{"id":1065744,"date":"2024-09-23T04:03:03","date_gmt":"2024-09-23T11:03:03","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-event&p=1065744"},"modified":"2024-09-23T20:10:51","modified_gmt":"2024-09-24T03:10:51","slug":"2022-causal-inference-machine-learning-workshop-zh-cn","status":"publish","type":"msr-event","link":"https:\/\/www.microsoft.com\/en-us\/research\/event\/2022-causal-inference-machine-learning-workshop-zh-cn\/","title":{"rendered":"2022\u56e0\u679c\u63a8\u65ad\u4e0e\u673a\u5668\u5b66\u4e60\u7814\u8ba8\u4f1a"},"content":{"rendered":"\n\n\n\n\n

\u56e0\u679c\u63a8\u65ad\u662f\u8fd1\u5e74\u6765\u6570\u636e\u79d1\u5b66\u548c\u4eba\u5de5\u667a\u80fd\u7814\u7a76\u7684\u70ed\u70b9\u4e4b\u4e00\uff0c\u5f97\u5230\u4e86\u5b66\u672f\u754c\u548c\u4e1a\u754c\u7684\u5e7f\u6cdb\u5173\u6ce8\u3002\u672c\u6b21\u7814\u8ba8\u4f1a\u65e8\u5728\u8fdb\u4e00\u6b65\u4fc3\u8fdb\u56e0\u679c\u63a8\u65ad\u9886\u57df\u548c\u673a\u5668\u5b66\u4e60\u9886\u57df\u56fd\u5185\u5b66\u8005\u7684\u5b66\u672f\u4ea4\u6d41\uff0c\u63a2\u8ba8\u56e0\u679c\u63a8\u65ad\u4e0e\u673a\u5668\u5b66\u4e60\u7684\u7ed3\u5408\u65b9\u5f0f\u3002\u672c\u6b21\u7814\u8ba8\u4f1a\u6709\u5e78\u9080\u8bf7\u4e8612\u4f4d\u76f8\u5173\u9886\u57df\u7814\u7a76\u7684\u4e13\u5bb6\u5b66\u8005\u8fdb\u884c\u5b66\u672f\u62a5\u544a\uff0c\u5e7f\u6cdb\u5f00\u5c55\u5b66\u672f\u63a2\u8ba8\uff0c\u4e3a\u76f8\u5173\u9886\u57df\u7684\u7814\u7a76\u4eba\u5458\u63d0\u4f9b\u4e00\u4e2a\u4e13\u4e1a\u7684\u4ea4\u6d41\u5e73\u53f0\u3002\u7814\u8ba8\u4f1a\u5c06\u4e8e2022\u5e744\u67082\u65e5\u5728\u4e2d\u56fd\u79d1\u5b66\u9662\u8ba1\u7b97\u6280\u672f\u7814\u7a76\u6240\u56db\u5c42\u62a5\u544a\u5385\u4e3e\u884c\uff0c\u7814\u8ba8\u4f1a\u4e0d\u6536\u53d6\u4f1a\u8bae\u6ce8\u518c\u8d39\uff0c\u5176\u4ed6\u8d39\u7528\u81ea\u7406\u3002<\/p>\n\n\n\n

2022\u201c\u56e0\u679c\u63a8\u65ad\u4e0e\u673a\u5668\u5b66\u4e60\u201d\u7814\u8ba8\u4f1a\u7ec4\u59d4\u4f1a<\/p>\n\n\n\n

\u4e3b\u5e2d\uff1a\u9a6c\u5fd7\u660e\u3001\u7a0b\u5b66\u65d7\u3001\u5218\u94c1\u5ca9<\/p>\n\n\n\n

\u59d4\u5458\uff1a\u90ed\u5609\u4e30\u3001\u9648\u8587\u3001\u90b9\u957f\u4eae\u3001\u5468\u5ddd\u3001\u5b5f\u742a\u3001\u5f20\u5112\u6e05\u3001\u5b59\u4e3d\u541b<\/p>\n\n\n\n\n\n\n\n

\u9898\u76ee<\/strong>\uff1a Robust and efficient estimation for treatment effect in causal inference<\/p>\n\n\n\n

\u6458\u8981<\/strong>: We develop new methods to evaluate various characteristics for treatment effect in causal inference, including but not limited on average treatment effect, median treatment effect, the Mann-Whitney statistic and etc. The proposed method combines the efficiency of model-based method and robustness of the nonparametric approach, including deep learning methods. It requires few model assumptions and is shown to be efficient if all specifications are correct, and doubly robust if some part is misspecified. Extensive numerical studies have been presented to demonstrate the advantages of the proposed method over others. *Joint work with Ling Zhou, Fanyin Zhou, Qiuxia Wang and Jing Qin<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb\uff1a<\/strong>\u897f\u5357\u8d22\u7ecf\u5927\u5b66\u6559\u6388\uff0c\u7edf\u8ba1\u7814\u7a76\u4e2d\u5fc3\u4e3b\u4efb\u3002\u56fd\u9645\u6570\u7406\u7edf\u8ba1\u5b66\u4f1aIMS-fellow\uff0c\u6559\u80b2\u90e8\u957f\u6c5f\u5b66\u8005\u7279\u8058\u6559\u6388\uff0c\u56fd\u5bb6\u6770\u51fa\u9752\u5e74\u79d1\u5b66\u57fa\u91d1\u83b7\u5f97\u8005\uff0c\u56fd\u5bb6\u767e\u5343\u4e07\u4eba\u624d\u5de5\u7a0b\u83b7\u5f97\u8005\uff0c\u4eab\u53d7\u56fd\u52a1\u9662\u653f\u5e9c\u7279\u6b8a\u6d25\u8d34\u4e13\u5bb6\u3002\u4e3b\u8981\u7814\u7a76\u65b9\u5411\u4e3a\u975e\u53c2\u6570\u65b9\u6cd5\u3001\u8f6c\u6362\u6a21\u578b\u3001\u751f\u5b58\u6570\u636e\u5206\u6790\u3001\u51fd\u6570\u578b\u6570\u636e\u5206\u6790\u3001\u6f5c\u53d8\u91cf\u5206\u6790\u3001\u65f6\u7a7a\u6570\u636e\u5206\u6790\u3002\u7814\u7a76\u6210\u679c\u53d1\u8868\u5728\u5305\u62ec\u56fd\u9645\u7edf\u8ba1\u5b66\u56db\u5927\u9876\u7ea7\u671f\u520aAoS\u3001JASA\u3001JRSSB\u3001Biometrika\u548c\u8ba1\u91cf\u7ecf\u6d4e\u5b66\u9876\u7ea7\u671f\u520aJOE\u53caJBES\u4e0a\u3002\u5148\u540e\u591a\u6b21\u4e3b\u6301\u56fd\u5bb6\u57fa\u91d1\u9879\u76ee\uff0c\u5305\u62ec\u56fd\u5bb6\u6770\u51fa\u9752\u5e74\u57fa\u91d1\u53ca\u81ea\u79d1\u91cd\u70b9\u9879\u76ee\u3002\u6797\u534e\u73cd\u6559\u6388\u662f\u56fd\u9645IMS-China\u3001IBS-CHINA\u53caICSA-China\u59d4\u5458\uff0c\u4e2d\u56fd\u73b0\u573a\u7edf\u8ba1\u7814\u7a76\u4f1a\u6570\u636e\u79d1\u5b66\u4e0e\u4eba\u5de5\u667a\u80fd\u5206\u4f1a\u7406\u4e8b\u957f\uff0c\u7b2c\u4e5d\u5c4a\u5168\u56fd\u5de5\u4e1a\u7edf\u8ba1\u5b66\u6559\u5b66\u7814\u7a76\u4f1a\u526f\u4f1a\u957f\uff0c\u4e2d\u56fd\u73b0\u573a\u7edf\u8ba1\u7814\u7a76\u4f1a\u591a\u4e2a\u5206\u4f1a\u7684\u526f\u7406\u4e8b\u957f\u3002\u5148\u540e\u662f\u56fd\u9645\u7edf\u8ba1\u5b66\u6743\u5a01\u671f\u520a\u300aBiometrics\u300b\u3001\u300aScandinavian Journal of Statistics\u300b\u3001\u300aJournal of Business & Economic Statistics\u300b\u3001\u300aCanadian Journal of Statistics\u300b\u3001 \u300aStatistics and Its Interface\u300b\u3001\u300aStatistical Theory and Related Fields\u300b\u7684Associate Editor\uff0c \u56fd\u5185\u6743\u5a01\u6216\u6838\u5fc3\u5b66\u672f\u671f\u520a\u300a\u6570\u5b66\u5b66\u62a5\u300b\uff08\u82f1\u6587\uff09\u3001\u300a\u5e94\u7528\u6982\u7387\u7edf\u8ba1\u300b\u3001\u300a\u7cfb\u7edf\u79d1\u5b66\u4e0e\u6570\u5b66\u300b\u3001\u300a\u6570\u7406\u7edf\u8ba1\u4e0e\u7ba1\u7406\u300b\u7f16\u59d4\u4f1a\u7f16\u59d4\u3002<\/p>\n\n\n\n\n\n

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

\u9898\u76ee\uff1a<\/strong>\u56e0\u679c\u542f\u53d1\u7684\u7a33\u5b9a\u5b66\u4e60<\/p>\n\n\n\n

\u6458\u8981\uff1a<\/strong>\u8fd1\u5e74\u6765\u4eba\u5de5\u667a\u80fd\u6280\u672f\u7684\u53d1\u5c55\uff0c\u5728\u8bf8\u591a\u5782\u76f4\u9886\u57df\u53d6\u5f97\u4e86\u6027\u80fd\u7a81\u7834\u3002\u4f46\u5f53\u6211\u4eec\u5c06\u8fd9\u4e9b\u6280\u672f\u5e94\u7528\u4e8e\u533b\u7597\u3001\u53f8\u6cd5\u3001\u5de5\u4e1a\u751f\u4ea7\u7b49\u98ce\u9669\u654f\u611f\u9886\u57df\u65f6\uff0c\u53d1\u73b0\u5f53\u524d\u4eba\u5de5\u667a\u80fd\u5728\u7a33\u5b9a\u6027\u3001\u53ef\u89e3\u91ca\u6027\u3001\u516c\u5e73\u6027\u3001\u53ef\u56de\u6eaf\u6027\u7b49\u201c\u56db\u6027\u201d\u65b9\u9762\u5b58\u5728\u4e25\u91cd\u7f3a\u9677\u3002\u7a76\u5176\u6df1\u5c42\u6b21\u539f\u56e0\uff0c\u5f53\u524d\u7edf\u8ba1\u673a\u5668\u5b66\u4e60\u7684\u57fa\u7840\u2014\u2014\u5173\u8054\u7edf\u8ba1\u81ea\u8eab\u4e0d\u7a33\u5b9a\u3001\u4e0d\u53ef\u89e3\u91ca\u3001\u4e0d\u516c\u5e73\u3001\u4e0d\u53ef\u56de\u6eaf\u53ef\u80fd\u662f\u95ee\u9898\u7684\u6839\u6e90\u3002\u76f8\u5bf9\u4e8e\u5173\u8054\u7edf\u8ba1\uff0c\u56e0\u679c\u7edf\u8ba1\u5728\u4fdd\u8bc1\u201c\u56db\u6027\u201d\u65b9\u9762\u5177\u6709\u66f4\u597d\u7684\u7406\u8bba\u57fa\u7840\u3002\u4f46\u5982\u4f55\u5c06\u56e0\u679c\u7edf\u8ba1\u878d\u5165\u673a\u5668\u5b66\u4e60\u6846\u67b6\uff0c\u662f\u4e00\u4e2a\u5f00\u653e\u5e76\u6709\u6311\u6218\u7684\u57fa\u7840\u6027\u95ee\u9898\u3002\u672c\u62a5\u544a\u4e2d\uff0c\u8bb2\u8005\u5c06\u91cd\u70b9\u4ecb\u7ecd\u5c06\u56e0\u679c\u63a8\u7406\u5f15\u5165\u9884\u6d4b\u6027\u95ee\u9898\u6240\u63d0\u51fa\u7684\u7a33\u5b9a\u5b66\u4e60\u7406\u8bba\u548c\u65b9\u6cd5\uff0c\u53ca\u5176\u5728\u89e3\u51b3OOD\u6cdb\u5316\u95ee\u9898\u65b9\u9762\u7684\u673a\u4f1a\u548c\u6311\u6218\u3002<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb\uff1a<\/strong>\u5d14\u9e4f\uff0c\u6e05\u534e\u5927\u5b66\u8ba1\u7b97\u673a\u7cfb\u957f\u8058\u526f\u6559\u6388\uff0c\u535a\u58eb\u751f\u5bfc\u5e08\u3002\u7814\u7a76\u5174\u8da3\u805a\u7126\u4e8e\u5927\u6570\u636e\u9a71\u52a8\u7684\u56e0\u679c\u63a8\u7406\u548c\u7a33\u5b9a\u9884\u6d4b\u3001\u5927\u89c4\u6a21\u7f51\u7edc\u8868\u5f81\u5b66\u4e60\u7b49\u3002\u5728\u6570\u636e\u6316\u6398\u53ca\u4eba\u5de5\u667a\u80fd\u9886\u57df\u9876\u7ea7\u56fd\u9645\u4f1a\u8bae\u53d1\u8868\u8bba\u6587100\u4f59\u7bc7\uff0c\u5148\u540e5\u6b21\u83b7\u5f97\u9876\u7ea7\u56fd\u9645\u4f1a\u8bae\u6216\u671f\u520a\u8bba\u6587\u5956\uff0c\u5e76\u5148\u540e\u4e24\u6b21\u5165\u9009\u6570\u636e\u6316\u6398\u9886\u57df\u9876\u7ea7\u56fd\u9645\u4f1a\u8baeKDD\u6700\u4f73\u8bba\u6587\u4e13\u520a\u3002\u62c5\u4efbIEEE TKDE\u3001ACM TOMM\u3001ACM TIST\u3001IEEE TBD\u7b49\u56fd\u9645\u9876\u7ea7\u671f\u520a\u7f16\u59d4\u3002\u66fe\u83b7\u5f97\u56fd\u5bb6\u81ea\u7136\u79d1\u5b66\u4e8c\u7b49\u5956\u3001\u6559\u80b2\u90e8\u81ea\u7136\u79d1\u5b66\u4e00\u7b49\u5956\u3001CCF-IEEE CS\u9752\u5e74\u79d1\u5b66\u5bb6\u5956\u3001ACM\u6770\u51fa\u79d1\u5b66\u5bb6\u3002<\/p>\n\n\n\n\n\n

