{"id":479499,"date":"2023-08-09T10:40:54","date_gmt":"2023-08-09T10:40:54","guid":{"rendered":""},"modified":"2023-09-05T11:18:57","modified_gmt":"2023-09-05T11:18:57","slug":"variational-autoencoders","status":"publish","type":"wiki","link":"https:\/\/oneproxy.pro\/cn\/wiki\/variational-autoencoders\/","title":{"rendered":"\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668"},"content":{"rendered":"<p>\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668 (VAE) \u662f\u4e00\u7c7b\u5c5e\u4e8e\u81ea\u52a8\u7f16\u7801\u5668\u5bb6\u65cf\u7684\u751f\u6210\u6a21\u578b\u3002\u5b83\u4eec\u662f\u65e0\u76d1\u7763\u5b66\u4e60\u7684\u5f3a\u5927\u5de5\u5177\uff0c\u5728\u673a\u5668\u5b66\u4e60\u548c\u4eba\u5de5\u667a\u80fd\u9886\u57df\u5f15\u8d77\u4e86\u6781\u5927\u5173\u6ce8\u3002VAE \u80fd\u591f\u5b66\u4e60\u590d\u6742\u6570\u636e\u7684\u4f4e\u7ef4\u8868\u793a\uff0c\u7279\u522b\u9002\u7528\u4e8e\u6570\u636e\u538b\u7f29\u3001\u56fe\u50cf\u751f\u6210\u548c\u5f02\u5e38\u68c0\u6d4b\u7b49\u4efb\u52a1\u3002<\/p>\n<h2>\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\u7684\u8d77\u6e90\u5386\u53f2\u53ca\u5176\u9996\u6b21\u63d0\u53ca<\/h2>\n<p>\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\u6700\u521d\u7531 Kingma \u548c Welling \u4e8e 2013 \u5e74\u63d0\u51fa\u3002\u5728\u4ed6\u4eec\u7684\u5f00\u521b\u6027\u8bba\u6587\u300a\u81ea\u52a8\u7f16\u7801\u53d8\u5206\u8d1d\u53f6\u65af\u300b\u4e2d\uff0c\u4ed6\u4eec\u63d0\u51fa\u4e86 VAE \u7684\u6982\u5ff5\uff0c\u4f5c\u4e3a\u4f20\u7edf\u81ea\u52a8\u7f16\u7801\u5668\u7684\u6982\u7387\u6269\u5c55\u3002\u8be5\u6a21\u578b\u7ed3\u5408\u4e86\u53d8\u5206\u63a8\u7406\u548c\u81ea\u52a8\u7f16\u7801\u5668\u7684\u601d\u60f3\uff0c\u63d0\u4f9b\u4e86\u4e00\u4e2a\u7528\u4e8e\u5b66\u4e60\u6570\u636e\u7684\u6982\u7387\u6f5c\u5728\u8868\u793a\u7684\u6846\u67b6\u3002<\/p>\n<h2>\u5173\u4e8e\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\u7684\u8be6\u7ec6\u4fe1\u606f<\/h2>\n<h3>\u6269\u5c55\u4e3b\u9898\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668<\/h3>\n<p>\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\u7684\u5de5\u4f5c\u539f\u7406\u662f\u5c06\u8f93\u5165\u6570\u636e\u7f16\u7801\u4e3a\u6f5c\u5728\u7a7a\u95f4\u8868\u793a\uff0c\u7136\u540e\u5c06\u5176\u89e3\u7801\u56de\u539f\u59cb\u6570\u636e\u7a7a\u95f4\u3002VAE \u80cc\u540e\u7684\u6838\u5fc3\u601d\u60f3\u662f\u5b66\u4e60\u6f5c\u5728\u7a7a\u95f4\u4e2d\u6570\u636e\u7684\u5e95\u5c42\u6982\u7387\u5206\u5e03\uff0c\u4ece\u800c\u53ef\u4ee5\u901a\u8fc7\u4ece\u5b66\u4e60\u5230\u7684\u5206\u5e03\u4e2d\u91c7\u6837\u6765\u751f\u6210\u65b0\u7684\u6570\u636e\u70b9\u3002\u8fd9\u4e00\u7279\u6027\u4f7f VAE \u6210\u4e3a\u5f3a\u5927\u7684\u751f\u6210\u6a21\u578b\u3002<\/p>\n<h2>\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\u7684\u5185\u90e8\u7ed3\u6784<\/h2>\n<h3>\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\u7684\u5de5\u4f5c\u539f\u7406<\/h3>\n<p>VAE \u7684\u67b6\u6784\u7531\u4e24\u4e2a\u4e3b\u8981\u7ec4\u4ef6\u7ec4\u6210\uff1a\u7f16\u7801\u5668\u548c\u89e3\u7801\u5668\u3002<\/p>\n<ol>\n<li>\n<p>\u7f16\u7801\u5668\uff1a\u7f16\u7801\u5668\u83b7\u53d6\u8f93\u5165\u6570\u636e\u70b9\u5e76\u5c06\u5176\u6620\u5c04\u5230\u6f5c\u5728\u7a7a\u95f4\uff0c\u5728\u8be5\u7a7a\u95f4\u4e2d\u5c06\u5176\u8868\u793a\u4e3a\u5747\u503c\u5411\u91cf\u548c\u65b9\u5dee\u5411\u91cf\u3002\u8fd9\u4e9b\u5411\u91cf\u5b9a\u4e49\u4e86\u6f5c\u5728\u7a7a\u95f4\u4e2d\u7684\u6982\u7387\u5206\u5e03\u3002<\/p>\n<\/li>\n<li>\n<p>\u91cd\u65b0\u53c2\u6570\u5316\u6280\u5de7\uff1a\u4e3a\u4e86\u5b9e\u73b0\u53cd\u5411\u4f20\u64ad\u548c\u9ad8\u6548\u8bad\u7ec3\uff0c\u4f7f\u7528\u4e86\u91cd\u65b0\u53c2\u6570\u5316\u6280\u5de7\u3002\u8be5\u6a21\u578b\u4e0d\u662f\u76f4\u63a5\u4ece\u6f5c\u5728\u7a7a\u95f4\u4e2d\u7684\u5b66\u4e60\u5206\u5e03\u4e2d\u91c7\u6837\uff0c\u800c\u662f\u4ece\u6807\u51c6\u9ad8\u65af\u5206\u5e03\u4e2d\u91c7\u6837\uff0c\u5e76\u4f7f\u7528\u4ece\u7f16\u7801\u5668\u83b7\u5f97\u7684\u5747\u503c\u548c\u65b9\u5dee\u5411\u91cf\u7f29\u653e\u548c\u79fb\u52a8\u6837\u672c\u3002<\/p>\n<\/li>\n<li>\n<p>\u89e3\u7801\u5668\uff1a\u89e3\u7801\u5668\u91c7\u7528\u91c7\u6837\u7684\u6f5c\u5728\u5411\u91cf\u5e76\u4ece\u4e2d\u91cd\u5efa\u539f\u59cb\u6570\u636e\u70b9\u3002<\/p>\n<\/li>\n<\/ol>\n<p>VAE \u7684\u76ee\u6807\u51fd\u6570\u5305\u62ec\u4e24\u4e2a\u4e3b\u8981\u9879\uff1a\u91cd\u5efa\u635f\u5931\uff08\u8861\u91cf\u91cd\u5efa\u7684\u8d28\u91cf\uff09\u548c KL \u6563\u5ea6\uff08\u9f13\u52b1\u5b66\u4e60\u5230\u7684\u6f5c\u5728\u5206\u5e03\u63a5\u8fd1\u6807\u51c6\u9ad8\u65af\u5206\u5e03\uff09\u3002<\/p>\n<h2>\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\u7684\u5173\u952e\u7279\u5f81\u5206\u6790<\/h2>\n<ul>\n<li>\n<p><strong>\u751f\u6210\u80fd\u529b<\/strong>\uff1aVAE \u53ef\u4ee5\u901a\u8fc7\u4ece\u5b66\u4e60\u5230\u7684\u6f5c\u5728\u7a7a\u95f4\u5206\u5e03\u4e2d\u91c7\u6837\u6765\u751f\u6210\u65b0\u7684\u6570\u636e\u70b9\uff0c\u4f7f\u5176\u53ef\u7528\u4e8e\u5404\u79cd\u751f\u6210\u4efb\u52a1\u3002<\/p>\n<\/li>\n<li>\n<p><strong>\u6982\u7387\u89e3\u91ca<\/strong>\uff1aVAE \u63d0\u4f9b\u6570\u636e\u7684\u6982\u7387\u89e3\u91ca\uff0c\u4ece\u800c\u80fd\u591f\u4f30\u8ba1\u4e0d\u786e\u5b9a\u6027\u5e76\u66f4\u597d\u5730\u5904\u7406\u7f3a\u5931\u6216\u566a\u58f0\u6570\u636e\u3002<\/p>\n<\/li>\n<li>\n<p><strong>\u7d27\u51d1\u6f5c\u5728\u8868\u5f81<\/strong>\uff1aVAE \u5b66\u4e60\u6570\u636e\u7684\u7d27\u51d1\u4e14\u8fde\u7eed\u7684\u6f5c\u5728\u8868\u793a\uff0c\u4ece\u800c\u5141\u8bb8\u6570\u636e\u70b9\u4e4b\u95f4\u8fdb\u884c\u5e73\u6ed1\u63d2\u503c\u3002<\/p>\n<\/li>\n<\/ul>\n<h2>\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\u7684\u7c7b\u578b<\/h2>\n<p>VAE \u53ef\u4ee5\u901a\u8fc7\u591a\u79cd\u65b9\u5f0f\u8fdb\u884c\u8c03\u6574\u548c\u6269\u5c55\uff0c\u4ee5\u9002\u5e94\u4e0d\u540c\u7c7b\u578b\u7684\u6570\u636e\u548c\u5e94\u7528\u7a0b\u5e8f\u3002\u4e00\u4e9b\u5e38\u89c1\u7684 VAE \u7c7b\u578b\u5305\u62ec\uff1a<\/p>\n<ol>\n<li>\n<p><strong>\u6761\u4ef6\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\uff08CVAE\uff09<\/strong>\uff1a\u8fd9\u4e9b\u6a21\u578b\u53ef\u4ee5\u6839\u636e\u5176\u4ed6\u8f93\u5165\uff08\u4f8b\u5982\u7c7b\u6807\u7b7e\u6216\u8f85\u52a9\u7279\u5f81\uff09\u6765\u8c03\u8282\u6570\u636e\u7684\u751f\u6210\u3002CVAE \u5bf9\u4e8e\u6761\u4ef6\u56fe\u50cf\u751f\u6210\u7b49\u4efb\u52a1\u975e\u5e38\u6709\u7528\u3002<\/p>\n<\/li>\n<li>\n<p><strong>\u5bf9\u6297\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\uff08AVAE\uff09<\/strong>\uff1aAVAE \u5c06 VAE \u4e0e\u751f\u6210\u5bf9\u6297\u7f51\u7edc (GAN) \u76f8\u7ed3\u5408\uff0c\u4ee5\u63d0\u9ad8\u751f\u6210\u6570\u636e\u7684\u8d28\u91cf\u3002<\/p>\n<\/li>\n<li>\n<p><strong>\u89e3\u7f20\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668<\/strong>\uff1a\u8fd9\u4e9b\u6a21\u578b\u65e8\u5728\u5b66\u4e60\u89e3\u5f00\u7684\u8868\u793a\uff0c\u5176\u4e2d\u6f5c\u5728\u7a7a\u95f4\u7684\u6bcf\u4e2a\u7ef4\u5ea6\u5bf9\u5e94\u4e8e\u6570\u636e\u7684\u7279\u5b9a\u7279\u5f81\u6216\u5c5e\u6027\u3002<\/p>\n<\/li>\n<li>\n<p><strong>\u534a\u76d1\u7763\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668<\/strong>\uff1aVAE \u53ef\u4ee5\u6269\u5c55\u5230\u5904\u7406\u534a\u76d1\u7763\u5b66\u4e60\u4efb\u52a1\uff0c\u5176\u4e2d\u53ea\u6709\u4e00\u5c0f\u90e8\u5206\u6570\u636e\u88ab\u6807\u8bb0\u3002<\/p>\n<\/li>\n<\/ol>\n<h2>\u53d8\u5206\u81ea\u7f16\u7801\u5668\u7684\u4f7f\u7528\u65b9\u6cd5\u3001\u95ee\u9898\u53ca\u89e3\u51b3\u65b9\u6cd5<\/h2>\n<p>\u7531\u4e8e\u5176\u751f\u6210\u80fd\u529b\u548c\u7d27\u51d1\u7684\u6f5c\u5728\u8868\u793a\uff0cVAE \u53ef\u4ee5\u5728\u5404\u4e2a\u9886\u57df\u5f97\u5230\u5e94\u7528\u3002\u4e00\u4e9b\u5e38\u89c1\u7528\u4f8b\u5305\u62ec\uff1a<\/p>\n<ol>\n<li>\n<p><strong>\u6570\u636e\u538b\u7f29<\/strong>\uff1aVAE \u53ef\u7528\u4e8e\u538b\u7f29\u6570\u636e\uff0c\u540c\u65f6\u4fdd\u7559\u5176\u57fa\u672c\u7279\u5f81\u3002<\/p>\n<\/li>\n<li>\n<p><strong>\u56fe\u50cf\u751f\u6210<\/strong>\uff1aVAE \u53ef\u4ee5\u751f\u6210\u65b0\u7684\u56fe\u50cf\uff0c\u4f7f\u5176\u5bf9\u4e8e\u521b\u610f\u5e94\u7528\u548c\u6570\u636e\u589e\u5f3a\u5177\u6709\u4ef7\u503c\u3002<\/p>\n<\/li>\n<li>\n<p><strong>\u5f02\u5e38\u68c0\u6d4b<\/strong>\uff1a\u5bf9\u5e95\u5c42\u6570\u636e\u5206\u5e03\u8fdb\u884c\u5efa\u6a21\u7684\u80fd\u529b\u4f7f VAE \u80fd\u591f\u68c0\u6d4b\u6570\u636e\u96c6\u4e2d\u7684\u5f02\u5e38\u6216\u5f02\u5e38\u503c\u3002<\/p>\n<\/li>\n<\/ol>\n<p>\u4f7f\u7528VAE\u76f8\u5173\u7684\u6311\u6218\u548c\u89e3\u51b3\u65b9\u6848\uff1a<\/p>\n<ul>\n<li>\n<p><strong>\u6a21\u5f0f\u5d29\u6e83<\/strong>\uff1a\u5728\u67d0\u4e9b\u60c5\u51b5\u4e0b\uff0c\u7531\u4e8e\u6a21\u5f0f\u5d29\u6e83\uff0cVAE \u53ef\u80fd\u4f1a\u4ea7\u751f\u6a21\u7cca\u6216\u4e0d\u5207\u5b9e\u9645\u7684\u6837\u672c\u3002\u7814\u7a76\u4eba\u5458\u63d0\u51fa\u4e86\u9000\u706b\u8bad\u7ec3\u548c\u6539\u8fdb\u67b6\u6784\u7b49\u6280\u672f\u6765\u89e3\u51b3\u6b64\u95ee\u9898\u3002<\/p>\n<\/li>\n<li>\n<p><strong>\u6f5c\u5728\u7a7a\u95f4\u53ef\u89e3\u91ca\u6027<\/strong>\uff1a\u89e3\u91ca VAE \u7684\u6f5c\u5728\u7a7a\u95f4\u53ef\u80fd\u5177\u6709\u6311\u6218\u6027\u3002\u89e3\u5f00 VAE \u548c\u53ef\u89c6\u5316\u6280\u672f\u6709\u52a9\u4e8e\u5b9e\u73b0\u66f4\u597d\u7684\u53ef\u89e3\u91ca\u6027\u3002<\/p>\n<\/li>\n<\/ul>\n<h2>\u4e3b\u8981\u7279\u70b9\u53ca\u4e0e\u540c\u7c7b\u672f\u8bed\u7684\u5176\u4ed6\u6bd4\u8f83<\/h2>\n<table>\n<thead>\n<tr>\n<th><strong>\u7279\u5f81<\/strong><\/th>\n<th><strong>\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668 (VAE)<\/strong><\/th>\n<th><strong>\u81ea\u52a8\u7f16\u7801\u5668<\/strong><\/th>\n<th><strong>\u751f\u6210\u5bf9\u6297\u7f51\u7edc (GAN)<\/strong><\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><strong>\u751f\u6210\u6a21\u578b<\/strong><\/td>\n<td>\u662f\u7684<\/td>\n<td>\u4e0d<\/td>\n<td>\u662f\u7684<\/td>\n<\/tr>\n<tr>\n<td><strong>\u6f5c\u5728\u7a7a\u95f4<\/strong><\/td>\n<td>\u8fde\u7eed\u548c\u6982\u7387<\/td>\n<td>\u8fde\u7eed\u7684<\/td>\n<td>\u968f\u673a\u566a\u58f0<\/td>\n<\/tr>\n<tr>\n<td><strong>\u57f9\u8bad\u76ee\u6807<\/strong><\/td>\n<td>\u91cd\u6784 + KL \u6563\u5ea6<\/td>\n<td>\u91cd\u5efa<\/td>\n<td>\u6700\u5c0f\u6700\u5927\u535a\u5f08<\/td>\n<\/tr>\n<tr>\n<td><strong>\u4e0d\u786e\u5b9a\u6027\u4f30\u8ba1<\/strong><\/td>\n<td>\u662f\u7684<\/td>\n<td>\u4e0d<\/td>\n<td>\u4e0d<\/td>\n<\/tr>\n<tr>\n<td><strong>\u5904\u7406\u7f3a\u5931\u6570\u636e<\/strong><\/td>\n<td>\u66f4\u597d\u7684<\/td>\n<td>\u96be\u7684<\/td>\n<td>\u96be\u7684<\/td>\n<\/tr>\n<tr>\n<td><strong>\u6f5c\u5728\u7a7a\u95f4\u7684\u53ef\u89e3\u91ca\u6027<\/strong><\/td>\n<td>\u7f13\u548c<\/td>\n<td>\u96be\u7684<\/td>\n<td>\u96be\u7684<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h2>\u4e0e\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\u76f8\u5173\u7684\u672a\u6765\u89c2\u70b9\u548c\u6280\u672f<\/h2>\n<p>\u53d8\u5206\u81ea\u7f16\u7801\u5668\u7684\u672a\u6765\u524d\u666f\u5149\u660e\uff0c\u6b63\u5728\u8fdb\u884c\u7684\u7814\u7a76\u91cd\u70b9\u662f\u589e\u5f3a\u5176\u529f\u80fd\u548c\u5e94\u7528\u3002\u4e00\u4e9b\u5173\u952e\u65b9\u5411\u5305\u62ec\uff1a<\/p>\n<ul>\n<li>\n<p><strong>\u6539\u8fdb\u7684\u751f\u6210\u6a21\u578b<\/strong>\uff1a\u7814\u7a76\u4eba\u5458\u6b63\u5728\u81f4\u529b\u4e8e\u6539\u8fdb