{"id":40187,"date":"2025-04-17T08:02:49","date_gmt":"2025-04-17T00:02:49","guid":{"rendered":"https:\/\/17aitech.com\/?p=40187"},"modified":"2025-04-17T08:02:49","modified_gmt":"2025-04-17T00:02:49","slug":"%e7%bb%a7vae%e4%b9%8b%e5%90%8e%ef%bc%8cadam%e4%b9%9f%e6%8b%bf%e5%88%b0%e4%ba%86iclr-%e6%97%b6%e9%97%b4%e6%a3%80%e9%aa%8c%e5%a5%96%ef%bc%8copenai%e8%81%94%e5%88%9b%e4%b8%a4%e6%ac%a1%e8%8e%b7%e5%a5%96","status":"publish","type":"post","link":"https:\/\/17aitech.com\/?p=40187","title":{"rendered":"\u7ee7VAE\u4e4b\u540e\uff0cAdam\u4e5f\u62ff\u5230\u4e86ICLR \u65f6\u95f4\u68c0\u9a8c\u5956\uff0cOpenAI\u8054\u521b\u4e24\u6b21\u83b7\u5956"},"content":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2025-04-15-11\" target=\"_blank\">\u7ee7VAE\u4e4b\u540e\uff0cAdam\u4e5f\u62ff\u5230\u4e86ICLR \u65f6\u95f4\u68c0\u9a8c\u5956\uff0cOpenAI\u8054\u521b\u4e24\u6b21\u83b7\u5956<\/a><\/p>\n<p>\u521a\u521a\uff0cICLR 2025 \u5ba3\u5e03\u4e86\u4eca\u5e74\u7684\u65f6\u95f4\u68c0\u9a8c\u5956\u83b7\u5956\u8bba\u6587\u3002\u8fd9\u4e2a\u5956\u9879\u65e8\u5728\u8868\u5f70\u5341\u5e74\u524d\u5728 ICLR 2015 \u4e0a\u53d1\u8868\u7684\u5bf9\u8be5\u9886\u57df\u4ea7\u751f\u6301\u4e45\u5f71\u54cd\u7684\u8bba\u6587\u3002<\/p>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-37c819374825294a9541ddc5f8619755.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-37c819374825294a9541ddc5f8619755.png\"><\/a><\/section>\n<p>\u7ecf\u4e45\u4e0d\u8870\u7684\u300cAdam \u7b97\u6cd5\u300d\u62ff\u5230\u4e86\u51a0\u519b\uff0c\u6bd4 Transformer \u8bba\u6587\u66f4\u65e9\u5f15\u5165\u6ce8\u610f\u529b\u673a\u5236\u7684\u300cNeural Machine Translation by Jointly Learning to Align and Translate\u300d\u62ff\u5230\u4e86\u4e9a\u519b\u3002\u8fd9\u6837\u7684\u7ed3\u679c\u53ef\u8c13\u300c\u5b9e\u81f3\u540d\u5f52\u300d\u3002<\/p>\n<p><strong>Adam\uff1a\u6df1\u5ea6\u5b66\u4e60\u9886\u57df\u5e94\u7528\u6700\u5e7f\u6cdb\u7684\u4f18\u5316\u7b97\u6cd5\u4e4b\u4e00<\/strong><\/p>\n<p>Adam \u7b97\u6cd5\uff0c\u5373 Adaptive Moment Estimation \u7b97\u6cd5\uff0c\u662f\u4e00\u79cd\u7ed3\u5408\u4e86\u52a8\u91cf\u6cd5\u548c RMSprop \u7b97\u6cd5\u601d\u60f3\u7684\u81ea\u9002\u5e94\u5b66\u4e60\u7387\u4f18\u5316\u7b97\u6cd5\u3002<\/p>\n<p>Adam \u9769\u547d\u6027\u5730\u6539\u53d8\u4e86\u795e\u7ecf\u7f51\u7edc\u8bad\u7ec3\uff0c\u5728\u5404\u79cd\u67b6\u6784\u548c\u4efb\u52a1\u4e2d\u5b9e\u73b0\u4e86\u663e\u8457\u66f4\u5feb\u7684\u6536\u655b\u901f\u5ea6\u548c\u66f4\u7a33\u5b9a\u7684\u8bad\u7ec3\u8fc7\u7a0b\u3002\u5982\u4eca\uff0c\u5b83\u5df2\u7ecf\u6210\u4e3a\u6df1\u5ea6\u5b66\u4e60\u4e2d\u6700\u5e7f\u6cdb\u91c7\u7528\u7684\u4f18\u5316\u7b97\u6cd5\u4e4b\u4e00\u3002<\/p>\n<p>Adam \u7b97\u6cd5\u7684\u521b\u65b0\u4e4b\u5904\u5728\u4e8e\uff0c\u5b83\u7ed3\u5408\u4e86 AdaGrad \u548c RMSProp \u7684\u4f18\u70b9\uff0c\u901a\u8fc7\u7ef4\u62a4\u7b2c\u4e00\u77e9\u4f30\u8ba1\uff08\u5747\u503c\uff09\u548c\u7b2c\u4e8c\u77e9\u4f30\u8ba1\uff08\u672a\u4e2d\u5fc3\u5316\u7684\u65b9\u5dee\uff09\u6765\u52a8\u6001\u8c03\u6574\u6bcf\u4e2a\u53c2\u6570\u7684\u5b66\u4e60\u7387\uff0c\u80fd\u591f\u5904\u7406\u7a00\u758f\u68af\u5ea6\u548c\u975e\u5e73\u7a33\u76ee\u6807\u51fd\u6570\u3002<\/p>\n<p>\u5b83\u7684\u91cd\u8981\u6027\u4f53\u73b0\u5728\u5176\u5e7f\u6cdb\u7684\u5e94\u7528\u573a\u666f\uff0c\u65e0\u8bba\u662f\u8ba1\u7b97\u673a\u89c6\u89c9\u3001\u81ea\u7136\u8bed\u8a00\u5904\u7406\u8fd8\u662f\u5f3a\u5316\u5b66\u4e60\u7b49\u9886\u57df\uff0cAdam \u7b97\u6cd5\u90fd\u4e3a\u6a21\u578b\u7684\u9ad8\u6548\u8bad\u7ec3\u63d0\u4f9b\u4e86\u6709\u529b\u652f\u6301\uff0c\u6781\u5927\u5730\u63d0\u9ad8\u4e86\u6a21\u578b\u7684\u8bad\u7ec3\u6548\u7387\u548c\u6027\u80fd\uff0c\u6210\u4e3a\u65e0\u6570\u6700\u5148\u8fdb\u6a21\u578b\u7684\u9ed8\u8ba4\u4f18\u5316\u5668\u3002<\/p>\n<p>\u6839\u636e Google Scholar \u7684\u6570\u636e\uff0cAdam \u8bba\u6587\u7684\u88ab\u5f15\u91cf\u5df2\u7ecf\u8d85\u8fc7 21 \u4e07\u3002<\/p>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-ee6d5b8ece9ad25d1d7bde621da85477.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-ee6d5b8ece9ad25d1d7bde621da85477.png\"><\/a><\/section>\n<p>\u8bba\u6587\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/abs\/1412.6980\u00a0<\/p>\n<p>\u8be5\u8bba\u6587\u603b\u5171\u6709\u4e24\u4f4d\u4f5c\u8005 \u2014\u2014Diederik P. Kingma \u548c Jimmy Lei Ba\u3002<\/p>\n<p>Diederik P. Kingma \u662f\u4e00\u4f4d\u8ba1\u7b97\u673a\u79d1\u5b66\u5bb6\u548c\u7814\u7a76\u5458\uff0c\u4e13\u6ce8\u4e8e\u673a\u5668\u5b66\u4e60\u548c\u751f\u6210\u5f0f\u5efa\u6a21\u7684\u53ef\u6269\u5c55\u65b9\u6cd5\u3002\u4ed6\u7684\u91cd\u8981\u8d21\u732e\u5305\u62ec\u53d8\u5206\u81ea\u7f16\u7801\u5668\uff08VAE\uff09\u3001Adam \u4f18\u5316\u5668\u3001Glow \u548c\u53d8\u5206\u6269\u6563\u6a21\u578b\u3002\u5176\u4e2d\uff0cVAE \u83b7\u5f97\u4e86\u53bb\u5e74\u7684 ICLR \u65f6\u95f4\u68c0\u9a8c\u5956\u3002\u6240\u4ee5\uff0c\u8fd9\u6b21\u662f Diederik P. Kingma \u7b2c\u4e8c\u6b21\u83b7\u5f97\u8be5\u5956\u9879\u3002\u800c\u622a\u81f3\u76ee\u524d\uff0c\u8be5\u5956\u9879\u603b\u5171\u5c31\u9881\u4e86\u4e24\u6b21\u3002<\/p>\n<p>2017 \u5e74\uff0c\u4ed6\u4ee5\u4f18\u5f02\u6210\u7ee9\u83b7\u5f97\u963f\u59c6\u65af\u7279\u4e39\u5927\u5b66\u535a\u58eb\u5b66\u4f4d\u30022015 \u5e74\uff0c\u4ed6\u4f5c\u4e3a\u521b\u59cb\u56e2\u961f\u6210\u5458\u53c2\u4e0e\u521b\u5efa\u4e86 OpenAI\u3002\u5728\u6b64\u4e4b\u524d\uff0c\u4ed6\u5171\u540c\u521b\u7acb\u4e86 Advanza \u516c\u53f8\uff0c\u8be5\u516c\u53f8\u4e8e 2016 \u5e74\u88ab\u6536\u8d2d\u3002<\/p>\n<p>\u4ed6\u7684\u804c\u4e1a\u751f\u6daf\u5305\u62ec\uff1a<\/p>\n<ul>\n<li>\n<p>2009 \u5e74\u548c 2012 \u5e74\uff1a\u5728\u7ebd\u7ea6\u5927\u5b66 Yann LeCun \u5b9e\u9a8c\u5ba4\u62c5\u4efb\u521d\u7ea7\u7814\u7a76\u79d1\u5b66\u5bb6\uff1b<\/p>\n<\/li>\n<li>\n<p>2010-2012 \u5e74\uff1aAdvanza \u8054\u5408\u521b\u59cb\u4eba\u517c\u6280\u672f\u8d1f\u8d23\u4eba\uff0c\u516c\u53f8\u4e8e 