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

\u9898\u76ee<\/strong>\uff1aDeconfounding with the Blessing of Dimensionality<\/p>\n\n\n\n

\u6458\u8981<\/strong>\uff1aIn this talk, I will give a selective overview of deconfounding methods that exploit the blessing of dimensionality in regression and graphical models with latent variables. For regression problems, I will discuss factor adjustment and spectral deconfounding methods. For graphical models, I will emphasize the low-rank plus sparse matrix decomposition approach. Finally, I will present two recent examples in compositional data analysis where surprisingly simple methods work extremely well.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb<\/strong>\uff1a\u6797\u4f1f\uff0c\u73b0\u4efb\u5317\u4eac\u5927\u5b66\u6570\u5b66\u79d1\u5b66\u5b66\u9662\u6982\u7387\u7edf\u8ba1\u7cfb\u3001\u7edf\u8ba1\u79d1\u5b66\u4e2d\u5fc3\u957f\u8058\u526f\u6559\u6388\u30022011\u5e74\u83b7\u5357\u52a0\u5dde\u5927\u5b66\u5e94\u7528\u6570\u5b66\u535a\u58eb\u5b66\u4f4d\uff0c2011\u81f32014\u5e74\u5728\u5bbe\u5915\u6cd5\u5c3c\u4e9a\u5927\u5b66\u505a\u535a\u58eb\u540e\u7814\u7a76\uff0c2014\u5e74\u52a0\u5165\u5317\u4eac\u5927\u5b66\u3002\u4e3b\u8981\u4ece\u4e8b\u9ad8\u7ef4\u7edf\u8ba1\u548c\u7edf\u8ba1\u673a\u5668\u5b66\u4e60\u7684\u7406\u8bba\u4e0e\u5e94\u7528\u7814\u7a76\uff0c\u4ee3\u8868\u6027\u6210\u679c\u53d1\u8868\u5728JASA\u3001Biometrika\u3001Biometrics\u3001IEEE TIT\u3001Operations Research\u3001Environmental Science & Technology\u3001\u300a\u4e2d\u56fd\u79d1\u5b66\uff1a\u6570\u5b66\u300b\u7b49\u7edf\u8ba1\u5b66\u53ca\u76f8\u5173\u9886\u57df\u9876\u7ea7\u671f\u520a\u4e0a\u30022015\u5e74\u5165\u9009\u56fd\u5bb6\u9ad8\u5c42\u6b21\u4eba\u624d\u8ba1\u5212\u9752\u5e74\u9879\u76ee\uff0c\u4e3b\u6301\u56fd\u5bb6\u91cd\u70b9\u7814\u53d1\u8ba1\u5212\u8bfe\u9898\u3001\u5317\u4eac\u5e02\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u91cd\u70b9\u7814\u7a76\u4e13\u9898\u9879\u76ee\u3001\u56fd\u5bb6\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u9762\u4e0a\u9879\u76ee\u7b49\u3002<\/p>\n\n\n\n\n\n

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

\u9898\u76ee<\/strong>\uff1a\u56e0\u679c\u63a8\u65ad\uff0c\u89c2\u5bdf\u6027\u7814\u7a76\u548c\u8bfa\u8d1d\u5c14\u7ecf\u6d4e\u5b66\u5956<\/p>\n\n\n\n

\u6458\u8981<\/strong>\uff1a\u8bfa\u8d1d\u5c14\u7ecf\u6d4e\u5b66\u59562021\u5e74\u6388\u4e88Card, Angrist, \u548cImbens\uff0c\u4ee5\u8868\u5f70\u4ed6\u4eec\u5728\u7ecf\u6d4e\u5b66\u7684\u5b9e\u8bc1\u7814\u7a76\u548c\u56e0\u679c\u63a8\u65ad\u65b9\u6cd5\u65b9\u9762\u7684\u8d21\u732e\u3002\u4e09\u4f4d\u7ecf\u6d4e\u5b66\u5bb6\u83b7\u5956\u7684\u79d1\u5b66\u80cc\u666f\u662f\u89c2\u5bdf\u6027\u6570\u636e\u7684\u56e0\u679c\u63a8\u65ad\u3002\u89c2\u5bdf\u6027\u7814\u7a76\u6709\u4e24\u5927\u6311\u6218\uff0c\u4e00\u662f\u6df7\u6742\u56e0\u7d20\uff0c\u4e8c\u662f\u9009\u62e9\u504f\u5dee\u3002Card\uff0cAngrist\u548cImbens\u83b7\u5f97\u8bfa\u8d1d\u5c14\u7ecf\u6d4e\u5b66\u5956\u7684\u4e3b\u8981\u6210\u679c\u662f\u4f7f\u7528\u6070\u5f53\u7684\u81ea\u7136\u8bd5\u9a8c\u3001\u5de5\u5177\u53d8\u91cf\u89e3\u51b3\u52b3\u52a8\u7ecf\u6d4e\u5b66\u4e2d\u7684\u51e0\u4e2a\u767e\u5e74\u96be\u9898\u3002\u5728\u6b64\u4e4b\u524d\uff0c1989\u5e74Haavelmo\u548c2000\u5e74Heckman\u83b7\u8bfa\u8d1d\u5c14\u5956\u7684\u4e3b\u8981\u8d21\u732e\u90fd\u4e0e\u56e0\u679c\u63a8\u65ad\u3001\u9009\u62e9\u504f\u5dee\u5bc6\u5207\u76f8\u5173\u3002\u6211\u5c06\u7ed3\u5408\u8fd9\u51e0\u5c4a\u8bfa\u8d1d\u5c14\u7ecf\u6d4e\u5b66\u5956\u7684\u79d1\u5b66\u80cc\u666f\uff0c\u4ecb\u7ecd\u56e0\u679c\u63a8\u65ad\u548c\u7f3a\u5931\u6570\u636e\u9886\u57df\u7684\u4e00\u4e9b\u8fdb\u5c55\uff0c\u5305\u62ec\u6211\u4eec\u63d0\u51fa\u7684\u4ee3\u7406\u63a8\u65ad\u4ee5\u53ca\u5728\u975e\u968f\u673a\u7f3a\u5931\u6570\u636e\u548c\u56de\u8c03\u6570\u636e\u5206\u6790\u65b9\u9762\u7684\u5de5\u4f5c\u3002<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb<\/strong>\uff1a\u82d7\u65fa\u73b0\u4e3a\u5317\u4eac\u5927\u5b66\u6982\u7387\u7edf\u8ba1\u7cfb\u52a9\u7406\u6559\u6388\u3001\u7814\u7a76\u5458\uff0c2018–2020\u66fe\u5728\u5317\u5927\u5149\u534e\u7ba1\u7406\u5b66\u9662\u4efb\u52a9\u7406\u6559\u6388\uff0c2008–2017 \u5e74\u5728\u5317\u4eac\u5927\u5b66\u8bfb\u672c\u79d1\u548c\u535a\u58eb\uff0c2017–2018 \u5728\u54c8\u4f5b\u5927\u5b66\u751f\u7269\u7edf\u8ba1\u7cfb\u505a\u535a\u58eb\u540e\u7814\u7a76\u3002\u7814\u7a76\u5174\u8da3\u5305\u62ec\u56e0\u679c\u63a8\u65ad\uff0c\u4eba\u5de5\u667a\u80fd\uff0c\u7f3a\u5931\u6570\u636e\u5206\u6790\u548c\u534a\u53c2\u6570\u7edf\u8ba1\uff0c\u4ee5\u53ca\u5728\u751f\u7269\u533b\u836f\uff0c\u6d41\u884c\u75c5\u5b66\u548c\u793e\u4f1a\u7ecf\u6d4e\u4e2d\u7684\u5e94\u7528\u3002\u4e2a\u4eba\u7f51\u9875https:\/\/www.math.pku.edu.cn\/teachers\/mwfy\/<\/p>\n\n\n\n\n\n

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

\u9898\u76ee\uff1a<\/strong>Supervised Causal Learning: A New Frontier of Causal Discovery<\/p>\n\n\n\n

\u6458\u8981\uff1a<\/strong>Supervised Causal Learning (SCL) aims to learn causal relations from observational data by accessing previously seen datasets associated with ground truth causal relations. The supervision-based method enjoys the benefit of \u201cfree\u201d acquisition of training data: with forward sampling techniques, we can generate additional datasets from as many synthetic causal graphs as needed. In this talk, I will first discuss a fundamental question of SCL: What are the benefits from supervision and how does it benefit? And its relationship with causal structure identifiability. Then a general two-phase learning paradigm for SCL is advocated. Following this paradigm, I will share one SCL approach which works remarkably well on discrete data. It shows promising results on validating the effectiveness of supervision for causal learning.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb\uff1a<\/strong>Rui Ding is a principal researcher in the DKI (Data, Knowledge, Intelligence) area at Microsoft Research Asia. His main focus is on robust and automated data analytics. Rui Ding is leading the research on AutoInsights. AutoInsights aims at automatically discovering interesting and meaningful data patterns, which are further equipped with techniques from causality and XAI domains to provide deeper insights, explanation and actions to bridge the gap towards automatic and robust data analytics. Key technologies have been\/are being shipped to Microsoft Power BI and Microsoft Office (Excel, Forms).<\/p>\n\n\n\n\n\n

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

\u9898\u76ee<\/strong>\uff1a Generalized regression estimators for average treatment effect with multicollinearity in high-dimensional covariates<\/p>\n\n\n\n

\u6458\u8981<\/strong>\uff1aTo obtain efficient propensity score based estimators for average treatment effect (ATE) with many covariates having multicollinearity, it’s difficult to  formally automate this variable selection rule in one statistical framework. In order to improve efficiency and solve multicollinearity issue, a two-stage estimation procedure is proposed in this paper. In the first stage, we adjust the usual Horvitz-Thompson estimator of the ATE by incorporating IVs to avoid model misspecification and then propose the generalized regression estimator by utilizing the auxiliary information from covariates related to the potential outcomes. In the second stage, we adapt the Elastic-net method to solve the multicollinearity issue and further improve the estimation efficiency based on the selected important covariates. The finite-sample performance of the proposed estimator is studied through simulation, and an application to employees’ weekly wages is also presented.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb<\/strong>\uff1a\u738b\u78ca\uff0c\u5357\u5f00\u5927\u5b66\u7edf\u8ba1\u4e0e\u6570\u636e\u79d1\u5b66\u5b66\u9662\u526f\u7814\u7a76\u5458\uff0c\u535a\u58eb\u751f\u5bfc\u5e08\u3002\u7814\u7a76\u65b9\u5411\u662f\u590d\u6742\u6570\u636e\u5206\u6790\u548c\u7edf\u8ba1\u5b66\u4e60\uff0c\u5df2\u5728Biometrika\u3001Bernoulli\u3001Statistica Sinica\u7b49\u7edf\u8ba1\u5b66\u6742\u5fd7\u53d1\u8868\u5b66\u672f\u8bba\u658730\u591a\u7bc7\uff0c\u4e3b\u6301\u56fd\u5bb6\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u9762\u4e0a\u9879\u76ee\u3001\u9752\u5e74\u9879\u76ee\u53ca\u5929\u6d25\u5e02\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u5404\u4e00\u9879\u3002<\/p>\n\n\n\n\n\n

