{"id":7013,"date":"2022-09-08T15:44:45","date_gmt":"2022-09-08T07:44:45","guid":{"rendered":"http:\/\/139.9.1.231\/?p=7013"},"modified":"2022-09-08T15:52:23","modified_gmt":"2022-09-08T07:52:23","slug":"ssl_sum","status":"publish","type":"post","link":"http:\/\/139.9.1.231\/index.php\/2022\/09\/08\/ssl_sum\/","title":{"rendered":"\u534a\u76d1\u7763\u5b66\u4e60\u7efc\u8ff0"},"content":{"rendered":"\n<p>    \u534a\u76d1\u7763\u5b66\u4e60(Semi-Supervised Learning\uff0cSSL) \u4f7f\u7528\u6807\u8bb0\u548c\u672a\u6807\u8bb0\u7684\u6570\u636e\u6765\u6267\u884c\u6709\u76d1\u7763\u7684\u5b66\u4e60\u6216\u65e0\u76d1\u7763\u7684\u5b66\u4e60\u4efb\u52a1\u3002<\/p>\n\n\n\n<p>    \u534a\u76d1\u7763\u5b66\u4e60\u53ef\u8fdb\u4e00\u6b65\u5212\u5206\u4e3a<strong>\u7eaf\uff08pure\uff09\u534a\u76d1\u7763\u5b66\u4e60<\/strong>\u548c<strong>\u76f4\u63a8\u5b66\u4e60\uff08transductive learning\uff09<\/strong>\u3002\u524d\u8005\u5047\u5b9a\u8bad\u7ec3\u6570\u636e\u4e2d\u7684\u672a\u6807\u8bb0\u6837\u672c\u5e76\u975e\u5f85\u9884\u6d4b\u7684\u6570\u636e\uff0c\u800c\u540e\u8005\u5219\u5047\u5b9a\u5b66\u4e60\u8fc7\u7a0b\u4e2d\u6240\u8003\u8651\u7684\u672a\u6807\u8bb0\u6837\u672c\u6070\u662f\u5f85\u9884\u6d4b\u6570\u636e\u3002\u7eaf\u534a\u76d1\u7763\u5b66\u4e60\u662f\u57fa\u4e8e\u201c\u5f00\u653e\u4e16\u754c\u201d\u5047\u8bbe\uff0c\u5e0c\u671b\u5b66\u5f97\u6a21\u578b\u80fd\u9002\u7528\u4e8e\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u672a\u89c2\u5bdf\u5230\u7684\u6570\u636e\uff0c\u800c\u76f4\u63a8\u5b66\u4e60\u662f\u57fa\u4e8e\u201c\u5c01\u95ed\u4e16\u754c\u201d\u5047\u8bbe\uff0c\u4ec5\u8bd5\u56fe\u5bf9\u5b66\u4e60\u8fc7\u7a0b\u4e2d\u89c2\u5bdf\u5230\u7684\u672a\u6807\u8bb0\u6570\u636e\u8fdb\u884c\u9884\u6d4b\u3002\u4e0b\u56fe\u76f4\u89c2\u7684\u8868\u73b0\u51fa<strong>\u4e3b\u52a8\u5b66\u4e60<\/strong>\u3001<strong>\u7eaf\u534a\u76d1\u7763\u5b66\u4e60<\/strong>\u3001<strong>\u76f4\u63a8\u5b66\u4e60<\/strong>\u7684\u533a\u522b\uff1a<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" width=\"932\" height=\"751\" src=\"http:\/\/139.9.1.231\/wp-content\/uploads\/2022\/09\/image-112.png\" alt=\"\" class=\"wp-image-7618\" srcset=\"http:\/\/139.9.1.231\/wp-content\/uploads\/2022\/09\/image-112.png 932w, http:\/\/139.9.1.231\/wp-content\/uploads\/2022\/09\/image-112-300x242.png 300w, http:\/\/139.9.1.231\/wp-content\/uploads\/2022\/09\/image-112-768x619.png 768w\" sizes=\"(max-width: 932px) 100vw, 932px\" \/><\/figure>\n\n\n\n<p>   \u867d\u7136\u8bad\u7ec3\u6570\u636e\u4e2d\u542b\u6709\u5927\u91cf\u65e0\u6807\u7b7e\u6570\u636e\uff0c\u4f46\u5176\u5b9e\u5728\u5f88\u591a\u534a\u76d1\u7763\u5b66\u4e60\u7b97\u6cd5\u4e2d\u7528\u7684\u8bad\u7ec3\u6570\u636e\u8fd8\u6709\u633a\u591a\u8981\u6c42\u7684\uff0c\u4e00\u822c\u9ed8\u8ba4\u7684\u6709\uff1a\u65e0\u6807\u7b7e\u6570\u636e\u4e00\u822c\u662f\u6709\u6807\u7b7e\u6570\u636e\u4e2d\u7684\u67d0\u4e00\u4e2a\u7c7b\u522b\u7684\uff08\u4e0d\u8981\u4e0d\u5c5e\u4e8e\u7684\uff0c\u4e5f\u4e0d\u8981\u5c5e\u4e8e\u591a\u4e2a\u7c7b\u522b\u7684\uff09\uff1b\u6709\u6807\u7b7e\u6570\u636e\u7684\u6807\u7b7e\u5e94\u8be5\u90fd\u662f\u5bf9\u7684\uff1b\u65e0\u6807\u7b7e\u6570\u636e\u4e00\u822c\u662f\u7c7b\u522b\u5e73\u8861\u7684\uff08\u5373\u6bcf\u4e00\u7c7b\u7684\u6837\u672c\u6570\u5dee\u4e0d\u591a\uff09\uff1b\u65e0\u6807\u7b7e\u6570\u636e\u7684\u5206\u5e03\u5e94\u8be5\u548c\u6709\u6807\u7b7e\u7684\u76f8\u540c\u6216\u7c7b\u4f3c\u7b49\u7b49\u3002<\/p>\n\n\n\n<p