2016 \u5e74\u6210\u529f\u9000\u51fa\uff1b<\/p>\n<\/li>\n<li>\n<p>2013-2017 \u5e74\uff1a\u5728\u963f\u59c6\u65af\u7279\u4e39\u5927\u5b66\u653b\u8bfb\u535a\u58eb\u5b66\u4f4d\uff0c\u5bfc\u5e08\u4e3a Max Welling\uff0c\u7814\u7a76\u9886\u57df\u4e3a\u6df1\u5ea6\u5b66\u4e60\u548c\u751f\u6210\u6a21\u578b\uff1b<\/p>\n<\/li>\n<li>\n<p>2015-2018 \u5e74\uff1aOpenAI \u521b\u59cb\u56e2\u961f\u6210\u5458\u548c\u7814\u7a76\u79d1\u5b66\u5bb6\uff0c\u9886\u5bfc\u7b97\u6cd5\u56e2\u961f\uff0c\u4e13\u6ce8\u4e8e\u57fa\u7840\u7814\u7a76\uff1b<\/p>\n<\/li>\n<li>\n<p>2018-2024 \u5e74\uff1a\u5728 Google Brain\/DeepMind \u62c5\u4efb\u7814\u7a76\u79d1\u5b66\u5bb6\uff0c\u9886\u5bfc\u591a\u4e2a\u7814\u7a76\u9879\u76ee\uff0c\u4e3b\u8981\u662f\u6587\u672c\u3001\u56fe\u50cf\u548c\u89c6\u9891\u7684\u751f\u6210\u6a21\u578b\uff1b<\/p>\n<\/li>\n<li>\n<p>2024 \u5e74\u81f3\u4eca\uff1a\u5728 Anthropic \u7814\u7a76\u5927\u89c4\u6a21\u673a\u5668\u5b66\u4e60\u3002<\/p>\n<\/li>\n<\/ul>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-c79e71c585a02bf758026193afcb3f8c.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-c79e71c585a02bf758026193afcb3f8c.png\"><\/a><\/section>\n<p>\u53e6\u4e00\u4f4d\u4f5c\u8005\u591a\u4f26\u591a\u5927\u5b66 Jimmy Lei Ba \u73b0\u5728\u662f\u591a\u4f26\u591a\u5927\u5b66\u7684\u52a9\u7406\u6559\u6388\u3002\u4ed6\u7684\u957f\u671f\u7814\u7a76\u76ee\u6807\u662f\u89e3\u51b3\u4e00\u4e2a\u8ba1\u7b97\u95ee\u9898\uff1a\u5982\u4f55\u6784\u5efa\u5177\u6709\u4eba\u7c7b\u6548\u7387\u548c\u9002\u5e94\u6027\u7684\u901a\u7528\u95ee\u9898\u89e3\u51b3\u673a\u5668\u3002\u4ed6\u7684\u7814\u7a76\u5174\u8da3\u4e3b\u8981\u96c6\u4e2d\u5728\u5f00\u53d1\u9ad8\u6548\u7684\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u5b66\u4e60\u7b97\u6cd5\u4e0a\u3002\u4ed6\u7684\u7814\u7a76\u9886\u57df\u4e0e NeurIPS\u3001ICLR \u548c ICML \u7b49\u7814\u7a76\u793e\u533a\u6709\u4ea4\u53c9\u3002\u6b64\u5916\uff0c\u4ed6\u5bf9\u5f3a\u5316\u5b66\u4e60\u3001\u81ea\u7136\u8bed\u8a00\u5904\u7406\u548c\u4eba\u5de5\u667a\u80fd\u4e5f\u6709\u5e7f\u6cdb\u7684\u5174\u8da3\u3002<\/p>\n<p>Jimmy Lei Ba \u5728 Geoffrey Hinton \u7684\u6307\u5bfc\u4e0b\u5b8c\u6210\u4e86\u535a\u58eb\u5b66\u4f4d\u3002\u4ed6\u7684\u7855\u58eb\uff082014 \u5e74\uff09\u548c\u672c\u79d1\u5b66\u4f4d\uff082011 \u5e74\uff09\u5747\u5728\u591a\u4f26\u591a\u5927\u5b66\u83b7\u5f97\uff0c\u5e08\u4ece Brendan Frey \u548c Ruslan Salakhutdinov\u3002\u4ed6\u662f CIFAR AI chair\uff0c\u66fe\u5728 2016 \u5e74\u83b7\u5f97 Facebook \u673a\u5668\u5b66\u4e60\u7814\u7a76\u751f\u5956\u5b66\u91d1\u3002<\/p>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-dc560bb4e6a38e9f9c9e2ba8b834637f.