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

\u9898\u76ee\uff1a<\/strong>Combinatorial Causal Bandit<\/p>\n\n\n\n

\u6458\u8981\uff1a<\/strong>Combinatorial causal bandit (CCB) is the integration of causal inference, multi-armed bandit and combinatorial optimization techniques. CCB is the following online learning task: A learning agent is given a causal graph with unknown distributions on how each variable X is causally influenced by its parents. The learning task is carried out in T rounds. In each round, the agent selects at most K variables to intervene, the output of a target variable Y is its reward, and the output of all observed variables is the feedback to the agent. The agent aims at learn from the past feedback to help selecting intervention variables in subsequent rounds, so that in the end the cumulative reward of all rounds is maximized. The performance metric is (cumulative) regret, which is defined as the difference between the cumulative reward of always selecting the optimal set of intervention variables and selecting intervention variables according to the learning algorithm. We propose the CCB framework, and study CCB for generalized linear causal models and design learning algorithms with near optimal regret bounds.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb\uff1a<\/strong>\u9648\u536b\u662f\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u9ad8\u7ea7\u7814\u7a76\u5458\uff0c\u4e5f\u5728\u6e05\u534e\u5927\u5b66\u3001\u4e0a\u6d77\u4ea4\u901a\u5927\u5b66\u3001\u4e2d\u79d1\u9662\u7b49\u591a\u6240\u9662\u6821\u548c\u7814\u7a76\u673a\u6784\u62c5\u4efb\u5ba2\u5ea7\u6559\u6388\u6216\u7814\u7a76\u5458\u3002\u4ed6\u662f\u4e2d\u56fd\u8ba1\u7b97\u673a\u5b66\u4f1a\u7406\u8bba\u8ba1\u7b97\u673a\u4e13\u4e1a\u59d4\u5458\u4f1a\u7684\u5e38\u52a1\u59d4\u5458\uff0c\u4e5f\u662f\u5168\u56fd\u5927\u6570\u636e\u4e13\u5bb6\u59d4\u5458\u4f1a\u7684\u59d4\u5458\u3002\u4ed6\u662f\u56fd\u9645\u7535\u6c14\u548c\u7535\u5b50\u5de5\u7a0b\u5e08\u5b66\u4f1a\u7684\u4f1a\u58eb\uff08IEEE Fellow\uff09\u3002\u9648\u536b\u4e3b\u8981\u7684\u7814\u7a76\u65b9\u5411\u5305\u62ec\u793e\u4ea4\u548c\u4fe1\u606f\u7f51\u7edc\uff0c\u5728\u7ebf\u5b66\u4e60\uff0c\u7f51\u7edc\u535a\u5f08\u8bba\u548c\u7ecf\u6d4e\u5b66\uff0c\u5206\u5e03\u5f0f\u8ba1\u7b97\uff0c\u5bb9\u9519\u7b49\u3002\u4ed6\u57282013\u5e74\u4e0e\u4eba\u5408\u8457\u4e00\u672c\u82f1\u6587\u4e13\u8457\uff0c\u57282020\u5e74\u72ec\u7acb\u64b0\u5199\u4e00\u672c\u4e2d\u6587\u4e13\u8457\u3002\u4ed6\u5728\u591a\u4e2a\u5b66\u672f\u671f\u520a\u62c5\u4efb\u7f16\u59d4\uff0c\u4e5f\u5728\u591a\u4e2a\u5b66\u672f\u4f1a\u8bae\u4e2d\u62c5\u4efb\u8fc7\u6280\u672f\u59d4\u5458\u4f1a\u4e3b\u5e2d\u548c\u59d4\u5458\u3002\u9648\u536b\u4e8e\u6e05\u534e\u5927\u5b66\u83b7\u5f97\u672c\u79d1\u548c\u7855\u58eb\u6bd5\u4e1a\uff0c\u4e8e\u7f8e\u56fd\u5eb7\u5948\u5c14\u5927\u5b66\u83b7\u5f97\u535a\u58eb\u5b66\u4f4d\u3002\u6709\u5173\u9648\u536b\u66f4\u591a\u7684\u4fe1\u606f\uff0c\u6b22\u8fce\u8bbf\u95ee\u4ed6\u7684\u4e3b\u9875\uff1ahttp:\/\/research.microsoft.com\/en-us\/people\/weic\/.<\/p>\n\n\n\n\n\n

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

\u9898\u76ee<\/strong>\uff1aConfederated learning and Inference<\/p>\n\n\n\n

\u6458\u8981<\/strong>: The theory of statistical learning and inference for large-scale\/high-dimensional data analysis has recently attracted considerable interest. The central analytic task in the development of confederated statistical learning and inference pertains to the method of integrating results yielded from multiple\/sequential data batches. This talk introduced an one-step meta method based on confidence inference functions, a communication efficient method without pooling individual datasets for unbalanced datasets, and an incremental learning algorithm for streaming datasets with correlated outcomes. Integrative causal inference of multiple similar clinical studies conducted at different sites are also investigated.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb<\/strong>\uff1a\u5468\u5cad\uff0c2004-2010\u5e74\u56db\u5ddd\u5927\u5b66\u6570\u5b66\u5b66\u9662\u672c\u79d1\u548c\u7855\u58eb\uff0c2014\u5e74\u897f\u5357\u8d22\u7ecf\u5927\u5b66\u535a\u58eb\uff0c2018\u5e74\u7f8e\u56fd\u5bc6\u897f\u6839\u5927\u5b66\u751f\u7269\u7edf\u8ba1\u7cfb\u535a\u58eb\u540e\uff0c2017\u5e74\u949f\u5bb6\u5e86\u6570\u5b66\u5956\u83b7\u5f97\u8005\uff0c\u5468\u5cad\u4e0e\u5408\u4f5c\u8005\u5728\u6570\u636e\u96c6\u6210\u3001\u9009\u62e9\u540e\u63a8\u65ad\u3001\u4e9a\u7ec4\u5206\u6790\u3001\u975e\u53c2\u6570\u7406\u8bba\u4e0e\u65b9\u6cd5\u3001\u56e0\u679c\u63a8\u65ad\u7b49\u9886\u57df\u53d6\u5f97\u4e86\u4e00\u7cfb\u5217\u7814\u7a76\u6210\u679c\uff0c\u5728Journal of the American Statistical Association (JASA), Journal of Economics (JoE), Journal of Machine Learning Research(JMLR), Annal of Applied Statistics(AOAS), Biometrics\u7b49\u56fd\u9645\u7edf\u8ba1\u5b66\u3001\u8ba1\u91cf\u7ecf\u6d4e\u5b66\u3001\u8ba1\u7b97\u673a\u9886\u57df\u671f\u520a\u4e0a\u53d1\u8868\u8bba\u658720\u4f59\u7bc7\u3002<\/p>\n\n\n\n\n\n

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

\u9898\u76ee<\/strong>\uff1aNonparametric Estimation of Continuous Treatment Effect with Measurement Error<\/p>\n\n\n\n

\u6458\u8981<\/strong>\uff1aWe consider estimating the average dose-response function (ADRF) nonparametrically for continuous-valued treatment. The existing literature of continuous treatment effect proposed consistent estimators only for error-free data. However, in observational studies concerned by the literature of treatment effect, the treatment data can be measured with error. There, existing techniques are not applicable and finding a proper modification is not straightforward. We identify the ADRF by a weighted conditional expectation and estimate the weights nonparametrically by maximising a local generalised empirical likelihood subject to an expanding set of conditional moment equations incorporated with the deconvolution kernels. We then construct a deconvolution kernel estimator of the weighted conditional expectation. We derive the $L_2$ and $L_\\infty$ convergence rates of our weights estimator and the asymptotic bias and variance of our ADRF estimator. We also provide the asymptotic linear expansion of our ADRF estimator in both the ordinary smooth and the supersmooth error cases, which can help conduct statistical inference. We provide a data-driven method to select our smoothing parameters based on the simulation-extrapolation (SIMEX) idea and propose a new extrapolation procedure to stabilise the computation. Monte-Carlo simulations show a satisfactory finite-sample performance of our method, and a real data study illustrates its practical value.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb<\/strong>\uff1a\u5f20\u653f\uff0c\u4e2d\u56fd\u4eba\u6c11\u5927\u5b66\u7edf\u8ba1\u4e0e\u5927\u6570\u636e\u7814\u7a76\u9662\u52a9\u7406\u6559\u6388\uff0c2015\u5e74\u4e8e\u9999\u6e2f\u4e2d\u6587\u5927\u5b66\u7edf\u8ba1\u7cfb\u83b7\u535a\u58eb\u5b66\u4f4d\u3002\u7814\u7a76\u65b9\u5411\u5305\u62ec\u56e0\u679c\u63a8\u65ad\u3001\u7f3a\u5931\u6570\u636e\u3001\u6c61\u67d3\u6570\u636e\u3001\u534a\u53c2\u6570\u6a21\u578b\u7684\u6709\u6548\u4f30\u8ba1\u3001\u975e\u53c2\u6570\u7edf\u8ba1\u63a8\u65ad\u3001\u968f\u673a\u5fae\u5206\u65b9\u7a0b\u3001\u968f\u673a\u5206\u6790\u7b49\u3002\u5728JRSS-B, JOE, Quantitative Economics, JBES, Statistica Sinica, Stochastic Processes and their Applications\u7b49\u7edf\u8ba1\u3001\u8ba1\u91cf\u7ecf\u6d4e\u3001\u6982\u7387\u8bba\u56fd\u9645\u671f\u520a\u4e0a\u53d1\u8868\u8bba\u6587\u5341\u4f59\u7bc7\u3002\u4e3b\u6301\u56fd\u5bb6\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u9752\u5e74\u57fa\u91d1\uff0c\u5317\u4eac\u5e02\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u9762\u4e0a\u9879\u76ee\u3002<\/p>\n\n\n\n\n\n

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

\u9898\u76ee<\/strong>: A novel stable higher-order influence function estimators for doubly-robust functionals<\/p>\n\n\n\n

\u6458\u8981<\/strong>: In this talk, we will first review the concept of higher-order influence functions (HOIFs) and HOIF-based estimators for a class of smooth statistical functionals commonly encountered in causality. We will demonstrate applications of HOIF-based estimators when deep-learning is being deployed in applied data analysis, including applied causal inference tasks, despite our theoretical understanding about deep learning being still quite limited. Motivated from some of our empirical experience, we recently developed a new class of HOIF-based estimators, which, somewhat surprisingly, enjoy both the nice theoretical properties of the original HOIF-based estimators proposed in 2008 by Robins et al. and possibly more importantly, the numerical stability in finite-sample settings. This new class of HOIF-based estimators (1) paves the way towards making HOIFs practically useful, (2) bridges the HOIF estimators developed in Robins et al. 2008 and the cross-fitting estimators in Newey and Robins 2018, and (3) coincides the HOIF estimator in the fixed-design setting. We envision that this new class of HOIF estimators will be useful in applied works, at least before the myth of deep learning is completely resolved. This talk is based on two working papers, one empirical paper with Kerollos N. Wanis and Jamie Robins and one theoretical paper with Chang Li (a senior undergraduate student at Shanghai Jiao Tong University).<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb<\/strong>\uff1aLin Liu is an Assistant Professor in the Institute of Natural Sciences, School of Mathematical Sciences, and SJTU-Yale Joint Center for Biostatistics and Data Science at Shanghai Jiao Tong University, and PI in the Shanghai AI lab. He completed his PhD in biostatistics at Harvard University, under the supervision of Franziska Michor & Jamie Robins. His research interests lie in the intersection between mathematical statistics, causal inference, machine learning, and applied statistics in biomedical sciences.<\/p>\n\n\n\n\n\n

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

\u9898\u76ee<\/strong>: Estimation and inference for high-dimensional nonparametric additive instrumental-variables regression<\/p>\n\n\n\n

