\u65b0\u8303\u5f0f\u3002SSL\u5728\u81ea\u7136\u8bed\u8a00\u548c\u56fe\u50cf\u5b66\u4e60\u4efb\u52a1\u4e2d\u5df2\u7ecf\u53d6\u5f97\u4e86\u5f88\u597d\u7684\u6548\u679c\u3002\u6700\u8fd1\uff0c\u5229\u7528\u56fe\u795e\u7ecf\u7f51\u7edc<\/span>(GNNs)<\/span>\u5c06\u8fd9\u79cd\u6210\u529f\u6269\u5c55\u5230\u56fe\u6570\u636e\u7684\u8d8b\u52bf\u3002<\/span><\/section>\n

<\/section>\n
\u5728\u672c\u7efc\u8ff0\u8bba\u6587\u4e2d\uff0c\u4f5c\u8005\u63d0\u4f9b\u4e86\u4f7f\u7528SSL\u8bad\u7ec3GNN\u7684\u4e0d\u540c\u65b9\u6cd5\u7684\u7edf\u4e00\u56de\u987e\u3002\u5177\u4f53\u6765\u8bf4\uff0c\u5c06SSL\u65b9\u6cd5\u5206\u4e3a\u5bf9\u6bd4\u6a21\u578b\u548c\u9884\u6d4b\u6a21\u578b\u3002<\/span><\/strong>\u5728\u8fd9\u4e24\u7c7b\u4e2d\uff0c\u6211\u4eec\u90fd\u4e3a\u65b9\u6cd5\u63d0\u4f9b\u4e86\u4e00\u4e2a\u7edf\u4e00\u7684\u6846\u67b6\uff0c\u4ee5\u53ca\u8fd9\u4e9b\u65b9\u6cd5\u5728\u6846\u67b6\u4e0b\u7684\u6bcf\u4e2a\u7ec4\u4ef6\u4e2d\u7684\u4e0d\u540c\u4e4b\u5904\u3002\u6211\u4eec\u5bf9GNNs SSL\u65b9\u6cd5\u7684\u7edf\u4e00\u5904\u7406\u63ed\u793a\u4e86\u5404\u79cd\u65b9\u6cd5\u7684\u5f02\u540c\uff0c\u4e3a\u5f00\u53d1\u65b0\u7684\u65b9\u6cd5\u548c\u7b97\u6cd5\u5960\u5b9a\u4e86\u57fa\u7840\u3002\u6211\u4eec\u8fd8\u603b\u7ed3\u4e86\u4e0d\u540c\u7684SSL\u8bbe\u7f6e\u548c\u6bcf\u4e2a\u8bbe\u7f6e\u4e2d\u4f7f\u7528\u7684\u76f8\u5e94\u6570\u636e\u96c6\u3002\u4e3a\u4e86\u4fc3\u8fdb\u65b9\u6cd5\u5f00\u53d1\u548c\u5b9e\u8bc1\u6bd4\u8f83\uff0c\u6211\u4eec\u4e3aGNNs\u4e2d\u7684SSL\u5f00\u53d1\u4e86\u4e00\u4e2a\u6807\u51c6\u5316\u6d4b\u8bd5\u5e8a\uff0c\u5305\u62ec\u901a\u7528\u57fa\u7ebf\u65b9\u6cd5\u3001\u6570\u636e\u96c6\u548c\u8bc4\u4f30\u6307\u6807\u7684\u5b9e\u73b0\u3002<\/span><\/section>\n

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

\n
\n


<\/p>\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

14 . \u56fe\u4e0a\u7684\u5143\u5b66\u4e60: \u65b9\u6cd5\u548c\u5e94\u7528<\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

<\/section>\n<\/section>\n<\/h3>\n<\/section>\n
Meta-Learning with Graph Neural Networks: Methods and Applications<\/span><\/strong><\/section>\n
\u4f5c\u8005\uff1aDebmalya Mandal, Sourav Medya, Brian Uzzi, Charu Aggarwal<\/span>
<\/section>\n
\u5355\u4f4d\uff1aColumbia University\u7b49<\/span><\/section>\n
\u94fe\u63a5\uff1a<\/span><\/section>\n
https:\/\/arxiv.org\/abs\/2103.00137<\/span><\/section>\n