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-dc560bb4e6a38e9f9c9e2ba8b834637f.png\"><\/a><\/section>\n<p><strong>\u4e9a\u519b\u8bba\u6587 \u00a0 \u6bd4 Transformer \u63d0\u524d\u4e09\u5e74\u5f15\u5165\u6ce8\u610f\u529b\u673a\u5236<\/strong><\/p>\n<p>\u83b7\u5f97 ICLR \u65f6\u95f4\u68c0\u9a8c\u5956\u4e9a\u519b\u7684\u8bba\u6587\u300cNeural Machine Translation by Jointly Learning to Align and Translate\u300d\u540c\u6837\u6765\u5934\u4e0d\u5c0f\uff0c\u56fe\u7075\u5956\u5f97\u4e3b Yoshua Bengio \u662f\u8be5\u8bba\u6587\u7684\u4f5c\u8005\u4e4b\u4e00\u3002<\/p>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-8ac814eb25c417cd1f9f4ce802b71607.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-8ac814eb25c417cd1f9f4ce802b71607.png\"><\/a><\/section>\n<p>\u8bba\u6587\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/abs\/1409.0473<\/p>\n<p>\u53bb\u5e74 12 \u6708\uff0cAI \u5927\u725b Andrej Karpathy \u4e13\u95e8\u53d1\u5e16\u4e3a\u8fd9\u7bc7\u8bba\u6587\u9e23\u4e0d\u5e73\uff0c\u79f0\u5b83\u662f\u300c\u771f\u6b63\u5f15\u5165\u6ce8\u610f\u529b\u673a\u5236\u7684\u8bba\u6587\u300d\uff0c\u4f46\u53d7\u5230\u7684\u5173\u6ce8\u8fdc\u8fdc\u4e0d\u53ca\u4e09\u5e74\u540e\u624d\u51fa\u73b0\u7684 Transformer \u8bba\u6587\u3002<\/p>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-77a73e7526e96a525db3bf70219479e6.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-77a73e7526e96a525db3bf70219479e6.png\"><\/a><\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-336f1989eec062724324e503ac5cde7f.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-336f1989eec062724324e503ac5cde7f.png\"><\/a><\/section>\n<p>\u5177\u4f53\u6765\u8bf4\uff0c\u8fd9\u7bc7\u8bba\u6587\u7684\u521b\u65b0\u4e4b\u5904\u5728\u4e8e\uff0c\u5b83\u5f15\u5165\u4e86\u4e00\u79cd\u6ce8\u610f\u529b\u5f62\u5f0f\uff0c\u4ece\u6839\u672c\u4e0a\u6539\u53d8\u4e86\u5e8f\u5217\u5230\u5e8f\u5217\u6a21\u578b\u5904\u7406\u4fe1\u606f\u7684\u65b9\u5f0f\u3002\u5728\u6b64\u4e4b\u524d\uff0c\u7f16\u7801\u5668 &#8211; \u89e3\u7801\u5668\u67b6\u6784\u901a\u5e38\u5c06\u6574\u4e2a\u8f93\u5165\u5e8f\u5217\u538b\u7f29\u4e3a\u56fa\u5b9a\u957f\u5ea6\u7684\u5411\u91cf\uff0c\u4ece\u800c\u4e3a\u8f83\u957f\u7684\u5e8f\u5217\u9020\u6210\u5185\u5b58\u74f6\u9888\u3002\u8be5\u65b9\u6cd5\u4f7f\u6a21\u578b\u80fd\u591f\u5728\u7ffb\u8bd1\u8fc7\u7a0b\u4e2d\u52a8\u6001\u5730\u300c\u5173\u6ce8\u300d\u6e90\u53e5\u5b50\u7684\u4e0d\u540c\u90e8\u5206\uff0c\u4ece\u800c\u5904\u7406\u76f8\u5173\u7684\u4e0a\u4e0b\u6587\u4fe1\u606f\u3002\u8fd9\u79cd\u6ce8\u610f\u529b\u673a\u5236\u5df2\u7ecf\u6210\u4e3a\u73b0\u4ee3\u6df1\u5ea6\u5b66\u4e60\u7684\u57fa\u77f3\uff0c\u8fdc\u8fdc\u8d85\u51fa\u4e86\u673a\u5668\u7ffb\u8bd1\u7684\u8303\u7574\uff0c\u5f62\u6210\u4e86 transformer \u548c\u5927\u578b\u8bed\u8a00\u6a21\u578b\u7684\u57fa\u7840\u3002\u8fd9\u7bc7\u8bba\u6587\u7684\u5b9e\u9645\u5f71\u54cd\u662f\u5de8\u5927\u7684\uff0c\u4f7f\u5176\u6210\u4e3a\u5bf9\u795e\u7ecf\u7f51\u7edc\u67b6\u6784\u6700\u6709\u5f71\u54cd\u529b\u7684\u8d21\u732e\u4e4b\u4e00\u3002<\/p>\n<p>\u8fd9\u7bc7\u8bba\u6587\u603b\u5171\u6709\u4e09\u4f4d\u4f5c\u8005\uff1aDzmitry Bahdanau\u3001Kyunghyun Cho \u548c Yoshua Bengio<\/p>\n<p>Dzmitry Bahdanau \u5728\u52a0\u62ff\u5927 Mila \u7814\u7a76\u6240\u5b8c\u6210\u4e86\u535a\u58eb\u5b66\u4f4d\uff0c\u5e08\u4ece Yoshua Bengio\u3002\u4ed6\u73b0\u5728\u662f\u4e00\u4f4d\u5728 ServiceNow \u65d7\u4e0b\uff08\u524d Element AI\uff09\u5de5\u4f5c\u7684\u7814\u7a76\u79d1\u5b66\u5bb6\u3002\u4ed6\u540c\u65f6\u4e5f\u662f Mila \u7684\u6838\u5fc3\u4ea7\u4e1a\u6210\u5458\u548c\u9ea6\u5409\u5c14\u5927\u5b66\u7684\u517c\u804c\u6559\u6388\u3002<\/p>\n<p>Dzmitry Bahdanau \u575a\u4fe1\u4eba\u7c7b\u8bed\u8a00\u6280\u672f\uff08HLT\uff0c\u4ed6\u8ba4\u4e3a\u8fd9\u662f\u6bd4 NLP \u66f4\u597d\u7684\u79f0\u547c\uff09\u5c06\u6539\u53d8\u4eba\u7c7b\u4e0e\u8f6f\u4ef6\u4ea4\u4e92\u548c\u83b7\u53d6\u77e5\u8bc6\u7684\u65b9\u5f0f\u3002\u4e8b\u5b9e\u4e0a\uff0c\u8fd9\u79cd\u6539\u53d8\u5df2\u7ecf\u5f00\u59cb\u53d1\u751f\uff08\u6bd4\u5982\u7f51\u7edc\u641c\u7d22\uff09\uff0c\u4f46\u8fd9\u4ec5\u4ec5\u662f\u4e2a\u5f00\u59cb\u3002Bahdanau \u5bf9 HLT \u6280\u672f\u6808\u5404\u4e2a\u5c42\u9762\u7684\u7814\u7a76\u95ee\u9898\u90fd\u6709\u6d53\u539a\u5174\u8da3\uff0c\u5305\u62ec\u6df1\u5ea6\u5b66\u4e60\u57fa\u7840\u3001\u57fa\u7840\u6a21\u578b\u8bad\u7ec3\u3001\u7279\u5b9a\u4efb\u52a1\u7b97\u6cd5\uff08\u5c24\u5176\u662f\u8bed\u4e49\u89e3\u6790\uff09\u4ee5\u53ca\u7528\u6237\u4e0e AI \u7cfb\u7edf\u7684\u4ea4\u4e92\u4f53\u9a8c\u3002\u4ed6\u8fd1\u671f\u6b63\u5728\u8fdb\u884c\u7684\u5de5\u4f5c\u4e3b\u8981\u96c6\u4e2d\u5728\u8bed\u4e49\u89e3\u6790\u548c\u9762\u5411\u4efb\u52a1\u7684\u5bf9\u8bdd\u65b9\u6cd5\u3001\u4ee3\u7801\u751f\u6210\u3001\u795e\u7ecf\u6a21\u578b\u7684\u7cfb\u7edf\u6027\uff08\u7ec4\u5408\u6027\uff09\u6cdb\u5316\u548c\u6837\u672c\u6548\u7387\u4e0a\u3002<\/p>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-ade2d19ba7fce9e99074a3dfd74119f4.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-ade2d19ba7fce9e99074a3dfd74119f4.png\"><\/a><\/section>\n<p>\u5728\u53d1\u7ed9 Karpathy \u7684\u90ae\u4ef6\u4e2d\uff0cDzmitry Bahdanau \u66fe\u8be6\u7ec6\u8c08\u8fc7\u4ed6\u4eec\u5f53\u5e74\u7814\u7a76\u300c\u6ce8\u610f\u529b\u673a\u5236\u300d\u7684\u5f80\u4e8b\u3002<\/p>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-5856e2c74f3da99928e2dbc3475d8d07.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-5856e2c74f3da99928e2dbc3475d8d07.png\"><\/a><\/section>\n<p>\u4ed6\u63d0\u5230\uff0c\u300c \u6211\u4f5c\u4e3a\u5b9e\u4e60\u751f\u6765\u5230 Yoshua \u7684\u5b9e\u9a8c\u5ba4\u2026\u2026 \u6211\u544a\u8bc9 Yoshua \u6211\u613f\u610f\u505a\u4efb\u4f55\u5de5\u4f5c\u3002Yoshua \u5b89\u6392\u6211\u53c2\u4e0e\u673a\u5668\u7ffb\u8bd1\u9879\u76ee\u2026\u2026 \u6709\u4e00\u5929\uff0c\u6211\u7a81\u7136\u60f3\u5230\u8ba9\u89e3\u7801\u5668 RNN \u5b66\u4e60\u5728\u6e90\u5e8f\u5217\u4e2d\u641c\u7d22\u653e\u7f6e\u6e38\u6807\u7684\u4f4d\u7f6e\u4f1a\u5f88\u597d\u3002\u8fd9\u6709\u70b9\u53d7\u5230\u7ffb\u8bd1\u7ec3\u4e60\u7684\u542f\u53d1\uff0c\u5728\u4e2d\u5b66\u5b66\u4e60\u82f1\u8bed\u65f6\u4f1a\u7528\u5230\u8fd9\u79cd\u7ec3\u4e60\u3002\u5f53\u4f60\u7ffb\u8bd1\u65f6\uff0c\u76ee\u5149\u5728\u6e90\u5e8f\u5217\u548c\u76ee\u6807\u5e8f\u5217\u4e4b\u95f4\u6765\u56de\u79fb\u52a8\u3002\u6211\u63d0\u51fa\u4e86\u5bf9 