\u6458\u8981<\/strong>: The method of instrumental variables provides a fundamental and practical tool for causal inference in the presence of unmeasured confounding between the treatments and the outcome in various empirical studies. Modern data such as the genetical genomics data in these studies can be high-dimensional. The<\/p>\n\n\n\n

high-dimensional linear instrumental-variables model has been considered in the literature due to its simplicity albeit the true relationship may be nonlinear. We propose a more data-driven approach by considering nonparametric additive models between the instrumental variables and the treatments while keeping the linear model assumption between the treatments and the outcome so that the coefficients therein can directly bear causal interpretation. We provide a two-stage framework for estimation and inference under this more general setup. The group lasso regularization is first employed to select optimal instruments for the high-dimensional nonparametric additive model, and then the outcome variable is regressed on the fitted values from the nonparametric additive model to identify and estimate important treatment effects. We provide non-asymptotic analysis for the estimation error of the proposed estimator. A debiased procedure is further employed to establish valid inference. Extensive numerical experiments show that our proposed method can rival or outperform existing approaches in the literature. We finally analyze the mouse obesity data with the proposed method and discuss new discoveries.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb<\/strong>\uff1a\u4e2d\u56fd\u4eba\u6c11\u5927\u5b66\u7edf\u8ba1\u5b66\u9662\uff0c\u751f\u7269\u7edf\u8ba1\u4e0e\u6d41\u884c\u75c5\u5b66\u7cfb\u8bb2\u5e08\uff0c\u5317\u4eac\u5927\u5b66\u6570\u5b66\u79d1\u5b66\u5b66\u9662\u535a\u58eb\u3002\u4e3b\u8981\u7814\u7a76\u9886\u57df\u4e3a\u56e0\u679c\u63a8\u65ad\u3001\u7f3a\u5931\u6570\u636e\u3001\u9ad8\u7ef4\u7edf\u8ba1\u7b49\u3002\u76ee\u524d\u5df2\u5728\u5305\u62ecBiometrika, Journal of Econometrics, Biometrics\u7b49\u56fd\u9645\u8457\u540d\u7edf\u8ba1\u671f\u520a\u4e0a\u53d1\u8868\u591a\u7bc7\u5b66\u672f\u8bba\u6587\u3002\u4e3b\u6301\u4e00\u9879\u56fd\u5bb6\u81ea\u7136\u79d1\u5b66\u9752\u5e74\u57fa\u91d1\u9879\u76ee\uff0c\u53c2\u4e0e\u5b8c\u6210\u591a\u9879\u56fd\u5bb6\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u9762\u4e0a\u9879\u76ee\u3002<\/p>\n\n\n\n\n\n

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

\u9898\u76ee\uff1a<\/strong>Improving out-of-Distribution Performance of Machine Learning Models from a Causal Perspective<\/p>\n\n\n\n

\u6458\u8981\uff1a<\/strong>Given the remarkable performance of modern machine learning models on various benchmarking datasets, people turn to the next challenges in their wider applications. Among these, out-of-distribution (OOD) generalization is a critical one, since in many real-world tasks, the deploying environment is different from the training one, causing a change in data distribution. Causality provides an insightful approach to analyze and handle the problem. It proposes the model should learn causal relations which represents the fundamental rule governing the data in all environments, in contrast to superficial relations that may only appear in a specific environment accidentally. In this talk, we introduce a model and its variants for prediction\/classification tasks which is designed following a causal reasoning process. We show their causal identification guarantees and OOD generalization analysis, and also the improved empirical performance.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb\uff1a<\/strong>\u5218\u7545\uff0c\u73b0\u4e3a\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u673a\u5668\u5b66\u4e60\u7ec4\u4e3b\u7ba1\u7814\u7a76\u5458\uff0c2019\u5e74\u4e8e\u6e05\u534e\u5927\u5b66\u8ba1\u7b97\u673a\u7cfb\u53d6\u5f97\u535a\u58eb\u5b66\u4f4d\u3002\u4e3b\u8981\u7814\u7a76\u65b9\u5411\u5305\u62ec\u8d1d\u53f6\u65af\u63a8\u65ad\u65b9\u6cd5\uff0c\u56e0\u679c\u6a21\u578b\uff0c\u751f\u6210\u5f0f\u6a21\u578b\u53ca\u5176\u4e0e\u7269\u7406\u5b66\u95ee\u9898\u7684\u7ed3\u5408\u3002<\/p>\n\n\n\n\n\n

<\/p>\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
\u65e5\u671f<\/th>\n\u65f6\u95f4<\/th>\n\u62a5\u544a\u4eba<\/th>\n\u62a5\u544a\u9898\u76ee<\/th>\n\u4e3b\u6301\u4eba<\/th>\n<\/tr>\n<\/thead>\n
4\u67081\u65e5<\/td>\n14:00-22:00<\/td>\n\u62a5\u5230\u6ce8\u518c<\/td>\n<\/tr>\n
4\u67082\u65e5<\/td>\n08:50-09:10<\/td>\n\u5f00\u5e55\u5f0f\u3001\u81f4\u8f9e<\/td>\n\u90ed\u5609\u4e30<\/td>\n<\/tr>\n
09:10-09:40<\/td>\n\u6797\u534e\u73cd\uff08\u897f\u5357\u8d22\u7ecf\u5927\u5b66\uff09<\/td>\nRobust and efficient estimation for treatment effect in causal inference<\/td>\n\u9648\u8587<\/td>\n<\/tr>\n
09:40-10:10<\/td>\n\u5d14\u9e4f\uff08\u6e05\u534e\u5927\u5b66\uff09<\/td>\n\u56e0\u679c\u542f\u53d1\u7684\u7a33\u5b9a\u5b66\u4e60<\/td>\n<\/tr>\n
10:10-10:40<\/td>\n\u6797\u4f1f\uff08\u5317\u4eac\u5927\u5b66\uff09<\/td>\nDeconfounding with the Blessing of Dimensionality<\/td>\n<\/tr>\n
10:40-11:00<\/td>\n\u8336\u6b47<\/td>\n<\/tr>\n
11:00-11:30<\/td>\n\u82d7\u65fa\uff08\u5317\u4eac\u5927\u5b66\uff09<\/td>\n\u56e0\u679c\u63a8\u65ad\uff0c\u89c2\u5bdf\u6027\u7814\u7a76\u548c\u8bfa\u8d1d\u5c14\u7ecf\u6d4e\u5b66\u5956<\/td>\n\u90b9\u957f\u4eae<\/td>\n<\/tr>\n
11:30-12:00<\/td>\n\u4e01\u9510\uff08\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\uff09<\/td>\nSupervised Causal Learning: A New Frontier of Causal Discovery<\/td>\n<\/tr>\n
12:00-12:30<\/td>\n\u738b\u78ca\uff08\u5357\u5f00\u5927\u5b66\uff09<\/td>\nGeneralized regression estimators for average treatment effect with multicollinearity in high-dimensional covariates<\/td>\n<\/tr>\n
12:30-14:00<\/td>\n\u5348\u9910<\/td>\n<\/tr>\n
14:00-14:30<\/td>\n\u9648\u536b\uff08\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\uff09<\/td>\nCombinatorial Causal Bandit<\/td>\n\u82d7\u65fa<\/td>\n<\/tr>\n
14:30-15:00<\/td>\n\u5468\u5cad\uff08\u897f\u5357\u8d22\u7ecf\u5927\u5b66\uff09<\/td>\nConfederated learning and Inference<\/td>\n<\/tr>\n
15:00-15:30<\/td>\n\u5f20\u653f\uff08\u4e2d\u56fd\u4eba\u6c11\u5927\u5b66\uff09<\/td>\nNonparametric Estimation of Continuous Treatment Effect with Measurement Error<\/td>\n<\/tr>\n
15:30-15:50<\/td>\n\u8336\u6b47<\/td>\n<\/tr>\n
15:50-16:20<\/td>\n\u5218\u6797\uff08\u4e0a\u6d77\u4ea4\u901a\u5927\u5b66\uff09<\/td>\nA novel stable higher-order influence function estimators for doubly-robust functionals<\/td>\n\u9648\u536b<\/td>\n<\/tr>\n
16:20-16:50<\/td>\n\u674e\u4f1f\uff08\u4e2d\u56fd\u4eba\u6c11\u5927\u5b66\uff09<\/td>\nEstimation and inference for high-dimensional nonparametric additive instrumental-variables regression<\/td>\n<\/tr>\n
16:50-17:20<\/td>\n\u5218\u7545\uff08\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\uff09<\/td>\nImproving out-of-Distribution Performance of Machine Learning Models from a Causal Perspective<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n\n\n\n
<\/div>\n\n\n","protected":false},"excerpt":{"rendered":"

\u56e0\u679c\u63a8\u65ad\u662f\u8fd1\u5e74\u6765\u6570\u636e\u79d1\u5b66\u548c\u4eba\u5de5\u667a\u80fd\u7814\u7a76\u7684\u70ed\u70b9\u4e4b\u4e00\uff0c\u5f97\u5230\u4e86\u5b66\u672f\u754c\u548c\u4e1a\u754c\u7684\u5e7f\u6cdb\u5173\u6ce8\u3002\u672c\u6b21\u7814\u8ba8\u4f1a\u65e8\u5728\u8fdb\u4e00\u6b65\u4fc3\u8fdb\u56e0\u679c\u63a8\u65ad\u9886\u57df\u548c\u673a\u5668\u5b66\u4e60\u9886\u57df\u56fd\u5185\u5b66\u8005\u7684\u5b66\u672f\u4ea4\u6d41\uff0c\u63a2\u8ba8\u56e0\u679c\u63a8\u65ad\u4e0e\u673a\u5668\u5b66\u4e60\u7684\u7ed3\u5408\u65b9\u5f0f\u3002\u672c\u6b21\u7814\u8ba8\u4f1a\u6709\u5e78\u9080\u8bf7\u4e8612\u4f4d\u76f8\u5173\u9886\u57df\u7814\u7a76\u7684\u4e13\u5bb6\u5b66\u8005\u8fdb\u884c\u5b66\u672f\u62a5\u544a\uff0c\u5e7f\u6cdb\u5f00\u5c55\u5b66\u672f\u63a2\u8ba8\uff0c\u4e3a\u76f8\u5173\u9886\u57df\u7684\u7814\u7a76\u4eba\u5458\u63d0\u4f9b\u4e00\u4e2a\u4e13\u4e1a\u7684\u4ea4\u6d41\u5e73\u53f0\u3002\u7814\u8ba8\u4f1a\u5c06\u4e8e2022\u5e744\u67082\u65e5\u5728\u4e2d\u56fd\u79d1\u5b66\u9662\u8ba1\u7b97\u6280\u672f\u7814\u7a76\u6240\u56db\u5c42\u62a5\u544a\u5385\u4e3e\u884c\uff0c\u7814\u8ba8\u4f1a\u4e0d\u6536\u53d6\u4f1a\u8bae\u6ce8\u518c\u8d39\uff0c\u5176\u4ed6\u8d39\u7528\u81ea\u7406\u3002 2022\u201c\u56e0\u679c\u63a8\u65ad\u4e0e\u673a\u5668\u5b66\u4e60\u201d\u7814\u8ba8\u4f1a\u7ec4\u59d4\u4f1a \u4e3b\u5e2d\uff1a\u9a6c\u5fd7\u660e\u3001\u7a0b\u5b66\u65d7\u3001\u5218\u94c1\u5ca9 \u59d4\u5458\uff1a\u90ed\u5609\u4e30\u3001\u9648\u8587\u3001\u90b9\u957f\u4eae\u3001\u5468\u5ddd\u3001\u5b5f\u742a\u3001\u5f20\u5112\u6e05\u3001\u5b59\u4e3d\u541b \u65e5\u671f \u65f6\u95f4 \u62a5\u544a\u4eba \u62a5\u544a\u9898\u76ee \u4e3b\u6301\u4eba 4\u67081\u65e5 14:00-22:00 \u62a5\u5230\u6ce8\u518c 4\u67082\u65e5 08:50-09:10 \u5f00\u5e55\u5f0f\u3001\u81f4\u8f9e \u90ed\u5609\u4e30 09:10-09:40 \u6797\u534e\u73cd\uff08\u897f\u5357\u8d22\u7ecf\u5927\u5b66\uff09 Robust and efficient estimation for treatment effect in causal inference \u9648\u8587 09:40-10:10 \u5d14\u9e4f\uff08\u6e05\u534e\u5927\u5b66\uff09 \u56e0\u679c\u542f\u53d1\u7684\u7a33\u5b9a\u5b66\u4e60 10:10-10:40 \u6797\u4f1f\uff08\u5317\u4eac\u5927\u5b66\uff09 Deconfounding with the Blessing of Dimensionality 10:40-11:00 \u8336\u6b47 11:00-11:30 \u82d7\u65fa\uff08\u5317\u4eac\u5927\u5b66\uff09 \u56e0\u679c\u63a8\u65ad\uff0c\u89c2\u5bdf\u6027\u7814\u7a76\u548c\u8bfa\u8d1d\u5c14\u7ecf\u6d4e\u5b66\u5956 \u90b9\u957f\u4eae 11:30-12:00 \u4e01\u9510\uff08\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\uff09 Supervised Causal Learning: A New Frontier of […]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr_startdate":"2022-04-02","msr_enddate":"2022-04-02","msr_location":"\u4e2d\u56fd\u79d1\u5b66\u9662\u8ba1\u7b97\u6280\u672f\u7814\u7a76\u6240","msr_expirationdate":"","msr_event_recording_link":"","msr_event_link":"","msr_event_link_redirect":false,"msr_event_time":"","msr_hide_region":false,"msr_private_event":false,"msr_hide_image_in_river":0,"footnotes":""},"research-area":[13556],"msr-region":[],"msr-event-type":[197944],"msr-video-type":[],"msr-locale":[268881],"msr-program-audience":[],"msr-post-option":[],"msr-impact-theme":[],"class_list":["post-1065744","msr-event","type-msr-event","status-publish","hentry","msr-research-area-artificial-intelligence","msr-event-type-hosted-by-microsoft","msr-locale-zh_cn"],"msr_about":"\n\n\n\n\n