class=\"has-bright-blue-background-color has-background\">   <strong>\u4e00\u822c\uff0c\u534a\u76d1\u7763\u5b66\u4e60\u7b97\u6cd5\u53ef\u5206\u4e3a\uff1aself-training\uff08\u81ea\u8bad\u7ec3\u7b97\u6cd5\uff09\u3001Graph-based Semi-supervised Learning\uff08\u57fa\u4e8e\u56fe\u7684\u534a\u76d1\u7763\u7b97\u6cd5\uff09\u3001Semi-supervised supported vector machine\uff08\u534a\u76d1\u7763\u652f\u6301\u5411\u91cf\u673a\uff0cS3VM\uff09\u3002\u7b80\u5355\u4ecb\u7ecd\u5982\u4e0b\uff1a<\/strong><\/p>\n\n\n\n<p>1.<strong>\u7b80\u5355\u81ea\u8bad\u7ec3<\/strong>\uff08simple self-training\uff09\uff1a\u7528\u6709\u6807\u7b7e\u6570\u636e\u8bad\u7ec3\u4e00\u4e2a\u5206\u7c7b\u5668\uff0c\u7136\u540e\u7528\u8fd9\u4e2a\u5206\u7c7b\u5668\u5bf9\u65e0\u6807\u7b7e\u6570\u636e\u8fdb\u884c\u5206\u7c7b\uff0c\u8fd9\u6837\u5c31\u4f1a\u4ea7\u751f\u4f2a\u6807\u7b7e\uff08pseudo label\uff09\u6216\u8f6f\u6807\u7b7e\uff08soft label\uff09\uff0c\u6311\u9009\u4f60\u8ba4\u4e3a\u5206\u7c7b\u6b63\u786e\u7684\u65e0\u6807\u7b7e\u6837\u672c\uff08\u6b64\u5904\u5e94\u8be5\u6709\u4e00\u4e2a<strong>\u6311\u9009\u51c6\u5219<\/strong>\uff09\uff0c\u628a\u9009\u51fa\u6765\u7684\u65e0\u6807\u7b7e\u6837\u672c\u7528\u6765\u8bad\u7ec3\u5206\u7c7b\u5668\u3002<\/p>\n\n\n\n<p>2.<strong>\u534f\u540c\u8bad\u7ec3<\/strong>\uff08co-training\uff09\uff1a\u5176\u5b9e\u4e5f\u662f self-training \u7684\u4e00\u79cd\uff0c\u4f46\u5176\u601d\u60f3\u662f\u597d\u7684\u3002\u5047\u8bbe\u6bcf\u4e2a\u6570\u636e\u53ef\u4ee5\u4ece\u4e0d\u540c\u7684\u89d2\u5ea6\uff08view\uff09\u8fdb\u884c\u5206\u7c7b\uff0c\u4e0d\u540c\u89d2\u5ea6\u53ef\u4ee5\u8bad\u7ec3\u51fa\u4e0d\u540c\u7684\u5206\u7c7b\u5668\uff0c\u7136\u540e\u7528\u8fd9\u4e9b\u4ece\u4e0d\u540c\u89d2\u5ea6\u8bad\u7ec3\u51fa\u6765\u7684\u5206\u7c7b\u5668\u5bf9\u65e0\u6807\u7b7e\u6837\u672c\u8fdb\u884c\u5206\u7c7b\uff0c\u518d\u9009\u51fa\u8ba4\u4e3a\u53ef\u4fe1\u7684\u65e0\u6807\u7b7e\u6837\u672c\u52a0\u5165\u8bad\u7ec3\u96c6\u4e2d\u3002\u7531\u4e8e\u8fd9\u4e9b\u5206\u7c7b\u5668\u4ece\u4e0d\u540c\u89d2\u5ea6\u8bad\u7ec3\u51fa\u6765\u7684\uff0c\u53ef\u4ee5\u5f62\u6210\u4e00\u79cd\u4e92\u8865\uff0c\u800c\u63d0\u9ad8\u5206\u7c7b\u7cbe\u5ea6\uff1b\u5c31\u5982\u540c\u4ece\u4e0d\u540c\u89d2\u5ea6\u53ef\u4ee5\u66f4\u597d\u5730\u7406\u89e3\u4e8b\u7269\u4e00\u6837\u3002<\/p>\n\n\n\n<p>3.<strong>\u534a\u76d1\u7763\u5b57\u5178\u5b66\u4e60<\/strong>\uff1a\u5176\u5b9e\u4e5f\u662f self-training \u7684\u4e00\u79cd\uff0c\u5148\u662f\u7528\u6709\u6807\u7b7e\u6570\u636e\u4f5c\u4e3a\u5b57\u5178\uff0c\u5bf9\u65e0\u6807\u7b7e\u6570\u636e\u8fdb\u884c\u5206\u7c7b\uff0c\u6311\u9009\u51fa\u4f60\u8ba4\u4e3a\u5206\u7c7b\u6b63\u786e\u7684\u65e0\u6807\u7b7e\u6837\u672c\uff0c\u52a0\u5165\u5b57\u5178\u4e2d\uff08\u6b64\u65f6\u7684\u5b57\u5178\u5c31\u53d8\u6210\u4e86\u534a\u76d1\u7763\u5b57\u5178\u4e86\uff09<\/p>\n\n\n\n<p>4.<strong>\u6807\u7b7e\u4f20\u64ad\u7b97\u6cd5<\/strong>\uff08Label Propagation Algorithm\uff09\uff1a\u662f\u4e00\u79cd\u57fa\u4e8e\u56fe\u7684\u534a\u76d1\u7763\u7b97\u6cd5\uff0c\u901a\u8fc7\u6784\u9020\u56fe\u7ed3\u6784\uff08\u6570\u636e\u70b9\u4e3a\u9876\u70b9\uff0c\u70b9\u4e4b\u95f4\u7684\u76f8\u4f3c\u6027\u4e3a\u8fb9\uff09\u6765\u5bfb\u627e<strong>\u8bad\u7ec3\u6570\u636e<\/strong>\u4e2d\u6709\u6807\u7b7e\u6570\u636e\u548c\u65e0\u6807\u7b7e\u6570\u636e\u7684\u5173\u7cfb\u3002\u662f\u7684\uff0c\u53ea\u662f\u8bad\u7ec3\u6570\u636e\u4e2d\uff0c\u8fd9\u662f\u4e00\u79cd\u76f4\u63a8\u5f0