VAE \u67b6\u6784\uff0c\u4ee5\u751f\u6210\u66f4\u9ad8\u8d28\u91cf\u3001\u66f4\u591a\u6837\u5316\u7684\u6837\u672c\u3002<\/p>\n<\/li>\n<li>\n<p><strong>\u89e3\u5f00\u8868\u5f81<\/strong>\uff1a\u5b66\u4e60\u89e3\u5f00\u8868\u5f81\u7684\u8fdb\u6b65\u5c06\u4f7f\u6211\u4eec\u80fd\u591f\u66f4\u597d\u5730\u63a7\u5236\u548c\u7406\u89e3\u751f\u6210\u8fc7\u7a0b\u3002<\/p>\n<\/li>\n<li>\n<p><strong>\u6df7\u5408\u6a21\u578b<\/strong>\uff1a\u5c06 VAE \u4e0e GAN \u7b49\u5176\u4ed6\u751f\u6210\u6a21\u578b\u76f8\u7ed3\u5408\uff0c\u53ef\u4ee5\u4ea7\u751f\u5177\u6709\u589e\u5f3a\u6027\u80fd\u7684\u65b0\u578b\u751f\u6210\u6a21\u578b\u3002<\/p>\n<\/li>\n<\/ul>\n<h2>\u4ee3\u7406\u670d\u52a1\u5668\u5982\u4f55\u4e0e\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\u914d\u5408\u4f7f\u7528<\/h2>\n<p>\u5728\u67d0\u4e9b\u60c5\u51b5\u4e0b\uff0c\u4ee3\u7406\u670d\u52a1\u5668\u53ef\u4ee5\u95f4\u63a5\u4e0e\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\u76f8\u5173\u8054\u3002VAE \u5728\u6570\u636e\u538b\u7f29\u548c\u56fe\u50cf\u751f\u6210\u4e2d\u5f97\u5230\u5e94\u7528\uff0c\u5176\u4e2d\u4ee3\u7406\u670d\u52a1\u5668\u53ef\u4ee5\u5728\u4f18\u5316\u6570\u636e\u4f20\u8f93\u548c\u7f13\u5b58\u65b9\u9762\u53d1\u6325\u4f5c\u7528\u3002\u4f8b\u5982\uff1a<\/p>\n<ol>\n<li>\n<p><strong>\u6570\u636e\u538b\u7f29\u548c\u89e3\u538b\u7f29<\/strong>\uff1a\u4ee3\u7406\u670d\u52a1\u5668\u53ef\u4ee5\u4f7f\u7528 VAE \u8fdb\u884c\u9ad8\u6548\u6570\u636e\u538b\u7f29\uff0c\u7136\u540e\u518d\u5c06\u5176\u4f20\u8f93\u7ed9\u5ba2\u6237\u7aef\u3002\u540c\u6837\uff0c\u5ba2\u6237\u7aef\u4e5f\u53ef\u4ee5\u4f7f\u7528 VAE \u6765\u89e3\u538b\u63a5\u6536\u5230\u7684\u6570\u636e\u3002<\/p>\n<\/li>\n<li>\n<p><strong>\u7f13\u5b58\u548c\u56fe\u50cf\u751f\u6210<\/strong>\uff1a\u5728\u5185\u5bb9\u5206\u53d1\u7f51\u7edc\u4e2d\uff0c\u4ee3\u7406\u670d\u52a1\u5668\u53ef\u4ee5\u5229\u7528 VAE \u9884\u5148\u751f\u6210\u7684\u56fe\u50cf\u6765\u5feb\u901f\u63d0\u4f9b\u7f13\u5b58\u5185\u5bb9\u3002<\/p>\n<\/li>\n<\/ol>\n<p>\u503c\u5f97\u6ce8\u610f\u7684\u662f\uff0cVAE \u548c\u4ee3\u7406\u670d\u52a1\u5668\u662f\u72ec\u7acb\u7684\u6280\u672f\uff0c\u4f46\u5b83\u4eec\u53ef\u4ee5\u4e00\u8d77\u4f7f\u7528\u4ee5\u6539\u5584\u7279\u5b9a\u5e94\u7528\u7a0b\u5e8f\u4e2d\u7684\u6570\u636e\u5904\u7406\u548c\u4f20\u9001\u3002<\/p>\n<h2>\u76f8\u5173\u94fe\u63a5<\/h2>\n<p>\u6709\u5173\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\u7684\u66f4\u591a\u4fe1\u606f\uff0c\u8bf7\u53c2\u8003\u4ee5\u4e0b\u8d44\u6e90\uff1a<\/p>\n<ol>\n<li>\n<p>\u201c\u81ea\u52a8\u7f16\u7801\u53d8\u5206\u8d1d\u53f6\u65af\u201d\u2014\u2014Diederik P. Kingma\u3001Max Welling\u3002 <a