BiRNN \u72b6\u6001\u8fdb\u884c\u8f6f\u641c\u7d22\u4f5c\u4e3a softmax \u7136\u540e\u8fdb\u884c\u52a0\u6743\u5e73\u5747\u7684\u60f3\u6cd5\u3002\u8fd9\u4e2a\u65b9\u6cd5\u4ece\u4e00\u5f00\u59cb\u5c31\u8868\u73b0\u51fa\u8272\uff0c\u4ee4\u4eba\u632f\u594b \u3002\u6211\u79f0\u8fd9\u79cd\u67b6\u6784\u4e3a RNNSearch\uff0c\u6211\u4eec\u8d76\u7d27\u53d1\u5e03\u4e86\u4e00\u7bc7 arXiv \u8bba\u6587\uff0c\u56e0\u4e3a\u6211\u4eec\u77e5\u9053 Ilya \u548c\u8c37\u6b4c\u7684\u540c\u4e8b\u4eec\u5728\u4ed6\u4eec\u5f3a\u5927\u7684 8 \u4e2a GPU LSTM \u6a21\u578b\uff08RNN Search \u4ecd\u5728 1 \u4e2a GPU \u4e0a\u8fd0\u884c\uff09\u4e0a\u9886\u5148\u4e8e\u6211\u4eec\u3002\u4e8b\u540e\u8bc1\u660e\uff0c\u8fd9\u4e2a\u540d\u79f0\u4e0d\u592a\u597d\u3002\u66f4\u597d\u7684\u540d\u79f0\uff08attention\uff09\u662f\u7531 Yoshua \u5728\u6700\u7ec8\u5b9a\u7a3f\u4e2d\u6dfb\u52a0\u7684\u3002\u300d<\/p>\n<p>\u53e6\u4e00\u4f4d\u4f5c\u8005 Kyunghyun Cho \u662f\u7ebd\u7ea6\u5927\u5b66\u8ba1\u7b97\u673a\u79d1\u5b66\u548c\u6570\u636e\u79d1\u5b66\u6559\u6388\uff0c\u540c\u65f6\u62c5\u4efb Genentech Research &amp; Early Development (gRED) Prescient Design \u56e2\u961f\u7684\u524d\u6cbf\u7814\u7a76\u6267\u884c\u603b\u76d1\u30022025 \u5e74\uff0c\u4ed6\u88ab\u4efb\u547d\u4e3a Glen de Vries \u5065\u5eb7\u7edf\u8ba1\u5b66\u6559\u6388\u3002\u5728\u83b7\u5956\u8bba\u6587\u5199\u4f5c\u671f\u95f4\uff0c\u4ed6\u66fe\u5728\u8499\u7279\u5229\u5c14\u5927\u5b66\u62c5\u4efb\u535a\u58eb\u540e\u7814\u7a76\u5458\uff0c\u5e08\u4ece Yoshua Bengio \u6559\u6388\u3002<\/p>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-b2395c0250a372c79ad05bec99cded88.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-b2395c0250a372c79ad05bec99cded88.png\"><\/a><\/section>\n<p>\u5728 ICLR \u65f6\u95f4\u68c0\u9a8c\u5956\u9881\u53d1\u4e4b\u9645\uff0c\u6216\u8bb8\u6211\u4eec\u4e5f\u53ef\u4ee5\u91cd\u8bfb\u8fd9\u4e9b\u7ecf\u5178\u8bba\u6587\uff0c\u4ece\u4e2d\u627e\u5230\u5173\u4e8e\u672a\u6765\u65b9\u5411\u7684\u65b0\u542f\u53d1\u3002<\/p>\n<p><sup>\u53c2\u8003\u94fe\u63a5\uff1ahttps:\/\/blog.iclr.cc\/2025\/04\/14\/announcing-the-test-of-time-award-winners-from-iclr-2015\/<\/sup><\/p>\n<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2025-04-15-11\" target=\"_blank\">\u7ee7VAE\u4e4b\u540e\uff0cAdam\u4e5f\u62ff\u5230\u4e86ICLR \u65f6\u95f4\u68c0\u9a8c\u5956\uff0cOpenAI\u8054\u521b\u4e24\u6b21\u83b7\u5956<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:\u7ee7VAE\u4e4b\u540e\uff0cAdam [&hellip;]<\/p>\n","protected":false},"author":3,"featured_media":0,"comment_status":"open","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[27],"tags":[55,81,65,67,66],"class_list":["post-40187","post","type-post","status-publish","format-standard","hentry","category-news","tag-55","tag-81","tag-65","tag-67","tag-66"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v26.4 