f\u7684\u534a\u76d1\u7763\u7b97\u6cd5\uff0c\u5373\u53ea\u5bf9\u8bad\u7ec3\u96c6\u4e2d\u7684\u65e0\u6807\u7b7e\u6570\u636e\u8fdb\u884c\u5206\u7c7b\uff0c\u8fd9\u5176\u5b9e\u611f\u89c9\u5f88\u50cf\u4e00\u4e2a\u6709\u76d1\u7763\u5206\u7c7b\u7b97\u6cd5&#8230;\uff0c\u4f46\u5176\u5b9e\u5e76\u4e0d\u662f\uff0c\u56e0\u4e3a\u5176\u6807\u7b7e\u4f20\u64ad\u7684\u8fc7\u7a0b\uff0c\u4f1a\u6d41\u7ecf\u65e0\u6807\u7b7e\u6570\u636e\uff0c\u5373\u6709\u4e9b\u65e0\u6807\u7b7e\u6570\u636e\u7684\u6807\u7b7e\u7684\u4fe1\u606f\uff0c\u662f\u4ece\u53e6\u4e00\u4e9b\u65e0\u6807\u7b7e\u6570\u636e\u4e2d\u6d41\u8fc7\u6765\u7684\uff0c\u8fd9\u5c31\u7528\u5230\u4e86\u65e0\u6807\u7b7e\u6570\u636e\u4e4b\u95f4\u7684\u8054\u7cfb<\/p>\n\n\n\n<p>5.<strong>\u534a\u76d1\u7763\u652f\u6301\u5411\u91cf\u673a<\/strong>\uff1a\u76d1\u7763\u652f\u6301\u5411\u91cf\u673a\u662f\u5229\u7528\u4e86\u7ed3\u6784\u98ce\u9669\u6700\u5c0f\u5316\u6765\u5206\u7c7b\u7684\uff0c\u534a\u76d1\u7763\u652f\u6301\u5411\u91cf\u673a\u8fd8\u7528\u4e0a\u4e86\u65e0\u6807\u7b7e\u6570\u636e\u7684\u7a7a\u95f4\u5206\u5e03\u4fe1\u606f\uff0c\u5373\u51b3\u7b56\u8d85\u5e73\u9762\u5e94\u8be5\u4e0e\u65e0\u6807\u7b7e\u6570\u636e\u7684\u5206\u5e03\u4e00\u81f4\uff08\u5e94\u8be5\u7ecf\u8fc7\u65e0\u6807\u7b7e\u6570\u636e\u5bc6\u5ea6\u4f4e\u7684\u5730\u65b9\uff09\uff08<strong>\u8fd9\u5176\u5b9e\u662f\u4e00\u79cd\u5047\u8bbe<\/strong>\uff0c\u4e0d\u6ee1\u8db3\u7684\u8bdd\u8fd9\u79cd\u65e0\u6807\u7b7e\u6570\u636e\u7684\u7a7a\u95f4\u5206\u5e03\u4fe1\u606f\u4f1a\u8bef\u5bfc\u51b3\u7b56\u8d85\u5e73\u9762\uff0c\u5bfc\u81f4\u6027\u80fd\u6bd4\u53ea\u7528\u6709\u6807\u7b7e\u6570\u636e\u65f6\u8fd8\u5dee\uff09<\/p>\n\n\n\n<p>\u5176\u5b9e\uff0c\u534a\u76d1\u7763\u5b66\u4e60\u7684\u65b9\u6cd5\u5927\u90fd\u5efa\u7acb\u5728\u5bf9\u6570\u636e\u7684\u67d0\u79cd\u5047\u8bbe\u4e0a\uff0c\u53ea\u6709\u6ee1\u8db3\u8fd9\u4e9b\u5047\u8bbe\uff0c\u534a\u76d1\u7763\u7b97\u6cd5\u624d\u80fd\u6709\u6027\u80fd\u7684\u4fdd\u8bc1\uff0c\u8fd9\u4e5f\u662f\u9650\u5236\u4e86\u534a\u76d1\u7763\u5b66\u4e60\u5e94\u7528\u7684\u4e00\u5927\u969c\u788d\u3002<\/p>\n\n\n\n<h2>\u534a\u76d1\u7763\u6df1\u5ea6\u5b66\u4e60<\/h2>\n\n\n\n<p>\u7ec8\u4e8e\u6765\u5230\u6b63\u9898\u2014\u2014\u534a\u76d1\u7763\u6df1\u5ea6\u5b66\u4e60\uff0c\u6df1\u5ea6\u5b66\u4e60\u9700\u8981\u7528\u5230\u5927\u91cf\u6709\u6807\u7b7e\u6570\u636e\uff0c\u5373\u4f7f\u5728\u5927\u6570\u636e\u65f6\u4ee3\uff0c\u5e72\u51c0\u80fd\u7528\u7684\u6709\u6807\u7b7e\u6570\u636e\u4e5f\u662f\u4e0d\u591a\u7684\uff0c\u7531\u6b64\u5f15\u53d1\u6df1\u5ea6\u5b66\u4e60\u4e0e\u534a\u76d1\u7763\u5b66\u4e60\u7684\u7ed3\u5408\u3002<\/p>\n\n\n\n<p>\u5982\u679c\u8981\u7ed9\u534a\u76d1\u7763\u6df1\u5ea6\u5b66\u4e60\u4e0b\u4e2a\u5b9a\u4e49\uff0c\u5927\u6982\u5c31\u662f\uff0c\u5728\u6709\u6807\u7b7e\u6570\u636e+\u65e0\u6807\u7b7e\u6570\u636e\u6df7\u5408\u6210\u7684\u8bad\u7ec3\u6570\u636e\u4e2d\u4f7f\u7528\u7684\u6df1\u5ea6\u5b66\u4e60\u7b97\u6cd5\u5427&#8230;orz.<\/p>\n\n\n\n<p>\u534a\u76d1\u7763\u6df1\u5ea6\u5b66\u4e60\u7b97\u6cd5\u4e2a\u4eba\u603b\u7ed3\u4e3a\u4e09\u7c7b\uff1a\u65e0\u6807\u7b7e\u6570\u636e\u9884\u8bad\u7ec3\u7f51\u7edc\u540e\u6709\u6807\u7b7e\u6570\u636e\u5fae\u8c03\uff08fine-tune\uff09\uff1b\u6709\u6807\u7b7e\u6570\u636e\u8bad\u7ec3\u7f51\u7edc\uff0c\u5229\u7528\u4ece\u7f51\u7edc\u4e2d\u5f97\u5230\u7684\u6df1\u5ea6\u7279\u5f81\u6765\u505a\u534a\u76d1\u7763\u7b97\u6cd5\uff1b\u8ba9\u7f51\u7edc work in semi-supervised fashion\u3002<\/p>\n\n\n\n<p class=\"has-light-pink-background-color has-background\"><strong>1.