\u56e0\u679c\u63a8\u65ad\u662f\u8fd1\u5e74\u6765\u6570\u636e\u79d1\u5b66\u548c\u4eba\u5de5\u667a\u80fd\u7814\u7a76\u7684\u70ed\u70b9\u4e4b\u4e00\uff0c\u5f97\u5230\u4e86\u5b66\u672f\u754c\u548c\u4e1a\u754c\u7684\u5e7f\u6cdb\u5173\u6ce8\u3002\u672c\u6b21\u7814\u8ba8\u4f1a\u65e8\u5728\u8fdb\u4e00\u6b65\u4fc3\u8fdb\u56e0\u679c\u63a8\u65ad\u9886\u57df\u548c\u673a\u5668\u5b66\u4e60\u9886\u57df\u56fd\u5185\u5b66\u8005\u7684\u5b66\u672f\u4ea4\u6d41\uff0c\u63a2\u8ba8\u56e0\u679c\u63a8\u65ad\u4e0e\u673a\u5668\u5b66\u4e60\u7684\u7ed3\u5408\u65b9\u5f0f\u3002\u672c\u6b21\u7814\u8ba8\u4f1a\u6709\u5e78\u9080\u8bf7\u4e8612\u4f4d\u76f8\u5173\u9886\u57df\u7814\u7a76\u7684\u4e13\u5bb6\u5b66\u8005\u8fdb\u884c\u5b66\u672f\u62a5\u544a\uff0c\u5e7f\u6cdb\u5f00\u5c55\u5b66\u672f\u63a2\u8ba8\uff0c\u4e3a\u76f8\u5173\u9886\u57df\u7684\u7814\u7a76\u4eba\u5458\u63d0\u4f9b\u4e00\u4e2a\u4e13\u4e1a\u7684\u4ea4\u6d41\u5e73\u53f0\u3002\u7814\u8ba8\u4f1a\u5c06\u4e8e2022\u5e744\u67082\u65e5\u5728\u4e2d\u56fd\u79d1\u5b66\u9662\u8ba1\u7b97\u6280\u672f\u7814\u7a76\u6240\u56db\u5c42\u62a5\u544a\u5385\u4e3e\u884c\uff0c\u7814\u8ba8\u4f1a\u4e0d\u6536\u53d6\u4f1a\u8bae\u6ce8\u518c\u8d39\uff0c\u5176\u4ed6\u8d39\u7528\u81ea\u7406\u3002<\/p>\n\n\n\n

2022\u201c\u56e0\u679c\u63a8\u65ad\u4e0e\u673a\u5668\u5b66\u4e60\u201d\u7814\u8ba8\u4f1a\u7ec4\u59d4\u4f1a<\/p>\n\n\n\n

\u4e3b\u5e2d\uff1a\u9a6c\u5fd7\u660e\u3001\u7a0b\u5b66\u65d7\u3001\u5218\u94c1\u5ca9<\/p>\n\n\n\n

\u59d4\u5458\uff1a\u90ed\u5609\u4e30\u3001\u9648\u8587\u3001\u90b9\u957f\u4eae\u3001\u5468\u5ddd\u3001\u5b5f\u742a\u3001\u5f20\u5112\u6e05\u3001\u5b59\u4e3d\u541b<\/p>\n\n\n\n\n\n\n\n

\u9898\u76ee<\/strong>\uff1a Robust and efficient estimation for treatment effect in causal inference<\/p>\n\n\n\n

\u6458\u8981<\/strong>: We develop new methods to evaluate various characteristics for treatment effect in causal inference, including but not limited on average treatment effect, median treatment effect, the Mann-Whitney statistic and etc. The proposed method combines the efficiency of model-based method and robustness of the nonparametric approach, including deep learning methods. It requires few model assumptions and is shown to be efficient if all specifications are correct, and doubly robust if some part is misspecified. Extensive numerical studies have been presented to demonstrate the advantages of the proposed method over others. *Joint work with Ling Zhou, Fanyin Zhou, Qiuxia Wang and Jing Qin<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb\uff1a<\/strong>\u897f\u5357\u8d22\u7ecf\u5927\u5b66\u6559\u6388\uff0c\u7edf\u8ba1\u7814\u7a76\u4e2d\u5fc3\u4e3b\u4efb\u3002\u56fd\u9645\u6570\u7406\u7edf\u8ba1\u5b66\u4f1aIMS-fellow\uff0c\u6559\u80b2\u90e8\u957f\u6c5f\u5b66\u8005\u7279\u8058\u6559\u6388\uff0c\u56fd\u5bb6\u6770\u51fa\u9752\u5e74\u79d1\u5b66\u57fa\u91d1\u83b7\u5f97\u8005\uff0c\u56fd\u5bb6\u767e\u5343\u4e07\u4eba\u624d\u5de5\u7a0b\u83b7\u5f97\u8005\uff0c\u4eab\u53d7\u56fd\u52a1\u9662\u653f\u5e9c\u7279\u6b8a\u6d25\u8d34\u4e13\u5bb6\u3002\u4e3b\u8981\u7814\u7a76\u65b9\u5411\u4e3a\u975e\u53c2\u6570\u65b9\u6cd5\u3001\u8f6c\u6362\u6a21\u578b\u3001\u751f\u5b58\u6570\u636e\u5206\u6790\u3001\u51fd\u6570\u578b\u6570\u636e\u5206\u6790\u3001\u6f5c\u53d8\u91cf\u5206\u6790\u3001\u65f6\u7a7a\u6570\u636e\u5206\u6790\u3002\u7814\u7a76\u6210\u679c\u53d1\u8868\u5728\u5305\u62ec\u56fd\u9645\u7edf\u8ba1\u5b66\u56db\u5927\u9876\u7ea7\u671f\u520aAoS\u3001JASA\u3001JRSSB\u3001Biometrika\u548c\u8ba1\u91cf\u7ecf\u6d4e\u5b66\u9876\u7ea7\u671f\u520aJOE\u53caJBES\u4e0a\u3002\u5148\u540e\u591a\u6b21\u4e3b\u6301\u56fd\u5bb6\u57fa\u91d1\u9879\u76ee\uff0c\u5305\u62ec\u56fd\u5bb6\u6770\u51fa\u9752\u5e74\u57fa\u91d1\u53ca\u81ea\u79d1\u91cd\u70b9\u9879\u76ee\u3002\u6797\u534e\u73cd\u6559\u6388\u662f\u56fd\u9645IMS-China\u3001IBS-CHINA\u53caICSA-China\u59d4\u5458\uff0c\u4e2d\u56fd\u73b0\u573a\u7edf\u8ba1\u7814\u7a76\u4f1a\u6570\u636e\u79d1\u5b66\u4e0e\u4eba\u5de5\u667a\u80fd\u5206\u4f1a\u7406\u4e8b\u957f\uff0c\u7b2c\u4e5d\u5c4a\u5168\u56fd\u5de5\u4e1a\u7edf\u8ba1\u5b66\u6559\u5b66\u7814\u7a76\u4f1a\u526f\u4f1a\u957f\uff0c\u4e2d\u56fd\u73b0\u573a\u7edf\u8ba1\u7814\u7a76\u4f1a\u591a\u4e2a\u5206\u4f1a\u7684\u526f\u7406\u4e8b\u957f\u3002\u5148\u540e\u662f\u56fd\u9645\u7edf\u8ba1\u5b66\u6743\u5a01\u671f\u520a\u300aBiometrics\u300b\u3001\u300aScandinavian Journal of Statistics\u300b\u3001\u300aJournal of Business & Economic Statistics\u300b\u3001\u300aCanadian Journal of Statistics\u300b\u3001 \u300aStatistics and Its Interface\u300b\u3001\u300aStatistical Theory and Related Fields\u300b\u7684Associate Editor\uff0c \u56fd\u5185\u6743\u5a01\u6216\u6838\u5fc3\u5b66\u672f\u671f\u520a\u300a\u6570\u5b66\u5b66\u62a5\u300b\uff08\u82f1\u6587\uff09\u3001\u300a\u5e94\u7528\u6982\u7387\u7edf\u8ba1\u300b\u3001\u300a\u7cfb\u7edf\u79d1\u5b66\u4e0e\u6570\u5b66\u300b\u3001\u300a\u6570\u7406\u7edf\u8ba1\u4e0e\u7ba1\u7406\u300b\u7f16\u59d4\u4f1a\u7f16\u59d4\u3002<\/p>\n\n\n\n\n\n

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

\u9898\u76ee\uff1a<\/strong>\u56e0\u679c\u542f\u53d1\u7684\u7a33\u5b9a\u5b66\u4e60<\/p>\n\n\n\n

\u6458\u8981\uff1a<\/strong>\u8fd1\u5e74\u6765\u4eba\u5de5\u667a\u80fd\u6280\u672f\u7684\u53d1\u5c55\uff0c\u5728\u8bf8\u591a\u5782\u76f4\u9886\u57df\u53d6\u5f97\u4e86\u6027\u80fd\u7a81\u7834\u3002\u4f46\u5f53\u6211\u4eec\u5c06\u8fd9\u4e9b\u6280\u672f\u5e94\u7528\u4e8e\u533b\u7597\u3001\u53f8\u6cd5\u3001\u5de5\u4e1a\u751f\u4ea7\u7b49\u98ce\u9669\u654f\u611f\u9886\u57df\u65f6\uff0c\u53d1\u73b0\u5f53\u524d\u4eba\u5de5\u667a\u80fd\u5728\u7a33\u5b9a\u6027\u3001\u53ef\u89e3\u91ca\u6027\u3001\u516c\u5e73\u6027\u3001\u53ef\u56de\u6eaf\u6027\u7b49\u201c\u56db\u6027\u201d\u65b9\u9762\u5b58\u5728\u4e25\u91cd\u7f3a\u9677\u3002\u7a76\u5176\u6df1\u5c42\u6b21\u539f\u56e0\uff0c\u5f53\u524d\u7edf\u8ba1\u673a\u5668\u5b66\u4e60\u7684\u57fa\u7840\u2014\u2014\u5173\u8054\u7edf\u8ba1\u81ea\u8eab\u4e0d\u7a33\u5b9a\u3001\u4e0d\u53ef\u89e3\u91ca\u3001\u4e0d\u516c\u5e73\u3001\u4e0d\u53ef\u56de\u6eaf\u53ef\u80fd\u662f\u95ee\u9898\u7684\u6839\u6e90\u3002\u76f8\u5bf9\u4e8e\u5173\u8054\u7edf\u8ba1\uff0c\u56e0\u679c\u7edf\u8ba1\u5728\u4fdd\u8bc1\u201c\u56db\u6027\u201d\u65b9\u9762\u5177\u6709\u66f4\u597d\u7684\u7406\u8bba\u57fa\u7840\u3002\u4f46\u5982\u4f55\u5c06\u56e0\u679c\u7edf\u8ba1\u878d\u5165\u673a\u5668\u5b66\u4e60\u6846\u67b6\uff0c\u662f\u4e00\u4e2a\u5f00\u653e\u5e76\u6709\u6311\u6218\u7684\u57fa\u7840\u6027\u95ee\u9898\u3002\u672c\u62a5\u544a\u4e2d\uff0c\u8bb2\u8005\u5c06\u91cd\u70b9\u4ecb\u7ecd\u5c06\u56e0\u679c\u63a8\u7406\u5f15\u5165\u9884\u6d4b\u6027\u95ee\u9898\u6240\u63d0\u51fa\u7684\u7a33\u5b9a\u5b66\u4e60\u7406\u8bba\u548c\u65b9\u6cd5\uff0c\u53ca\u5176\u5728\u89e3\u51b3OOD\u6cdb\u5316\u95ee\u9898\u65b9\u9762\u7684\u673a\u4f1a\u548c\u6311\u6218\u3002<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb\uff1a<\/strong>\u5d14\u9e4f\uff0c\u6e05\u534e\u5927\u5b66\u8ba1\u7b97\u673a\u7cfb\u957f\u8058\u526f\u6559\u6388\uff0c\u535a\u58eb\u751f\u5bfc\u5e08\u3002\u7814\u7a76\u5174\u8da3\u805a\u7126\u4e8e\u5927\u6570\u636e\u9a71\u52a8\u7684\u56e0\u679c\u63a8\u7406\u548c\u7a33\u5b9a\u9884\u6d4b\u3001\u5927\u89c4\u6a21\u7f51\u7edc\u8868\u5f81\u5b66\u4e60\u7b49\u3002\u5728\u6570\u636e\u6316\u6398\u53ca\u4eba\u5de5\u667a\u80fd\u9886\u57df\u9876\u7ea7\u56fd\u9645\u4f1a\u8bae\u53d1\u8868\u8bba\u6587100\u4f59\u7bc7\uff0c\u5148\u540e5\u6b21\u83b7\u5f97\u9876\u7ea7\u56fd\u9645\u4f1a\u8bae\u6216\u671f\u520a\u8bba\u6587\u5956\uff0c\u5e76\u5148\u540e\u4e24\u6b21\u5165\u9009\u6570\u636e\u6316\u6398\u9886\u57df\u9876\u7ea7\u56fd\u9645\u4f1a\u8baeKDD\u6700\u4f73\u8bba\u6587\u4e13\u520a\u3002\u62c5\u4efbIEEE TKDE\u3001ACM TOMM\u3001ACM TIST\u3001IEEE TBD\u7b49\u56fd\u9645\u9876\u7ea7\u671f\u520a\u7f16\u59d4\u3002\u66fe\u83b7\u5f97\u56fd\u5bb6\u81ea\u7136\u79d1\u5b66\u4e8c\u7b49\u5956\u3001\u6559\u80b2\u90e8\u81ea\u7136\u79d1\u5b66\u4e00\u7b49\u5956\u3001CCF-IEEE CS\u9752\u5e74\u79d1\u5b66\u5bb6\u5956\u3001ACM\u6770\u51fa\u79d1\u5b66\u5bb6\u3002<\/p>\n\n\n\n\n\n