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>\u7ee7VAE\u4e4b\u540e\uff0cAdam\u4e5f\u62ff\u5230\u4e86ICLR \u65f6\u95f4\u68c0\u9a8c\u5956\uff0cOpenAI\u8054\u521b\u4e24\u6b21\u83b7\u5956 - \u4e00\u8d77AI\u6280\u672f<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/17aitech.com\/?p=40187\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/17aitech.com\/?p=40187\",\"url\":\"https:\/\/17aitech.com\/?p=40187\",\"name\":\"\u7ee7VAE\u4e4b\u540e\uff0cAdam\u4e5f\u62ff\u5230\u4e86ICLR \u65f6\u95f4\u68c0\u9a8c\u5956\uff0cOpenAI\u8054\u521b\u4e24\u6b21\u83b7\u5956 - \u4e00\u8d77AI\u6280\u672f\",\"isPartOf\":{\"@id\":\"https:\/\/17aitech.com\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/17aitech.com\/?p=40187#primaryimage\"},\"image\":{\"@id\":\"https:\/\/17aitech.com\/?p=40187#primaryimage\"},\"thumbnailUrl\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-37c819374825294a9541ddc5f8619755.png\",\"datePublished\":\"2025-04-17T00:02:49+00:00\",\"author\":{\"@id\":\"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d\"},\"breadcrumb\":{\"@id\":\"https:\/\/17aitech.com\/?p=40187#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/17aitech.com\/?p=40187\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/17aitech.com\/?p=40187#primaryimage\",\"url\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-37c819374825294a9541ddc5f8619755.png\",\"contentUrl\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-37c819374825294a9541ddc5f8619755.png\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/17aitech.com\/?p=40187#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/17aitech.com\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u7ee7VAE\u4e4b\u540e\uff0cAdam\u4e5f\u62ff\u5230\u4e86ICLR \u65f6\u95f4\u68c0\u9a8c\u5956\uff0cOpenAI\u8054\u521b\u4e24\u6b21\u83b7\u5956\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/17aitech.com\/#website\",\"url\":\"https:\/\/17aitech.com\/\",\"name\":\"\u4e00\u8d77AI\u6280\u672f\",\"description\":\"\u8ba9AI\u77e5\u8bc6\u89e6\u624b\u53ef\u53ca\",\"alternateName\":\"\u4e00\u8d77AI\u6280\u672f\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/17aitech.com\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d\",\"name\":\"AI\u5c0f\u52a9\u624b\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/17aitech.com\/#\/schema\/person\/image\/\",\"url\":\"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png\",\"contentUrl\":\"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png\",\"caption\":\"AI\u5c0f\u52a9\u624b\"},\"description\":\"\u8fd9\u4e2a\u4eba\u5f88\u61d2\uff0c\u4ec0\u4e48\u90fd\u6ca1\u6709\u7559\u4e0b\uff5e\",\"url\":\"https:\/\/17aitech.com\/?page_id=33738&user=3\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"\u7ee7VAE\u4e4b\u540e\uff0cAdam\u4e5f\u62ff\u5230\u4e86ICLR \u65f6\u95f4\u68c0\u9a8c\u5956\uff0cOpenAI\u8054\u521b\u4e24\u6b21\u83b7\u5956 - \u4e00\u8d77AI\u6280\u672f","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/17aitech.com\/?p=40187","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/17aitech.com\/?p=40187","url":"https:\/\/17aitech.com\/?p=40187","name":"\u7ee7VAE\u4e4b\u540e\uff0cAdam\u4e5f\u62ff\u5230\u4e86ICLR \u65f6\u95f4\u68c0\u9a8c\u5956\uff0cOpenAI\u8054\u521b\u4e24\u6b21\u83b7\u5956 - \u4e00\u8d77AI\u6280\u672f","isPartOf":{"@id":"https:\/\/17aitech.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/17aitech.com\/?p=40187#primaryimage"},"image":{"@id":"https:\/\/17aitech.com\/?p=40187#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-37c819374825294a9541ddc5f8619755.png","datePublished":"2025-04-17T00:02:49+00:00","author":{"@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"breadcrumb":{"@id":"https:\/\/17aitech.com\/?p=40187#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/17aitech.com\/?p=40187"]}]},{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/?p=40187#primaryimage","url":"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-37c819374825294a9541ddc5f8619755.png","contentUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/04\/frc-37c819374825294a9541ddc5f8619755.png"},{"@type":"BreadcrumbList","@id":"https:\/\/17aitech.com\/?p=40187#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/17aitech.com\/"},{"@type":"ListItem","position":2,"name":"\u7ee7VAE\u4e4b\u540e\uff0cAdam\u4e5f\u62ff\u5230\u4e86ICLR \u65f6\u95f4\u68c0\u9a8c\u5956\uff0cOpenAI\u8054\u521b\u4e24\u6b21\u83b7\u5956"}]},{"@type":"WebSite","@id":"https:\/\/17aitech.com\/#website","url":"https:\/\/17aitech.com\/","name":"\u4e00\u8d77AI\u6280\u672f","description":"\u8ba9AI\u77e5\u8bc6\u89e6\u624b\u53ef\u53ca","alternateName":"\u4e00\u8d77AI\u6280\u672f","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/17aitech.com\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d","name":"AI\u5c0f\u52a9\u624b","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/#\/schema\/person\/image\/","url":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","contentUrl":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","caption":"AI\u5c0f\u52a9\u624b"},"description":"\u8fd9\u4e2a\u4eba\u5f88\u61d2\uff0c\u4ec0\u4e48\u90fd\u6ca1\u6709\u7559\u4e0b\uff5e","url":"https:\/\/17aitech.com\/?page_id=33738&user=3"}]}},"_links":{"self":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/40187","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=40187"}],"version-history":[{"count":0,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/40187\/revisions"}],"wp:attachment":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=40187"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=40187"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=40187"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}