<\/section>\n
\u56fe\u795e\u7ecf\u7f51\u7edc<\/span>\uff08GNN\uff09<\/span>\u662f\u5bf9\u56fe\u6570\u636e\u8fdb\u884c\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u7684\u6cdb\u5316\uff0cGNN\u5df2\u5e7f\u6cdb\u7528\u4e8e\u4ece\u836f\u7269\u53d1\u73b0\u5230\u63a8\u8350\u7cfb\u7edf\u7684\u5404\u4e2a\u9886\u57df\u3002\u4f46\u662f\uff0c\u5f53\u53ef\u7528\u6837\u672c\u5f88\u5c11\u65f6\uff0c\u6b64\u7c7b\u5e94\u7528\u4e0a\u7684GNN\u53d7\u9650\u5236\u3002\u5143\u5b66\u4e60\u4e00\u76f4\u662f\u89e3\u51b3\u673a\u5668\u5b66\u4e60\u4e2d\u6837\u672c\u4e0d\u8db3\u7684\u91cd\u8981\u6846\u67b6\uff0c\u8fd1\u5e74\u6765\uff0c\u7814\u7a76\u4eba\u5458\u5df2\u5f00\u59cb\u5c06\u5143\u5b66\u4e60\u5e94\u7528\u4e8eGNN\u3002\u5728\u8fd9\u9879\u5de5\u4f5c\u4e2d\uff0c\u4f5c\u8005\u5bf9\u56fe\u4e0a\u4e0d\u540c\u7684\u5143\u5b66\u4e60\u65b9\u6cd5\u8fdb\u884c\u4e86\u5168\u9762\u8c03\u67e5\uff0c\u5e76\u4e14\u6839\u636e\u7ed3\u6784\uff0c\u8868\u793a\u5f62\u5f0f\u548c\u5e94\u7528\u5bf9\u6587\u732e\u8fdb\u884c\u5206\u7c7b\u3002\u6700\u540e\uff0c\u8ba8\u8bba\u4e86\u4e00\u4e9b\u6709\u610f\u601d\u7684\u7814\u7a76\u65b9\u5411\u548c\u672a\u89e3\u51b3\u7684\u95ee\u9898\u3002<\/span><\/section>\n

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

<\/p>\n


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

\n
\n

 <\/span>
<\/p>\n<\/section>\n<\/h1>\n

\n
\n
\n
\n
\n

<\/section>\n

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

15 . \u7efc\u8ff0\uff1a\u56fe\u7ed3\u6784\u5b66\u4e60\u7528\u4e8e\u9c81\u68d2\u6027\u8868\u793a<\/span><\/strong><\/strong><\/span><\/p>\n<\/section>\n

\n

<\/section>\n

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

<\/section>\n<\/section>\n<\/h3>\n<\/section>\n<\/section>\n<\/section>\n
Deep Graph Structure Learning for Robust Representations: A Survey<\/span><\/strong><\/section>\n
\u5355\u4f4d<\/span>\uff1a\u4e2d\u79d1\u9662\u81ea\u52a8\u5316\u7814\u7a76\u6240<\/span><\/section>\n
\u4f5c\u8005<\/span>\uff1aYanqiao Zhu , Weizhi Xu , Jinghao Zhang , Qiang Liu, Shu Wu\u2020 and Liang Wang<\/span><\/section>\n
\u94fe\u63a5\uff1a<\/span><\/section>\n
https:\/\/arxiv.org\/pdf\/2103.03036.pdf<\/span><\/section>\n