\u65e0\u6807\u7b7e\u6570\u636e\u9884\u8bad\u7ec3\uff0c\u6709\u6807\u7b7e\u6570\u636e\u5fae\u8c03<\/strong><\/p>\n\n\n\n<p>\u5bf9\u4e8e\u795e\u7ecf\u7f51\u7edc\u6765\u8bf4\uff0c\u4e00\u4e2a\u597d\u7684\u521d\u59cb\u5316\u53ef\u4ee5\u4f7f\u5f97\u7ed3\u679c\u66f4\u7a33\u5b9a\uff0c\u8fed\u4ee3\u6b21\u6570\u66f4\u5c11\u3002\u56e0\u6b64\u5982\u4f55\u5229\u7528\u65e0\u6807\u7b7e\u6570\u636e\u8ba9\u7f51\u7edc\u6709\u4e00\u4e2a\u597d\u7684\u521d\u59cb\u5316\u5c31\u6210\u4e3a\u4e00\u4e2a\u7814\u7a76\u70b9\u4e86\u3002<\/p>\n\n\n\n<p>\u76ee\u524d\u6211\u89c1\u8fc7\u7684\u521d\u59cb\u5316\u65b9\u5f0f\u6709\u4e24\u79cd\uff1a\u65e0\u76d1\u7763\u9884\u8bad\u7ec3\uff0c\u548c\u4f2a\u6709\u76d1\u7763\u9884\u8bad\u7ec3<\/p>\n\n\n\n<p><strong>\u65e0\u76d1\u7763\u9884\u8bad\u7ec3<\/strong>\uff1a\u4e00\u662f\u7528\u6240\u6709\u6570\u636e\u9010\u5c42\u91cd\u6784\u9884\u8bad\u7ec3\uff0c\u5bf9\u7f51\u7edc\u7684\u6bcf\u4e00\u5c42\uff0c\u90fd\u505a\u91cd\u6784\u81ea\u7f16\u7801\uff0c\u5f97\u5230\u53c2\u6570\u540e\u7528\u6709\u6807\u7b7e\u6570\u636e\u5fae\u8c03\uff1b\u4e8c\u662f\u7528\u6240\u6709\u6570\u636e\u8bad\u7ec3\u91cd\u6784\u81ea\u7f16\u7801\u7f51\u7edc\uff0c\u7136\u540e\u628a\u81ea\u7f16\u7801\u7f51\u7edc\u7684\u53c2\u6570\uff0c\u4f5c\u4e3a\u521d\u59cb\u53c2\u6570\uff0c\u7528\u6709\u6807\u7b7e\u6570\u636e\u5fae\u8c03\u3002<\/p>\n\n\n\n<p><strong>\u4f2a\u6709\u76d1\u7763\u9884\u8bad\u7ec3<\/strong>\uff1a\u901a\u8fc7\u67d0\u79cd\u65b9\u5f0f\/\u7b97\u6cd5\uff08\u5982\u534a\u76d1\u7763\u7b97\u6cd5\uff0c\u805a\u7c7b\u7b97\u6cd5\u7b49\uff09\uff0c\u7ed9\u65e0\u6807\u7b7e\u6570\u636e\u9644\u4e0a\u4f2a\u6807\u7b7e\u4fe1\u606f\uff0c\u5148\u7528\u8fd9\u4e9b\u4f2a\u6807\u7b7e\u4fe1\u606f\u6765\u9884\u8bad\u7ec3\u7f51\u7edc\uff0c\u7136\u540e\u5728\u7528\u6709\u6807\u7b7e\u6570\u636e\u6765\u5fae\u8c03\u3002\uff08MAE\uff1a mask \u7f16\u7801\u5668\uff09<\/p>\n\n\n\n<p class=\"has-light-pink-background-color has-background\"><strong>2.\u5229\u7528\u4ece\u7f51\u7edc\u5f97\u5230\u7684\u6df1\u5ea6\u7279\u5f81\u6765\u505a\u534a\u76d1\u7763\u7b97\u6cd5<\/strong><\/p>\n\n\n\n<p>    \u795e\u7ecf\u7f51\u7edc\u4e0d\u662f\u9700\u8981\u6709\u6807\u7b7e\u6570\u636e\u5417\uff1f\u6211\u7ed9\u4f60\u9020\u4e00\u4e9b\u6709\u6807\u7b7e\u6570\u636e\u51fa\u6765\uff01\u8fd9\u5c31\u662f\u7b2c\u4e8c\u7c7b\u7684\u601d\u60f3\u4e86\uff0c\u76f8\u5f53\u4e8e\u4e00\u79cd\u95f4\u63a5\u7684 self-training \u5427\u3002\u4e00\u822c\u6d41\u7a0b\u662f\uff1a<\/p>\n\n\n\n<p>    \u5148\u7528\u6709\u6807\u7b7e\u6570\u636e\u8bad\u7ec3\u7f51\u7edc\uff08\u6b64\u65f6\u7f51\u7edc\u4e00\u822c\u8fc7\u62df\u5408&#8230;\uff09\uff0c\u4ece\u8be5\u7f51\u7edc\u4e2d\u63d0\u53d6\u6240\u6709\u6570\u636e\u7684\u7279\u5f81\uff0c\u4ee5\u8fd9\u4e9b\u7279\u5f81\u6765\u7528<strong>\u67d0\u79cd\u5206\u7c7b\u7b97\u6cd5<\/strong>\u5bf9\u65e0\u6807\u7b7e\u6570\u636e\u8fdb\u884c\u5206\u7c7b\uff0c\u6311\u9009\u4f60\u8ba4\u4e3a\u5206\u7c7b\u6b63\u786e\u7684\u65e0\u6807\u7b7e\u6570\u636e\u52a0\u5165\u5230\u8bad\u7ec3\u96c6\uff0c\u518d\u8bad\u7ec3\u7f51\u7edc\uff1b\u5982\u6b64\u5faa\u73af\u3002<\/p>\n\n\n\n<p>   \u7531\u4e8e\u7f51\u7edc\u5f97\u5230\u65b0\u7684\u6570\u636e\uff08\u6311\u9009\u51fa\u6765\u5206\u7c7b\u540e\u7684\u65e0\u6807\u7b7e\u6570\u636e\uff09\u4f1a\u66f4\u65b0\u63d0\u5347\uff0c\u4f7f\u5f97\u540e\u7eed\u63d0\u51fa\u6765\u7684\u7279\u5f81\u66f4\u597d\uff0c\u540e\u9762\u5bf9\u65e0\u6807\u7b7e\u6570\u636e\u5206\u7c7b\u5c31\u66f4\u7cbe\u786e\uff0c\u6311\u9009\u540e\u52a0\u5165\u5230\u8bad\u7ec3\u96c6\u4e2d\u53c8\u7ee7\u7eed\u63d0\u5347\u7f51\u7edc\uff0c\u611f\u89c9\u60f3\u6cd5\u5f88\u597d\uff0c\u4f46\u603b\u6709\u54ea\u91cc\u4e0d\u5bf9&#8230;orz<\/p>\n\n\n\n<p>   \u4e2a\u4eba\u731c\u6d4b\u8fd9\u4e2a\u60f3\u6cd5\u4e0d\u80fd\u5f88\u597d\u5730 work \u7684\u539f\u56e0\u53ef\u80fd\u662f\u566a\u58f0\uff0c\u4f60\u6311\u9009\u52a0\u5165\u5230\u8bad\u7ec3\u65e0\u6807\u7b7e\u6570\u636e\u4e00\u822c\u90fd\u5e26\u6709\u6807\u7b7e\u566a\u58f0\uff08\u5c31\u662f\u67d0\u4e9b\u65e0\u6807\u7b7e\u6570\u636e\u88ab\u5206\u7c7b\u9519\u8bef\uff09\uff0c\u8fd9\u79cd\u566a\u58f0\u4f1a\u8bef\u5bfc\u7f51\u7edc\u4e14\u88ab\u7f51\u7edc\u5b66\u4e60\u8bb0\u5fc6\u3002<\/p>\n\n\n\n<p class=\"has-light-pink-background-color has-background\"><strong>3.\u8ba9\u7f51\u7edc work in semi-supervised fashion<\/strong><\/p>\n\n\n\n<p>\u524d\u9762\u76841.\u548c2.\u867d\u7136\u662f\u90fd\u7528\u4e86\u6709\u6807\u7b7e\u6570\u636e\u548c\u65e0\u6807\u7b7e\u6570\u636e\uff0c\u4f46\u5c31\u795e\u7ecf\u7f51\u7edc\u672c\u8eab\u800c\u8a00\uff0c\u5176\u5b9e\u8fd8\u662f\u8fd0\u884c\u5728\u4e00\u79cd\u6709\u76d1\u7763\u7684\u65b9\u5f0f\u4e0a\u3002<\/p>\n\n\n\n<p>\u54ea\u80fd\u4e0d\u80fd\u8ba9\u6df1\u5ea6\u5b66\u4e60\u771f\u6b63\u5730\u6210\u4e3a\u4e00\u79cd\u534a\u76d1\u7763\u7b97\u6cd5\u5462\uff0c\u5f53\u7136\u662f\u53ef\u4ee5\u554a\u3002\u8b6c\u5982\u4e0b\u9762\u8fd9\u4e9b\u65b9\u6cd5\uff1a<\/p>\n\n\n\n<p><strong>Pseudo-Label : The Simple and Efficient Semi-Supervised Learning Method for Deep Neural Networks<\/strong><\/p>\n\n\n\n<p>\u8fd9\u662f\u4e00\u7bc7\u53d1\u8868\u5728 ICML 2013 \u7684\u6587\u7ae0\uff0c\u662f\u4e00\u4e2a\u76f8\u5f53\u7b80\u5355\u7684\u8ba9\u7f51\u7edc work in semi-supervised fashion \u7684\u65b9\u6cd5\u3002<strong>\u5c31\u662f\u628a\u7f51\u7edc\u5bf9\u65e0\u6807\u7b7e\u6570\u636e\u7684\u9884\u6d4b\uff0c\u4f5c\u4e3a\u65e0\u6807\u7b7e\u6570\u636e\u7684\u6807\u7b7e\uff08\u5373 Pseudo label\uff09<\/strong>\uff0c\u7528\u6765\u5bf9\u7f51\u7edc\u8fdb\u884c\u8bad\u7ec3\uff0c\u5176\u601d\u60f3\u5c31\u662f\u4e00\u79cd\u7b80\u5355\u81ea\u8bad\u7ec3\u3002\u4f46\u65b9\u6cd5\u867d\u7136\u7b80\u5355\uff0c\u4f46\u662f\u6548\u679c\u5f88\u597d\uff0c\u6bd4\u5355\u7eaf\u7528\u6709\u6807\u7b7e\u6570\u636e\u6709\u4e0d\u5c11\u7684\u63d0\u5347\u3002<\/p>\n\n\n\n<p>\u7f51\u7edc\u4f7f\u7528\u7684\u4ee3\u4ef7\u51fd\u6570\u5982\u4e0b\uff1a<\/p>\n\n\n\n<p>L=\u2211m=1n\u2211i=1CL(yim,fim)+\u03b1(t)\u2211m=1n\u2032\u2211i=1CL(y\u2032im,f\u2032im)<\/p>\n\n\n\n<p>\u4ee3\u4ef7\u51fd\u6570\u7684\u524d\u9762\u662f\u6709\u6807\u7b7e\u6570\u636e\u7684\u4ee3\u4ef7\uff0c\u540e\u9762\u7684\u65e0\u6807\u7b7e\u6570\u636e\u7684\u4ee3\u4ef7\uff0c\u5728\u65e0\u6807\u7b7e\u6570\u636e\u7684\u4ee3\u4ef7\u4e2d\uff0cy\u2032\u65e0\u6807\u7b7e\u6570\u636e\u7684 pseudo label\uff0c\u662f\u76f4\u63a5\u53d6\u7f51\u7edc\u5bf9\u65e0\u6807\u7b7e\u6570\u636e\u7684\u9884\u6d4b\u7684\u6700\u5927\u503c\u4e3a\u6807\u7b7e\u3002<\/p>\n\n\n\n<p>\u867d\u7136\u601d\u60f3\u7b80\u5355\uff0c\u4f46\u662f\u8fd8\u662f\u6709\u4e9b\u4e1c\u897f\u9700\u8981\u6ce8\u610f\u7684\uff0c\u5c31\u662f\u8fd9\u4e2a\u03b1(t)\uff0c\u5176\u51b3\u5b9a\u7740\u65e0\u6807\u7b7e\u6570\u636e\u7684\u4ee3\u4ef7\u5728\u7f51\u7edc\u66f4\u65b0\u7684\u4f5c\u7528\uff0c\u9009\u62e9\u5408\u9002\u7684\u03b1(t)\u5f88\u91cd\u8981\uff0c\u592a\u5927\u6027\u80fd\u9000\u5316\uff0c\u592a\u5c0f\u63d0\u5347\u6709\u9650\u3002\u5728\u7f51\u7edc\u521d\u59cb\u65f6\uff0c\u7f51\u7edc\u7684\u9884\u6d4b\u65f6\u4e0d\u592a\u51c6\u786e\u7684\uff0c\u56e0\u6b64\u751f\u6210\u7684 