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

\u9898\u76ee<\/strong>\uff1aDeconfounding with the Blessing of Dimensionality<\/p>\n\n\n\n

\u6458\u8981<\/strong>\uff1aIn this talk, I will give a selective overview of deconfounding methods that exploit the blessing of dimensionality in regression and graphical models with latent variables. For regression problems, I will discuss factor adjustment and spectral deconfounding methods. For graphical models, I will emphasize the low-rank plus sparse matrix decomposition approach. Finally, I will present two recent examples in compositional data analysis where surprisingly simple methods work extremely well.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb<\/strong>\uff1a\u6797\u4f1f\uff0c\u73b0\u4efb\u5317\u4eac\u5927\u5b66\u6570\u5b66\u79d1\u5b66\u5b66\u9662\u6982\u7387\u7edf\u8ba1\u7cfb\u3001\u7edf\u8ba1\u79d1\u5b66\u4e2d\u5fc3\u957f\u8058\u526f\u6559\u6388\u30022011\u5e74\u83b7\u5357\u52a0\u5dde\u5927\u5b66\u5e94\u7528\u6570\u5b66\u535a\u58eb\u5b66\u4f4d\uff0c2011\u81f32014\u5e74\u5728\u5bbe\u5915\u6cd5\u5c3c\u4e9a\u5927\u5b66\u505a\u535a\u58eb\u540e\u7814\u7a76\uff0c2014\u5e74\u52a0\u5165\u5317\u4eac\u5927\u5b66\u3002\u4e3b\u8981\u4ece\u4e8b\u9ad8\u7ef4\u7edf\u8ba1\u548c\u7edf\u8ba1\u673a\u5668\u5b66\u4e60\u7684\u7406\u8bba\u4e0e\u5e94\u7528\u7814\u7a76\uff0c\u4ee3\u8868\u6027\u6210\u679c\u53d1\u8868\u5728JASA\u3001Biometrika\u3001Biometrics\u3001IEEE TIT\u3001Operations Research\u3001Environmental Science & Technology\u3001\u300a\u4e2d\u56fd\u79d1\u5b66\uff1a\u6570\u5b66\u300b\u7b49\u7edf\u8ba1\u5b66\u53ca\u76f8\u5173\u9886\u57df\u9876\u7ea7\u671f\u520a\u4e0a\u30022015\u5e74\u5165\u9009\u56fd\u5bb6\u9ad8\u5c42\u6b21\u4eba\u624d\u8ba1\u5212\u9752\u5e74\u9879\u76ee\uff0c\u4e3b\u6301\u56fd\u5bb6\u91cd\u70b9\u7814\u53d1\u8ba1\u5212\u8bfe\u9898\u3001\u5317\u4eac\u5e02\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u91cd\u70b9\u7814\u7a76\u4e13\u9898\u9879\u76ee\u3001\u56fd\u5bb6\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u9762\u4e0a\u9879\u76ee\u7b49\u3002<\/p>\n\n\n\n\n\n

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

\u9898\u76ee<\/strong>\uff1a\u56e0\u679c\u63a8\u65ad\uff0c\u89c2\u5bdf\u6027\u7814\u7a76\u548c\u8bfa\u8d1d\u5c14\u7ecf\u6d4e\u5b66\u5956<\/p>\n\n\n\n

\u6458\u8981<\/strong>\uff1a\u8bfa\u8d1d\u5c14\u7ecf\u6d4e\u5b66\u59562021\u5e74\u6388\u4e88Card, Angrist, \u548cImbens\uff0c\u4ee5\u8868\u5f70\u4ed6\u4eec\u5728\u7ecf\u6d4e\u5b66\u7684\u5b9e\u8bc1\u7814\u7a76\u548c\u56e0\u679c\u63a8\u65ad\u65b9\u6cd5\u65b9\u9762\u7684\u8d21\u732e\u3002\u4e09\u4f4d\u7ecf\u6d4e\u5b66\u5bb6\u83b7\u5956\u7684\u79d1\u5b66\u80cc\u666f\u662f\u89c2\u5bdf\u6027\u6570\u636e\u7684\u56e0\u679c\u63a8\u65ad\u3002\u89c2\u5bdf\u6027\u7814\u7a76\u6709\u4e24\u5927\u6311\u6218\uff0c\u4e00\u662f\u6df7\u6742\u56e0\u7d20\uff0c\u4e8c\u662f\u9009\u62e9\u504f\u5dee\u3002Card\uff0cAngrist\u548cImbens\u83b7\u5f97\u8bfa\u8d1d\u5c14\u7ecf\u6d4e\u5b66\u5956\u7684\u4e3b\u8981\u6210\u679c\u662f\u4f7f\u7528\u6070\u5f53\u7684\u81ea\u7136\u8bd5\u9a8c\u3001\u5de5\u5177\u53d8\u91cf\u89e3\u51b3\u52b3\u52a8\u7ecf\u6d4e\u5b66\u4e2d\u7684\u51e0\u4e2a\u767e\u5e74\u96be\u9898\u3002\u5728\u6b64\u4e4b\u524d\uff0c1989\u5e74Haavelmo\u548c2000\u5e74Heckman\u83b7\u8bfa\u8d1d\u5c14\u5956\u7684\u4e3b\u8981\u8d21\u732e\u90fd\u4e0e\u56e0\u679c\u63a8\u65ad\u3001\u9009\u62e9\u504f\u5dee\u5bc6\u5207\u76f8\u5173\u3002\u6211\u5c06\u7ed3\u5408\u8fd9\u51e0\u5c4a\u8bfa\u8d1d\u5c14\u7ecf\u6d4e\u5b66\u5956\u7684\u79d1\u5b66\u80cc\u666f\uff0c\u4ecb\u7ecd\u56e0\u679c\u63a8\u65ad\u548c\u7f3a\u5931\u6570\u636e\u9886\u57df\u7684\u4e00\u4e9b\u8fdb\u5c55\uff0c\u5305\u62ec\u6211\u4eec\u63d0\u51fa\u7684\u4ee3\u7406\u63a8\u65ad\u4ee5\u53ca\u5728\u975e\u968f\u673a\u7f3a\u5931\u6570\u636e\u548c\u56de\u8c03\u6570\u636e\u5206\u6790\u65b9\u9762\u7684\u5de5\u4f5c\u3002<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb<\/strong>\uff1a\u82d7\u65fa\u73b0\u4e3a\u5317\u4eac\u5927\u5b66\u6982\u7387\u7edf\u8ba1\u7cfb\u52a9\u7406\u6559\u6388\u3001\u7814\u7a76\u5458\uff0c2018--2020\u66fe\u5728\u5317\u5927\u5149\u534e\u7ba1\u7406\u5b66\u9662\u4efb\u52a9\u7406\u6559\u6388\uff0c2008--2017 \u5e74\u5728\u5317\u4eac\u5927\u5b66\u8bfb\u672c\u79d1\u548c\u535a\u58eb\uff0c2017--2018 \u5728\u54c8\u4f5b\u5927\u5b66\u751f\u7269\u7edf\u8ba1\u7cfb\u505a\u535a\u58eb\u540e\u7814\u7a76\u3002\u7814\u7a76\u5174\u8da3\u5305\u62ec\u56e0\u679c\u63a8\u65ad\uff0c\u4eba\u5de5\u667a\u80fd\uff0c\u7f3a\u5931\u6570\u636e\u5206\u6790\u548c\u534a\u53c2\u6570\u7edf\u8ba1\uff0c\u4ee5\u53ca\u5728\u751f\u7269\u533b\u836f\uff0c\u6d41\u884c\u75c5\u5b66\u548c\u793e\u4f1a\u7ecf\u6d4e\u4e2d\u7684\u5e94\u7528\u3002\u4e2a\u4eba\u7f51\u9875https:\/\/www.math.pku.edu.cn\/teachers\/mwfy\/<\/p>\n\n\n\n\n\n

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

\u9898\u76ee\uff1a<\/strong>Supervised Causal Learning: A New Frontier of Causal Discovery<\/p>\n\n\n\n

\u6458\u8981\uff1a<\/strong>Supervised Causal Learning (SCL) aims to learn causal relations from observational data by accessing previously seen datasets associated with ground truth causal relations. The supervision-based method enjoys the benefit of \u201cfree\u201d acquisition of training data: with forward sampling techniques, we can generate additional datasets from as many synthetic causal graphs as needed. In this talk, I will first discuss a fundamental question of SCL: What are the benefits from supervision and how does it benefit? And its relationship with causal structure identifiability. Then a general two-phase learning paradigm for SCL is advocated. Following this paradigm, I will share one SCL approach which works remarkably well on discrete data. It shows promising results on validating the effectiveness of supervision for causal learning.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb\uff1a<\/strong>Rui Ding is a principal researcher in the DKI (Data, Knowledge, Intelligence) area at Microsoft Research Asia. His main focus is on robust and automated data analytics. Rui Ding is leading the research on AutoInsights. AutoInsights aims at automatically discovering interesting and meaningful data patterns, which are further equipped with techniques from causality and XAI domains to provide deeper insights, explanation and actions to bridge the gap towards automatic and robust data analytics. Key technologies have been\/are being shipped to Microsoft Power BI and Microsoft Office (Excel, Forms).<\/p>\n\n\n\n\n\n

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

\u9898\u76ee<\/strong>\uff1a Generalized regression estimators for average treatment effect with multicollinearity in high-dimensional covariates<\/p>\n\n\n\n

\u6458\u8981<\/strong>\uff1aTo obtain efficient propensity score based estimators for average treatment effect (ATE) with many covariates having multicollinearity, it's difficult to  formally automate this variable selection rule in one statistical framework. In order to improve efficiency and solve multicollinearity issue, a two-stage estimation procedure is proposed in this paper. In the first stage, we adjust the usual Horvitz-Thompson estimator of the ATE by incorporating IVs to avoid model misspecification and then propose the generalized regression estimator by utilizing the auxiliary information from covariates related to the potential outcomes. In the second stage, we adapt the Elastic-net method to solve the multicollinearity issue and further improve the estimation efficiency based on the selected important covariates. The finite-sample performance of the proposed estimator is studied through simulation, and an application to employees' weekly wages is also presented.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb<\/strong>\uff1a\u738b\u78ca\uff0c\u5357\u5f00\u5927\u5b66\u7edf\u8ba1\u4e0e\u6570\u636e\u79d1\u5b66\u5b66\u9662\u526f\u7814\u7a76\u5458\uff0c\u535a\u58eb\u751f\u5bfc\u5e08\u3002\u7814\u7a76\u65b9\u5411\u662f\u590d\u6742\u6570\u636e\u5206\u6790\u548c\u7edf\u8ba1\u5b66\u4e60\uff0c\u5df2\u5728Biometrika\u3001Bernoulli\u3001Statistica Sinica\u7b49\u7edf\u8ba1\u5b66\u6742\u5fd7\u53d1\u8868\u5b66\u672f\u8bba\u658730\u591a\u7bc7\uff0c\u4e3b\u6301\u56fd\u5bb6\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u9762\u4e0a\u9879\u76ee\u3001\u9752\u5e74\u9879\u76ee\u53ca\u5929\u6d25\u5e02\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u5404\u4e00\u9879\u3002<\/p>\n\n\n\n\n\n