<\/section>\n
\u56fe\u795e\u7ecf\u7f51\u7edc<\/span>\uff08GNN\uff09<\/span>\u88ab\u5e7f\u6cdb\u7528\u4e8e\u5206\u6790\u56fe\u7ed3\u6784\u6570\u636e\u3002\u5927\u591a\u6570GNN\u65b9\u6cd5\u5bf9\u56fe\u7ed3\u6784\u7684\u8d28\u91cf\u9ad8\u5ea6\u654f\u611f\uff0c\u5e76\u4e14\u901a\u5e38\u9700\u8981\u7528\u4e8e\u5b66\u4e60\u4fe1\u606f\u5d4c\u5165\u7684\u5b8c\u7f8e\u56fe\u7ed3\u6784\u3002\u4f46\u662f\uff0c\u56fe\u4e2d\u7684\u566a\u58f0\u65e0\u5904\u4e0d\u5728\uff0c\u56e0\u6b64\u6709\u5fc5\u8981\u4e3a\u73b0\u5b9e\u4e16\u754c\u4e2d\u7684\u95ee\u9898\u5b66\u4e60\u53ef\u9760\u7684\u8868\u793a\u5f62\u5f0f\u3002\u4e3a\u4e86\u63d0\u9ad8GNN\u6a21\u578b\u7684\u9c81\u68d2\u6027\uff0c\u56f4\u7ed5\u56fe\u7ed3\u6784\u5b66\u4e60<\/span>\uff08GSL\uff09<\/span>\u7684\u4e2d\u5fc3\u6982\u5ff5\u63d0\u51fa\u4e86\u8bb8\u591a\u7814\u7a76\uff0c\u65e8\u5728\u5171\u540c\u5b66\u4e60\u4f18\u5316\u7684\u56fe\u7ed3\u6784\u548c\u76f8\u5e94\u7684\u8868\u793a\u5f62\u5f0f\u3002\u4e3a\u6b64\uff0c\u5728\u63d0\u51fa\u7684\u8c03\u67e5\u4e2d\uff0c\u4f5c\u8005\u5e7f\u6cdb\u56de\u987e\u4e86GSL\u65b9\u6cd5\u5728\u5b66\u4e60\u9c81\u68d2\u8868\u793a\u4e2d\u7684\u6700\u65b0\u8fdb\u5c55\u3002\u5177\u4f53\u6765\u8bf4\uff0c\u9996\u5148\u5b9a\u4e49\u4e86GSL\u7684\u4e00\u822c\u8303\u5f0f\uff0c\u7136\u540e\u6309\u6a21\u578b\u7ed3\u6784\u5efa\u6a21\u65b9\u5f0f\u5206\u7c7b\u7684\u6700\u65b0\u65b9\u6cd5\u8fdb\u56de\u987e\uff0c\u7136\u540e\u5c06\u5728\u5176\u4ed6\u56fe\u5f62\u4efb\u52a1\u4e2d\u7eb3\u5165GSL\u601d\u60f3\u7684\u5e94\u7528\u7a0b\u5e8f\u8fdb\u884c\u5206\u7c7b\u3002\u6700\u540e\uff0c\u6307\u51fa\u4e86\u5f53\u524d\u7814\u7a76\u4e2d\u7684\u4e00\u4e9b\u95ee\u9898\uff0c\u5e76\u8ba8\u8bba\u4e86\u672a\u6765\u7684\u65b9\u5411\u3002<\/span><\/section>\n

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

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


<\/p>\n


<\/p>\n

\n
\n
\n
\n

\u56fe\u7f51\u7edc\u5165\u95e8\u8def\u5f84<\/strong><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

\n

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

\u4ece\u77e5\u8bc6\u56fe\u8c31\u5230\u6982\u7387\u56fe\u6a21\u578b\uff0c\u4ece\u86cb\u767d\u8d28\u76f8\u4e92\u4f5c\u7528\u7f51\u7edc\u5230\u793e\u4ea4\u7f51\u7edc\uff0c\u4ece\u57fa\u672c\u7684\u903b\u8f91\u7ebf\u8def\u5230\u56e0\u7279\u7f51\uff0c\u56fe\u4e0e\u7f51\u7edc\u65e0\u5904\u4e0d\u5728\u3002\u7531\u4e8e\u4f20\u7edf\u673a\u5668\u5b66\u4e60\u65b9\u6cd5\u96be\u4ee5\u5904\u7406\u56fe\u6570\u636e\u4fe1\u606f\uff0c\u8fd1\u5e74\u6765\u5174\u8d77\u4e86\u56fe\u795e\u7ecf\u7f51\u7edc\u65b9\u6cd5\uff0c\u5728\u8bf8\u591a\u771f\u5b9e\u95ee\u9898\u7684\u5efa\u6a21\u548c\u89e3\u51b3\u4e0a\uff0c\u53d6\u5f97\u826f\u597d\u6548\u679c\u3002<\/span><\/p>\n


<\/p>\n

\u4f5c\u4e3a\u6b63\u706b\u70ed\u7684\u65b0\u9886\u57df\uff0c\u56fe\u7f51\u7edc\u65b9\u6cd5\u8fd8\u6ca1\u6709\u975e\u5e38\u6210\u719f\u7684\u6559\u6750\u6559\u7a0b\uff0c\u6211\u4eec\u6574\u7406\u4e86 \u4e00\u4efd\u7ed3\u5408\u591a\u95e8\u57fa\u7840\u8bfe\u7a0b\u3001\u591a\u7bc7\u7ecf\u5178\u8bba\u6587\u7684\u5b66\u4e60\u8def\u5f84\uff0c\u4f9b\u4f60\u5165\u95e8\u53c2\u8003\u3002<\/span><\/p>\n


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

<\/p>\n

\u626b\u7801\u67e5\u9605\u56fe\u7f51\u7edc\u5165\u95e8\u8def\u5f84<\/span><\/p>\n


<\/section>\n

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