href=\"https:\/\/arxiv.org\/abs\/1312.6114\" target=\"_new\" rel=\"noopener nofollow\">https:\/\/arxiv.org\/abs\/1312.6114<\/a><\/p>\n<\/li>\n<li>\n<p>\u201c\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\u6559\u7a0b\u201d\u2014\u2014Carl Doersch\u3002 <a href=\"https:\/\/arxiv.org\/abs\/1606.05908\" target=\"_new\" rel=\"noopener nofollow\">https:\/\/arxiv.org\/abs\/1606.05908<\/a><\/p>\n<\/li>\n<li>\n<p>\u201c\u4e86\u89e3\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668 (VAE)\u201d \u2013 Janardhan Rao Doppa \u7684\u535a\u5ba2\u6587\u7ae0\u3002 <a href=\"https:\/\/towardsdatascience.com\/understanding-variational-autoencoders-vaes-f70510919f73\" target=\"_new\" rel=\"noopener nofollow\">https:\/\/towardsdatascience.com\/understanding-variational-autoencoders-vaes-f70510919f73<\/a><\/p>\n<\/li>\n<li>\n<p>\u201c\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668 (VAE) \u7684\u751f\u6210\u6a21\u578b\u7b80\u4ecb\u201d \u2013 Jie Fu \u7684\u535a\u5ba2\u6587\u7ae0\u3002 <a href=\"https:\/\/towardsdatascience.com\/introduction-to-generative-models-with-variational-autoencoders-vae-and-adversarial-177e1b1a4497\" target=\"_new\" rel=\"noopener nofollow\">https:\/\/towardsdatascience.com\/introduction-to-generative-models-with-variational-autoencoders-vae-and-adversarial-177e1b1a4497<\/a><\/p>\n<\/li>\n<\/ol>\n<p>\u901a\u8fc7\u63a2\u7d22\u8fd9\u4e9b\u8d44\u6e90\uff0c\u60a8\u53ef\u4ee5\u66f4\u6df1\u5165\u5730\u4e86\u89e3\u53d8\u5206\u81ea\u52a8\u7f16\u7801\u5668\u53ca\u5176\u5728\u673a\u5668\u5b66\u4e60\u9886\u57df\u53ca\u5176\u4ed6\u9886\u57df\u7684\u5404\u79cd\u5e94\u7528\u3002<\/p>","protected":false},"featured_media":470813,"menu_order":0,"template":"","meta":{"_acf_changed":false,"content-type":"","inline_featured_image":false,"footnotes":""},"class_list":["post-479499","wiki","type-wiki","status-publish","has-post-thumbnail","hentry"],"acf":{"faq_title":"Frequently Asked Questions about <mark>Variational Autoencoders<\/mark>","faq_items":[{"question":"What are Variational Autoencoders (VAEs)?","answer":"<p>Variational Autoencoders (VAEs) are a class of generative models that can learn a compact representation of complex data. They are particularly useful for tasks like data compression, image generation, and anomaly detection.<\/p>"},{"question":"How do Variational Autoencoders work?","answer":"<p>VAEs consist of two main components: the encoder and the decoder. The encoder maps input data to a latent space representation, while the decoder reconstructs the original data from the latent representation. VAEs use probabilistic inference and a reparameterization trick to enable efficient training and generative capabilities.