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

\u9898\u76ee\uff1a<\/strong>Combinatorial Causal Bandit<\/p>\n\n\n\n

\u6458\u8981\uff1a<\/strong>Combinatorial causal bandit (CCB) is the integration of causal inference, multi-armed bandit and combinatorial optimization techniques. CCB is the following online learning task: A learning agent is given a causal graph with unknown distributions on how each variable X is causally influenced by its parents. The learning task is carried out in T rounds. In each round, the agent selects at most K variables to intervene, the output of a target variable Y is its reward, and the output of all observed variables is the feedback to the agent. The agent aims at learn from the past feedback to help selecting intervention variables in subsequent rounds, so that in the end the cumulative reward of all rounds is maximized. The performance metric is (cumulative) regret, which is defined as the difference between the cumulative reward of always selecting the optimal set of intervention variables and selecting intervention variables according to the learning algorithm. We propose the CCB framework, and study CCB for generalized linear causal models and design learning algorithms with near optimal regret bounds.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb\uff1a<\/strong>\u9648\u536b\u662f\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u9ad8\u7ea7\u7814\u7a76\u5458\uff0c\u4e5f\u5728\u6e05\u534e\u5927\u5b66\u3001\u4e0a\u6d77\u4ea4\u901a\u5927\u5b66\u3001\u4e2d\u79d1\u9662\u7b49\u591a\u6240\u9662\u6821\u548c\u7814\u7a76\u673a\u6784\u62c5\u4efb\u5ba2\u5ea7\u6559\u6388\u6216\u7814\u7a76\u5458\u3002\u4ed6\u662f\u4e2d\u56fd\u8ba1\u7b97\u673a\u5b66\u4f1a\u7406\u8bba\u8ba1\u7b97\u673a\u4e13\u4e1a\u59d4\u5458\u4f1a\u7684\u5e38\u52a1\u59d4\u5458\uff0c\u4e5f\u662f\u5168\u56fd\u5927\u6570\u636e\u4e13\u5bb6\u59d4\u5458\u4f1a\u7684\u59d4\u5458\u3002\u4ed6\u662f\u56fd\u9645\u7535\u6c14\u548c\u7535\u5b50\u5de5\u7a0b\u5e08\u5b66\u4f1a\u7684\u4f1a\u58eb\uff08IEEE Fellow\uff09\u3002\u9648\u536b\u4e3b\u8981\u7684\u7814\u7a76\u65b9\u5411\u5305\u62ec\u793e\u4ea4\u548c\u4fe1\u606f\u7f51\u7edc\uff0c\u5728\u7ebf\u5b66\u4e60\uff0c\u7f51\u7edc\u535a\u5f08\u8bba\u548c\u7ecf\u6d4e\u5b66\uff0c\u5206\u5e03\u5f0f\u8ba1\u7b97\uff0c\u5bb9\u9519\u7b49\u3002\u4ed6\u57282013\u5e74\u4e0e\u4eba\u5408\u8457\u4e00\u672c\u82f1\u6587\u4e13\u8457\uff0c\u57282020\u5e74\u72ec\u7acb\u64b0\u5199\u4e00\u672c\u4e2d\u6587\u4e13\u8457\u3002\u4ed6\u5728\u591a\u4e2a\u5b66\u672f\u671f\u520a\u62c5\u4efb\u7f16\u59d4\uff0c\u4e5f\u5728\u591a\u4e2a\u5b66\u672f\u4f1a\u8bae\u4e2d\u62c5\u4efb\u8fc7\u6280\u672f\u59d4\u5458\u4f1a\u4e3b\u5e2d\u548c\u59d4\u5458\u3002\u9648\u536b\u4e8e\u6e05\u534e\u5927\u5b66\u83b7\u5f97\u672c\u79d1\u548c\u7855\u58eb\u6bd5\u4e1a\uff0c\u4e8e\u7f8e\u56fd\u5eb7\u5948\u5c14\u5927\u5b66\u83b7\u5f97\u535a\u58eb\u5b66\u4f4d\u3002\u6709\u5173\u9648\u536b\u66f4\u591a\u7684\u4fe1\u606f\uff0c\u6b22\u8fce\u8bbf\u95ee\u4ed6\u7684\u4e3b\u9875\uff1ahttp:\/\/research.microsoft.com\/en-us\/people\/weic\/.<\/p>\n\n\n\n\n\n

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

\u9898\u76ee<\/strong>\uff1aConfederated learning and Inference<\/p>\n\n\n\n

\u6458\u8981<\/strong>: The theory of statistical learning and inference for large-scale\/high-dimensional data analysis has recently attracted considerable interest. The central analytic task in the development of confederated statistical learning and inference pertains to the method of integrating results yielded from multiple\/sequential data batches. This talk introduced an one-step meta method based on confidence inference functions, a communication efficient method without pooling individual datasets for unbalanced datasets, and an incremental learning algorithm for streaming datasets with correlated outcomes. Integrative causal inference of multiple similar clinical studies conducted at different sites are also investigated.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb<\/strong>\uff1a\u5468\u5cad\uff0c2004-2010\u5e74\u56db\u5ddd\u5927\u5b66\u6570\u5b66\u5b66\u9662\u672c\u79d1\u548c\u7855\u58eb\uff0c2014\u5e74\u897f\u5357\u8d22\u7ecf\u5927\u5b66\u535a\u58eb\uff0c2018\u5e74\u7f8e\u56fd\u5bc6\u897f\u6839\u5927\u5b66\u751f\u7269\u7edf\u8ba1\u7cfb\u535a\u58eb\u540e\uff0c2017\u5e74\u949f\u5bb6\u5e86\u6570\u5b66\u5956\u83b7\u5f97\u8005\uff0c\u5468\u5cad\u4e0e\u5408\u4f5c\u8005\u5728\u6570\u636e\u96c6\u6210\u3001\u9009\u62e9\u540e\u63a8\u65ad\u3001\u4e9a\u7ec4\u5206\u6790\u3001\u975e\u53c2\u6570\u7406\u8bba\u4e0e\u65b9\u6cd5\u3001\u56e0\u679c\u63a8\u65ad\u7b49\u9886\u57df\u53d6\u5f97\u4e86\u4e00\u7cfb\u5217\u7814\u7a76\u6210\u679c\uff0c\u5728Journal of the American Statistical Association (JASA), Journal of Economics (JoE), Journal of Machine Learning Research(JMLR), Annal of Applied Statistics(AOAS), Biometrics\u7b49\u56fd\u9645\u7edf\u8ba1\u5b66\u3001\u8ba1\u91cf\u7ecf\u6d4e\u5b66\u3001\u8ba1\u7b97\u673a\u9886\u57df\u671f\u520a\u4e0a\u53d1\u8868\u8bba\u658720\u4f59\u7bc7\u3002<\/p>\n\n\n\n\n\n

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

\u9898\u76ee<\/strong>\uff1aNonparametric Estimation of Continuous Treatment Effect with Measurement Error<\/p>\n\n\n\n

\u6458\u8981<\/strong>\uff1aWe consider estimating the average dose-response function (ADRF) nonparametrically for continuous-valued treatment. The existing literature of continuous treatment effect proposed consistent estimators only for error-free data. However, in observational studies concerned by the literature of treatment effect, the treatment data can be measured with error. There, existing techniques are not applicable and finding a proper modification is not straightforward. We identify the ADRF by a weighted conditional expectation and estimate the weights nonparametrically by maximising a local generalised empirical likelihood subject to an expanding set of conditional moment equations incorporated with the deconvolution kernels. We then construct a deconvolution kernel estimator of the weighted conditional expectation. We derive the $L_2$ and $L_\\infty$ convergence rates of our weights estimator and the asymptotic bias and variance of our ADRF estimator. We also provide the asymptotic linear expansion of our ADRF estimator in both the ordinary smooth and the supersmooth error cases, which can help conduct statistical inference. We provide a data-driven method to select our smoothing parameters based on the simulation-extrapolation (SIMEX) idea and propose a new extrapolation procedure to stabilise the computation. Monte-Carlo simulations show a satisfactory finite-sample performance of our method, and a real data study illustrates its practical value.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb<\/strong>\uff1a\u5f20\u653f\uff0c\u4e2d\u56fd\u4eba\u6c11\u5927\u5b66\u7edf\u8ba1\u4e0e\u5927\u6570\u636e\u7814\u7a76\u9662\u52a9\u7406\u6559\u6388\uff0c2015\u5e74\u4e8e\u9999\u6e2f\u4e2d\u6587\u5927\u5b66\u7edf\u8ba1\u7cfb\u83b7\u535a\u58eb\u5b66\u4f4d\u3002\u7814\u7a76\u65b9\u5411\u5305\u62ec\u56e0\u679c\u63a8\u65ad\u3001\u7f3a\u5931\u6570\u636e\u3001\u6c61\u67d3\u6570\u636e\u3001\u534a\u53c2\u6570\u6a21\u578b\u7684\u6709\u6548\u4f30\u8ba1\u3001\u975e\u53c2\u6570\u7edf\u8ba1\u63a8\u65ad\u3001\u968f\u673a\u5fae\u5206\u65b9\u7a0b\u3001\u968f\u673a\u5206\u6790\u7b49\u3002\u5728JRSS-B, JOE, Quantitative Economics, JBES, Statistica Sinica, Stochastic Processes and their Applications\u7b49\u7edf\u8ba1\u3001\u8ba1\u91cf\u7ecf\u6d4e\u3001\u6982\u7387\u8bba\u56fd\u9645\u671f\u520a\u4e0a\u53d1\u8868\u8bba\u6587\u5341\u4f59\u7bc7\u3002\u4e3b\u6301\u56fd\u5bb6\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u9752\u5e74\u57fa\u91d1\uff0c\u5317\u4eac\u5e02\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u9762\u4e0a\u9879\u76ee\u3002<\/p>\n\n\n\n\n\n

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

\u9898\u76ee<\/strong>: A novel stable higher-order influence function estimators for doubly-robust functionals<\/p>\n\n\n\n

\u6458\u8981<\/strong>: In this talk, we will first review the concept of higher-order influence functions (HOIFs) and HOIF-based estimators for a class of smooth statistical functionals commonly encountered in causality. We will demonstrate applications of HOIF-based estimators when deep-learning is being deployed in applied data analysis, including applied causal inference tasks, despite our theoretical understanding about deep learning being still quite limited. Motivated from some of our empirical experience, we recently developed a new class of HOIF-based estimators, which, somewhat surprisingly, enjoy both the nice theoretical properties of the original HOIF-based estimators proposed in 2008 by Robins et al. and possibly more importantly, the numerical stability in finite-sample settings. This new class of HOIF-based estimators (1) paves the way towards making HOIFs practically useful, (2) bridges the HOIF estimators developed in Robins et al. 2008 and the cross-fitting estimators in Newey and Robins 2018, and (3) coincides the HOIF estimator in the fixed-design setting. We envision that this new class of HOIF estimators will be useful in applied works, at least before the myth of deep learning is completely resolved. This talk is based on two working papers, one empirical paper with Kerollos N. Wanis and Jamie Robins and one theoretical paper with Chang Li (a senior undergraduate student at Shanghai Jiao Tong University).<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb<\/strong>\uff1aLin Liu is an Assistant Professor in the Institute of Natural Sciences, School of Mathematical Sciences, and SJTU-Yale Joint Center for Biostatistics and Data Science at Shanghai Jiao Tong University, and PI in the Shanghai AI lab. He completed his PhD in biostatistics at Harvard University, under the supervision of Franziska Michor & Jamie Robins. His research interests lie in the intersection between mathematical statistics, causal inference, machine learning, and applied statistics in biomedical sciences.<\/p>\n\n\n\n\n\n

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

\u9898\u76ee<\/strong>: Estimation and inference for high-dimensional nonparametric additive instrumental-variables regression<\/p>\n\n\n\n