pseudo label \u7684\u51c6\u786e\u6027\u4e5f\u4e0d\u9ad8\u3002\u5728\u521d\u59cb\u8bad\u7ec3\u65f6\uff0c\u03b1(t)\u8981\u8bbe\u4e3a 0\uff0c\u7136\u540e\u518d\u6162\u6162\u589e\u52a0\uff0c\u8bba\u6587\u4e2d\u7ed9\u51fa\u5176\u589e\u957f\u51fd\u6570\u3002\u5728\u540e\u9762\u7684\u4ecb\u7ecd\u4e2d\uff0c\u6709\u4e24\u7bc7\u8bba\u6587\u90fd\u4f7f\u7528\u4e86\u4e00\u79cd\u9ad8\u65af\u578b\u7684\u722c\u5347\u51fd\u6570\u3002<\/p>\n\n\n\n<p>\u611f\u89c9\u8fd9\u79cd\u65e0\u6807\u7b7e\u6570\u636e\u4ee3\u4ef7\u8fbe\u5230\u4e00\u79cd\u6b63\u5219\u5316\u7684\u6548\u679c\uff0c\u5176\u51cf\u5c11\u4e86\u7f51\u7edc\u5728\u6709\u9650\u6709\u6807\u7b7e\u6570\u636e\u4e0b\u7684\u8fc7\u62df\u5408\uff0c\u4f7f\u5f97\u7f51\u7edc\u6cdb\u5316\u5730\u66f4\u597d\u3002<\/p>\n\n\n\n<p><strong>Semi-Supervised Learning with Ladder Networks<\/strong><\/p>\n\n\n\n<p>    2015\u5e74\u8bde\u751f\u534a\u76d1\u7763 ladderNet\uff0cladderNet\u662f\u5176\u4ed6\u6587\u7ae0\u4e2d\u5148\u63d0\u51fa\u6765\u7684\u60f3\u6cd5\uff0c\u4f46\u8fd9\u7bc7\u6587\u7ae0\u4f7f\u5b83 work in semi-supervised fashion\uff0c\u800c\u4e14\u6548\u679c\u975e\u5e38\u597d\uff0c\u8fbe\u5230\u4e86\u5f53\u65f6\u7684 state-of-the-art \u6027\u80fd\u3002<\/p>\n\n\n\n<p>   ladderNet \u662f\u6709\u76d1\u7763\u7b97\u6cd5\u548c\u65e0\u76d1\u7763\u7b97\u6cd5\u7684\u6709\u673a\u7ed3\u5408\u3002\u524d\u9762\u63d0\u5230\uff0c\u5f88\u591a\u534a\u76d1\u7763\u6df1\u5ea6\u5b66\u4e60\u7b97\u6cd5\u662f\u7528\u65e0\u76d1\u7763\u9884\u8bad\u7ec3\u8fd9\u79cd\u65b9\u5f0f\u5bf9\u65e0\u6807\u7b7e\u6570\u636e\u8fdb\u884c\u5229\u7528\u7684\uff0c\u4f46\u4e8b\u5b9e\u4e0a\uff0c\u8fd9\u79cd\u628a\u65e0\u76d1\u7763\u5b66\u4e60\u5f3a\u52a0\u5728\u6709\u76d1\u7763\u5b66\u4e60\u4e0a\u7684\u65b9\u5f0f\u6709\u7f3a\u70b9\uff1a\u4e24\u79cd\u5b66\u4e60\u7684\u76ee\u7684\u4e0d\u4e00\u81f4\uff0c\u5176\u5b9e\u5e76\u4e0d\u80fd\u5f88\u597d\u517c\u5bb9\u3002<\/p>\n\n\n\n<p>   \u65e0\u76d1\u7763\u9884\u8bad\u7ec3\u4e00\u822c\u662f\u7528\u91cd\u6784\u6837\u672c\u8fdb\u884c\u8bad\u7ec3\uff0c\u5176\u7f16\u7801\uff08\u5b66\u4e60\u7279\u5f81\uff09\u7684\u76ee\u7684\u662f\u5c3d\u53ef\u80fd\u5730\u4fdd\u7559\u6837\u672c\u7684\u4fe1\u606f\uff1b\u800c\u6709\u76d1\u7763\u5b66\u4e60\u662f\u7528\u4e8e\u5206\u7c7b\uff0c\u5e0c\u671b\u53ea\u4fdd\u7559\u5176\u672c\u8d28\u7279\u5f81\uff0c\u53bb\u9664\u4e0d\u5fc5\u8981\u7684\u7279\u5f81\u3002<\/p>\n\n\n\n<p>   ladderNet \u901a\u8fc7 skip connection \u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\uff0c\u901a\u8fc7\u5728\u6bcf\u5c42\u7684\u7f16\u7801\u5668\u548c\u89e3\u7801\u5668\u4e4b\u95f4\u6dfb\u52a0\u8df3\u8dc3\u8fde\u63a5\uff08skip connection\uff09\uff0c\u51cf\u8f7b\u6a21\u578b\u8f83\u9ad8\u5c42\u8868\u793a\u7ec6\u8282\u7684\u538b\u529b\uff0c\u4f7f\u5f97\u65e0\u76d1\u7763\u5b66\u4e60\u548c\u6709\u76d1\u7763\u5b66\u4e60\u80fd\u7ed3\u5408\u5728\u4e00\u8d77\uff0c\u5e76\u5728\u6700\u9ad8\u5c42\u6dfb\u52a0\u5206\u7c7b\u5668\uff0cladderNet \u5c31\u53d8\u8eab\u6210\u4e00\u4e2a\u534a\u76d1\u7763\u6a21\u578b\u3002<\/p>\n\n\n\n<p>ladderNet \u6709\u673a\u5730\u7ed3\u5408\u4e86\u65e0\u76d1\u7763\u5b66\u4e60\u548c\u6709\u76d1\u7763\u5b66\u4e60\uff0c\u89e3\u51b3\u517c\u5bb9\u6027\u95ee\u9898\uff0c\u53d1\u5c55\u51fa\u4e00\u4e2a\u7aef\u5bf9\u7aef\u7684\u534a\u76d1\u7763\u6df1\u5ea6\u6a21\u578b\u3002<\/p>\n\n\n\n<p>PS\uff1a\u8bba\u6587\u6709\u7ed9\u51fa\u4ee3\u7801<\/p>\n\n\n\n<p><strong>Temporal