<\/p>"},{"question":"What makes Variational Autoencoders unique?","answer":"<p>VAEs offer a probabilistic interpretation of data, allowing for uncertainty estimation and better handling of missing or noisy data. Their generative capability enables them to generate new data points by sampling from the learned latent space distribution.<\/p>"},{"question":"What types of Variational Autoencoders exist?","answer":"<p>Several types of VAEs cater to different applications. Conditional VAEs (CVAE) can condition data generation on additional inputs, while disentangled VAEs aim to learn interpretable and disentangled representations. Semi-supervised VAEs handle tasks with limited labeled data, and adversarial VAEs combine VAEs with Generative Adversarial Networks (GANs) for improved data generation.<\/p>"},{"question":"How are Variational Autoencoders used in practice?","answer":"<p>VAEs find applications in various domains. They are used for data compression, image generation, and anomaly detection. Additionally, VAEs can help improve data transmission and caching in proxy servers, enhancing content delivery network performance.<\/p>"},{"question":"What are the challenges associated with Variational Autoencoders?","answer":"<p>VAEs may encounter mode collapse, resulting in blurry or unrealistic samples. Interpreting the latent space can also be challenging. Researchers are continuously working on improved architectures and disentangled representations to address these challenges.<\/p>"},{"question":"What does the future hold for Variational Autoencoders?","answer":"<p>The future of VAEs looks promising, with ongoing research focusing on improving generative models, disentangled representations, and hybrid models. These advancements will unlock new possibilities in creative applications and data handling.<\/p>"},{"question":"How can proxy servers collaborate with Variational Autoencoders?","answer":"<p>Proxy servers can indirectly collaborate with VAEs in data compression and decompression for efficient data transmission. Additionally, VAE-generated images can be cached to enhance content delivery in proxy servers and content delivery networks.<\/p>"}]},"_links":{"self":[{"href":"https:\/\/oneproxy.pro\/cn\/wp-json\/wp\/v2\/wiki\/479499","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/oneproxy.pro\/cn\/wp-json\/wp\/v2\/wiki"}],"about":[{"href":"https:\/\/oneproxy.pro\/cn\/wp-json\/wp\/v2\/types\/wiki"}],"version-history":[{"count":0,"href":"https:\/\/oneproxy.pro\/cn\/wp-json\/wp\/v2\/wiki\/479499\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/oneproxy.pro\/cn\/wp-json\/wp\/v2\/media\/470813"}],"wp:attachment":[{"href":"https:\/\/oneproxy.pro\/cn\/wp-json\/wp\/v2\/media?parent=479499"}],"curies":[{"name":"\u53ef\u6e7f\u6027\u7c89\u5242","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}