\u6458\u8981<\/strong>: The method of instrumental variables provides a fundamental and practical tool for causal inference in the presence of unmeasured confounding between the treatments and the outcome in various empirical studies. Modern data such as the genetical genomics data in these studies can be high-dimensional. The<\/p>\n\n\n\n

high-dimensional linear instrumental-variables model has been considered in the literature due to its simplicity albeit the true relationship may be nonlinear. We propose a more data-driven approach by considering nonparametric additive models between the instrumental variables and the treatments while keeping the linear model assumption between the treatments and the outcome so that the coefficients therein can directly bear causal interpretation. We provide a two-stage framework for estimation and inference under this more general setup. The group lasso regularization is first employed to select optimal instruments for the high-dimensional nonparametric additive model, and then the outcome variable is regressed on the fitted values from the nonparametric additive model to identify and estimate important treatment effects. We provide non-asymptotic analysis for the estimation error of the proposed estimator. A debiased procedure is further employed to establish valid inference. Extensive numerical experiments show that our proposed method can rival or outperform existing approaches in the literature. We finally analyze the mouse obesity data with the proposed method and discuss new discoveries.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb<\/strong>\uff1a\u4e2d\u56fd\u4eba\u6c11\u5927\u5b66\u7edf\u8ba1\u5b66\u9662\uff0c\u751f\u7269\u7edf\u8ba1\u4e0e\u6d41\u884c\u75c5\u5b66\u7cfb\u8bb2\u5e08\uff0c\u5317\u4eac\u5927\u5b66\u6570\u5b66\u79d1\u5b66\u5b66\u9662\u535a\u58eb\u3002\u4e3b\u8981\u7814\u7a76\u9886\u57df\u4e3a\u56e0\u679c\u63a8\u65ad\u3001\u7f3a\u5931\u6570\u636e\u3001\u9ad8\u7ef4\u7edf\u8ba1\u7b49\u3002\u76ee\u524d\u5df2\u5728\u5305\u62ecBiometrika, Journal of Econometrics, Biometrics\u7b49\u56fd\u9645\u8457\u540d\u7edf\u8ba1\u671f\u520a\u4e0a\u53d1\u8868\u591a\u7bc7\u5b66\u672f\u8bba\u6587\u3002\u4e3b\u6301\u4e00\u9879\u56fd\u5bb6\u81ea\u7136\u79d1\u5b66\u9752\u5e74\u57fa\u91d1\u9879\u76ee\uff0c\u53c2\u4e0e\u5b8c\u6210\u591a\u9879\u56fd\u5bb6\u81ea\u7136\u79d1\u5b66\u57fa\u91d1\u9762\u4e0a\u9879\u76ee\u3002<\/p>\n\n\n\n\n\n

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

\u9898\u76ee\uff1a<\/strong>Improving out-of-Distribution Performance of Machine Learning Models from a Causal Perspective<\/p>\n\n\n\n

\u6458\u8981\uff1a<\/strong>Given the remarkable performance of modern machine learning models on various benchmarking datasets, people turn to the next challenges in their wider applications. Among these, out-of-distribution (OOD) generalization is a critical one, since in many real-world tasks, the deploying environment is different from the training one, causing a change in data distribution. Causality provides an insightful approach to analyze and handle the problem. It proposes the model should learn causal relations which represents the fundamental rule governing the data in all environments, in contrast to superficial relations that may only appear in a specific environment accidentally. In this talk, we introduce a model and its variants for prediction\/classification tasks which is designed following a causal reasoning process. We show their causal identification guarantees and OOD generalization analysis, and also the improved empirical performance.<\/p>\n\n\n\n

\u4e2a\u4eba\u7b80\u4ecb\uff1a<\/strong>\u5218\u7545\uff0c\u73b0\u4e3a\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u673a\u5668\u5b66\u4e60\u7ec4\u4e3b\u7ba1\u7814\u7a76\u5458\uff0c2019\u5e74\u4e8e\u6e05\u534e\u5927\u5b66\u8ba1\u7b97\u673a\u7cfb\u53d6\u5f97\u535a\u58eb\u5b66\u4f4d\u3002\u4e3b\u8981\u7814\u7a76\u65b9\u5411\u5305\u62ec\u8d1d\u53f6\u65af\u63a8\u65ad\u65b9\u6cd5\uff0c\u56e0\u679c\u6a21\u578b\uff0c\u751f\u6210\u5f0f\u6a21\u578b\u53ca\u5176\u4e0e\u7269\u7406\u5b66\u95ee\u9898\u7684\u7ed3\u5408\u3002<\/p>\n\n\n\n\n\n

<\/p>\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
\u65e5\u671f<\/th>\n\u65f6\u95f4<\/th>\n\u62a5\u544a\u4eba<\/th>\n\u62a5\u544a\u9898\u76ee<\/th>\n\u4e3b\u6301\u4eba<\/th>\n<\/tr>\n<\/thead>\n
4\u67081\u65e5<\/td>\n14:00-22:00<\/td>\n\u62a5\u5230\u6ce8\u518c<\/td>\n<\/tr>\n
4\u67082\u65e5<\/td>\n08:50-09:10<\/td>\n\u5f00\u5e55\u5f0f\u3001\u81f4\u8f9e<\/td>\n\u90ed\u5609\u4e30<\/td>\n<\/tr>\n
09:10-09:40<\/td>\n\u6797\u534e\u73cd\uff08\u897f\u5357\u8d22\u7ecf\u5927\u5b66\uff09<\/td>\nRobust and efficient estimation for treatment effect in causal inference<\/td>\n\u9648\u8587<\/td>\n<\/tr>\n
09:40-10:10<\/td>\n\u5d14\u9e4f\uff08\u6e05\u534e\u5927\u5b66\uff09<\/td>\n\u56e0\u679c\u542f\u53d1\u7684\u7a33\u5b9a\u5b66\u4e60<\/td>\n<\/tr>\n
10:10-10:40<\/td>\n\u6797\u4f1f\uff08\u5317\u4eac\u5927\u5b66\uff09<\/td>\nDeconfounding with the Blessing of Dimensionality<\/td>\n<\/tr>\n
10:40-11:00<\/td>\n\u8336\u6b47<\/td>\n<\/tr>\n
11:00-11:30<\/td>\n\u82d7\u65fa\uff08\u5317\u4eac\u5927\u5b66\uff09<\/td>\n\u56e0\u679c\u63a8\u65ad\uff0c\u89c2\u5bdf\u6027\u7814\u7a76\u548c\u8bfa\u8d1d\u5c14\u7ecf\u6d4e\u5b66\u5956<\/td>\n\u90b9\u957f\u4eae<\/td>\n<\/tr>\n
11:30-12:00<\/td>\n\u4e01\u9510\uff08\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\uff09<\/td>\nSupervised Causal Learning: A New Frontier of Causal Discovery<\/td>\n<\/tr>\n
12:00-12:30<\/td>\n\u738b\u78ca\uff08\u5357\u5f00\u5927\u5b66\uff09<\/td>\nGeneralized regression estimators for average treatment effect with multicollinearity in high-dimensional covariates<\/td>\n<\/tr>\n
12:30-14:00<\/td>\n\u5348\u9910<\/td>\n<\/tr>\n
14:00-14:30<\/td>\n\u9648\u536b\uff08\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\uff09<\/td>\nCombinatorial Causal Bandit<\/td>\n\u82d7\u65fa<\/td>\n<\/tr>\n
14:30-15:00<\/td>\n\u5468\u5cad\uff08\u897f\u5357\u8d22\u7ecf\u5927\u5b66\uff09<\/td>\nConfederated learning and Inference<\/td>\n<\/tr>\n
15:00-15:30<\/td>\n\u5f20\u653f\uff08\u4e2d\u56fd\u4eba\u6c11\u5927\u5b66\uff09<\/td>\nNonparametric Estimation of Continuous Treatment Effect with Measurement Error<\/td>\n<\/tr>\n
15:30-15:50<\/td>\n\u8336\u6b47<\/td>\n<\/tr>\n
15:50-16:20<\/td>\n\u5218\u6797\uff08\u4e0a\u6d77\u4ea4\u901a\u5927\u5b66\uff09<\/td>\nA novel stable higher-order influence function estimators for doubly-robust functionals<\/td>\n\u9648\u536b<\/td>\n<\/tr>\n
16:20-16:50<\/td>\n\u674e\u4f1f\uff08\u4e2d\u56fd\u4eba\u6c11\u5927\u5b66\uff09<\/td>\nEstimation and inference for high-dimensional nonparametric additive instrumental-variables regression<\/td>\n<\/tr>\n
16:50-17:20<\/td>\n\u5218\u7545\uff08\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\uff09<\/td>\nImproving out-of-Distribution Performance of Machine Learning Models from a Causal Perspective<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n\n\n\n
<\/div>\n\n\n","tab-content":[],"msr_startdate":"2022-04-02","msr_enddate":"2022-04-02","msr_event_time":"","msr_location":"\u4e2d\u56fd\u79d1\u5b66\u9662\u8ba1\u7b97\u6280\u672f\u7814\u7a76\u6240","msr_event_link":"","msr_event_recording_link":"","msr_startdate_formatted":"April 2, 2022","msr_register_text":"Watch now","msr_cta_link":"","msr_cta_text":"","msr_cta_bi_name":"","featured_image_thumbnail":null,"event_excerpt":"\u56e0\u679c\u63a8\u65ad\u662f\u8fd1\u5e74\u6765\u6570\u636e\u79d1\u5b66\u548c\u4eba\u5de5\u667a\u80fd\u7814\u7a76\u7684\u70ed\u70b9\u4e4b\u4e00\uff0c\u5f97\u5230\u4e86\u5b66\u672f\u754c\u548c\u4e1a\u754c\u7684\u5e7f\u6cdb\u5173\u6ce8\u3002\u672c\u6b21\u7814\u8ba8\u4f1a\u65e8\u5728\u8fdb\u4e00\u6b65\u4fc3\u8fdb\u56e0\u679c\u63a8\u65ad\u9886\u57df\u548c\u673a\u5668\u5b66\u4e60\u9886\u57df\u56fd\u5185\u5b66\u8005\u7684\u5b66\u672f\u4ea4\u6d41\uff0c\u63a2\u8ba8\u56e0\u679c\u63a8\u65ad\u4e0e\u673a\u5668\u5b66\u4e60\u7684\u7ed3\u5408\u65b9\u5f0f\u3002\u672c\u6b21\u7814\u8ba8\u4f1a\u6709\u5e78\u9080\u8bf7\u4e8612\u4f4d\u76f8\u5173\u9886\u57df\u7814\u7a76\u7684\u4e13\u5bb6\u5b66\u8005\u8fdb\u884c\u5b66\u672f\u62a5\u544a\uff0c\u5e7f\u6cdb\u5f00\u5c55\u5b66\u672f\u63a2\u8ba8\uff0c\u4e3a\u76f8\u5173\u9886\u57df\u7684\u7814\u7a76\u4eba\u5458\u63d0\u4f9b\u4e00\u4e2a\u4e13\u4e1a\u7684\u4ea4\u6d41\u5e73\u53f0\u3002\u7814\u8ba8\u4f1a\u5c06\u4e8e2022\u5e744\u67082\u65e5\u5728\u4e2d\u56fd\u79d1\u5b66\u9662\u8ba1\u7b97\u6280\u672f\u7814\u7a76\u6240\u56db\u5c42\u62a5\u544a\u5385\u4e3e\u884c\uff0c\u7814\u8ba8\u4f1a\u4e0d\u6536\u53d6\u4f1a\u8bae\u6ce8\u518c\u8d39\uff0c\u5176\u4ed6\u8d39\u7528\u81ea\u7406\u3002 2022\u201c\u56e0\u679c\u63a8\u65ad\u4e0e\u673a\u5668\u5b66\u4e60\u201d\u7814\u8ba8\u4f1a\u7ec4\u59d4\u4f1a \u4e3b\u5e2d\uff1a\u9a6c\u5fd7\u660e\u3001\u7a0b\u5b66\u65d7\u3001\u5218\u94c1\u5ca9 \u59d4\u5458\uff1a\u90ed\u5609\u4e30\u3001\u9648\u8587\u3001\u90b9\u957f\u4eae\u3001\u5468\u5ddd\u3001\u5b5f\u742a\u3001\u5f20\u5112\u6e05\u3001\u5b59\u4e3d\u541b \u65e5\u671f \u65f6\u95f4 \u62a5\u544a\u4eba \u62a5\u544a\u9898\u76ee \u4e3b\u6301\u4eba 4\u67081\u65e5 14:00-22:00 \u62a5\u5230\u6ce8\u518c 4\u67082\u65e5 08:50-09:10 \u5f00\u5e55\u5f0f\u3001\u81f4\u8f9e \u90ed\u5609\u4e30 09:10-09:40 \u6797\u534e\u73cd\uff08\u897f\u5357\u8d22\u7ecf\u5927\u5b66\uff09 Robust a...","msr_research_lab":[1012650],"related-researchers":[],"msr_impact_theme":[],"related-academic-programs":[],"related-groups":[],"related-projects":[],"related-opportunities":[],"related-publications":[],"related-videos":[],"related-posts":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event\/1065744","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-event"}],"version-history":[{"count":8,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event\/1065744\/revisions"}],"predecessor-version":[{"id":1081026,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event\/1065744\/revisions\/1081026"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1065744"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1065744"},{"taxonomy":"msr-region","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-region?post=1065744"},{"taxonomy":"msr-event-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event-type?post=1065744"},{"taxonomy":"msr-video-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video-type?post=1065744"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1065744"},{"taxonomy":"msr-program-audience","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-program-audience?post=1065744"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1065744"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=1065744"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}