Ensembling for Semi-supervised Learning<\/strong><\/p>\n\n\n\n<p>Temporal ensembling \u662f Pseudo label \u7684\u53d1\u5c55\uff0c\u76ee\u7684\u662f\u6784\u9020\u66f4\u597d\u7684 pseudo label\uff08\u6587\u4e2d\u79f0\u4e3a target\uff0c\u6211\u8ba4\u4e3a\u662f\u4e00\u81f4\u7684\uff09\u3002<\/p>\n\n\n\n<p>\u591a\u4e2a\u72ec\u7acb\u8bad\u7ec3\u7684\u7f51\u7edc\u7684\u96c6\u6210\u53ef\u53d6\u5f97\u66f4\u597d\u7684\u9884\u6d4b\uff0c\u8bba\u6587\u6269\u5c55\u4e86\u8fd9\u4e2a\u89c2\u70b9\uff0c\u63d0\u51fa\u81ea\u96c6\u6210\uff08self-ensembling\uff09\uff0c\u901a\u8fc7\u540c\u4e00\u4e2a\u6a21\u578b\u5728\u4e0d\u540c\u7684\u8fed\u4ee3\u671f\uff0c\u4e0d\u540c\u7684\u6570\u636e\u589e\u5f3a\u548c\u6b63\u5219\u5316\u7684\u6761\u4ef6\u4e0b\u8fdb\u884c\u96c6\u6210\uff0c\u6765\u6784\u9020\u66f4\u597d\u7684 target\u3002<\/p>\n\n\n\n<p>\u8bba\u6587\u63d0\u51fa\u4e86\u4e24\u79cd\u4e0d\u540c\u7684\u5b9e\u73b0\uff1a&nbsp;\u03a0&nbsp;model \u548c temporal ensembling<\/p>\n\n\n\n<p>\u4e24\u4e2a\u6a21\u578b\u7684\u4ee3\u4ef7\u51fd\u6570\u90fd\u662f\u4e00\u6837\u7684\uff0c\u4e0e Pseudo Label \u7684\u4ee3\u4ef7\u51fd\u6570\u7c7b\u4f3c\uff0c\u4e00\u4e2a\u6709\u76d1\u7763 loss\uff0c\u4e00\u4e2a\u65e0\u76d1\u7763 loss\uff0c\u4e2d\u95f4\u6709\u4e2a\u6743\u7cfb\u6570\u51fd\u6570\uff0c\u4e0e Pseudo Label \u7684\u533a\u522b\u5728\u4e8e\uff0cPseudo Label \u7684\u7b2c\u4e8c\u9879\u662f\u65e0\u6807\u7b7e loss\uff0c\u662f\u53ea\u9488\u5bf9\u65e0\u6807\u7b7e\u6570\u636e\u7684\uff08\u5982\u679c\u6211\u6ca1\u7406\u89e3\u9519..orz\uff09\uff0c\u800c Temporal ensembling \u7684\u7b2c\u4e8c\u9879\u662f \u65e0\u76d1\u7763 loss\uff0c\u662f\u9762\u5411\u5168\u90e8\u6570\u636e\u7684\u3002<\/p>\n\n\n\n<p><strong>\u03a0&nbsp;model<\/strong>&nbsp;\u7684\u65e0\u76d1\u7763\u4ee3\u4ef7\u662f\u5bf9\u540c\u4e00\u4e2a\u8f93\u5165\u5728\u4e0d\u540c\u7684\u6b63\u5219\u548c\u6570\u636e\u589e\u5f3a\u6761\u4ef6\u4e0b\u7684\u4e00\u81f4\u6027\u3002\u5373\u8981\u6c42\u5728\u4e0d\u540c\u7684\u6761\u4ef6\u4e0b\uff0c\u6a21\u578b\u7684\u4f30\u8ba1\u8981\u4e00\u81f4\uff0c\u4ee5\u9f13\u52b1\u7f51\u7edc\u5b66\u4e60\u6570\u636e\u5185\u5728\u7684\u4e0d\u53d8\u6027\u3002<\/p>\n\n\n\n<p>\u7f3a\u70b9\u4e5f\u662f\u76f8\u5f53\u660e\u663e\uff0c\u6bcf\u4e2a\u8fed\u4ee3\u671f\u8981\u5bf9\u540c\u4e00\u4e2a\u8f93\u5165\u5728\u4e0d\u540c\u7684\u6b63\u5219\u548c\u6570\u636e\u589e\u5f3a\u7684\u6761\u4ef6\u4e0b\u9884\u6d4b\u4e24\u6b21\uff0c\u76f8\u5bf9\u8017\u65f6\u3002\u8fd8\u597d\u4e0d\u540c\u7684\u6b63\u5219\u53ef\u4ee5\u4f7f\u7528 dropout \u6765\u5b9e\u73b0\uff0c\u4e0d\u7136\u4e5f\u5f88\u9ebb\u70e6\u3002<\/p>\n\n\n\n<p><strong>temporal ensembling<\/strong>&nbsp;\u6a21\u578b\u662f\u5bf9\u6bcf\u4e00\u6b21\u8fed\u4ee3\u671f\u7684\u9884\u6d4b\u8fdb\u884c\u79fb\u52a8\u5e73\u5747\u6765\u6784\u9020\u66f4\u597d\u7684 target\uff0c\u7136\u540e\u7528\u8fd9\u4e2a target \u6765\u8ba1\u7b97\u65e0\u76d1\u7763 loss\uff0c\u7ee7\u800c\u66f4\u65b0\u7f51\u7edc\u3002<\/p>\n\n\n\n<p>\u7f3a\u70b9\u4e5f\u6709\uff0c\u8bb0\u5f55\u79fb\u52a8\u5e73\u5747\u7684 target \u9700\u8981\u8f83\u591a\u7a7a\u95f4\u3002\u4f46 temporal ensembling \u7684\u6f5c\u529b\u4e5f\u66f4\u5927\uff0c\u53ef\u4ee5\u6536\u96c6\u66f4\u591a\u7684\u4fe1\u606f\uff0c\u5982\u4e8c\u9636\u539f\u59cb\u77e9\uff0c\u53ef\u57fa\u4e8e\u8fd9\u4e9b\u4fe1\u606f\u5bf9\u4e0d\u540c\u7684\u9884\u6d4b\u52a0\u6743\u7b49\u3002<\/p>\n\n\n\n<p>Temporal ensembling \u8fd8\u5bf9\u6807\u7b7e\u566a\u58f0\u5177\u6709\u9c81\u68d2\u6027\uff0c\u5373\u4f7f\u6709\u6807\u7b7e\u6570\u636e\u7684\u6807\u7b7e\u6709\u8bef\u7684\u8bdd\uff0c\u65e0\u76d1\u7763 loss \u53ef\u4ee5\u5e73\u6ed1\u8fd9\u79cd\u9519\u8bef\u6807\u7b7e\u7684\u5f71\u54cd\u3002<\/p>\n\n\n\n<p><strong>Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results<\/strong><\/p>\n\n\n\n<p>Mean Teacher \u8fd9\u7bc7\u6587\u7ae0\u4e00\u4e0a\u6765\u5c31\u8bf4\u201c\u6a21\u578b\u6210\u529f\u7684\u5173\u952e\u5728\u4e8e target \u7684\u8d28\u91cf\u201d\uff0c\u4e00\u8bed\u9053\u7834\u5929\u673a\u554a\u3002\u800c\u63d0\u9ad8 target \u7684\u8d28\u91cf\u7684\u65b9\u6cd5\u76ee\u524d\u6709\u4e24\uff1a1.\u7cbe\u5fc3\u9009\u62e9\u6837\u672c\u566a\u58f0\uff1b2. \u627e\u5230\u4e00\u4e2a\u66f4\u597d\u7684 Teacher model\u3002\u800c\u8bba\u6587\u91c7\u7528\u4e86\u7b2c\u4e8c\u79cd\u65b9\u6cd5\u3002<\/p>\n\n\n\n<p>Mean teacher \u4e5f\u662f\u575a\u4fe1\u201c\u5e73\u5747\u5f97\u5c31\u662f\u6700\u597d\u7684\u201d\uff08\u4e0d\u77e5\u9053\u662f\u4e0d\u662f\u5e73\u5747\u53ef\u4ee5\u53bb\u566a\u7684\u539f\u56e0&#8230;orz\uff09\uff0c\u4f46\u662f\u65f6\u5e8f\u4e0a\u7684\u5e73\u5747\u5df2\u7ecf\u88ab temporal ensembling \u505a\u4e86\uff0c\u56e0\u6b64 Mean teacher \u63d0\u51fa\u4e86\u4e00\u4e2a\u5927\u80c6\u7684\u60f3\u6cd5\uff0c\u6211\u4eec<strong>\u5bf9\u6a21\u578b\u7684\u53c2\u6570\u8fdb\u884c\u79fb\u52a8\u5e73\u5747<\/strong>\uff08weight-averaged\uff09\uff0c\u4f7f\u7528\u8fd9\u4e2a\u79fb\u52a8\u5e73\u5747\u6a21\u578b\u53c2\u6570\u7684\u5c31\u662f teacher model \u4e86\uff0c\u7136\u540e\u7528 teacher model \u6765\u6784\u9020\u9ad8\u8d28\u91cf target\u3002<\/p>\n\n\n\n<p>\u4e00\u601d\u7d22\u5c31\u89c9\u5f97\u8fd9\u60f3\u6cd5\u597d\uff0c\u5bf9\u6a21\u578b\u7684\u53c2\u6570\u8fdb\u884c\u5e73\u5747\uff0c\u6bcf\u6b21\u66f4\u65b0\u7684\u7f51\u7edc\u7684\u65f6\u5019\u5c31\u80fd\u66f4\u65b0 teacher model\uff0c\u5c31\u80fd\u5f97\u5230 target\uff0c\u4e0d\u7528\u50cf temporal ensembling \u90a3\u6837\u7b49\u4e00\u4e2a\u8fed\u4ee3\u671f\u8fd9\u4e48\u4e45\uff0c\u8fd9\u5bf9 online model \u662f\u81f4\u547d\u7684\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u534a\u76d1\u7763\u5b66\u4e60(Semi-Supervised Learning\uff0cSSL) \u4f7f\u7528\u6807\u8bb0\u548c\u672a\u6807\u8bb0\u7684\u6570\u636e\u6765\u6267\u884c\u6709\u76d1\u7763\u7684\u5b66 &hellip; <a href=\"http:\/\/139.9.1.231\/index.php\/2022\/09\/08\/ssl_sum\/\" class=\"more-link\">\u7ee7\u7eed\u9605\u8bfb<span class=\"screen-reader-text\">\u534a\u76d1\u7763\u5b66\u4e60\u7efc\u8ff0<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[27],"tags":[],"_links":{"self":[{"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/posts\/7013"}],"collection":[{"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/comments?post=7013"}],"version-history":[{"count":16,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/posts\/7013\/revisions"}],"predecessor-version":[{"id":7630,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/posts\/7013\/revisions\/7630"}],"wp:attachment":[{"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/media?parent=7013"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/categories?post=7013"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/tags?post=7013"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}