{"id":8552,"date":"2024-06-04T23:01:01","date_gmt":"2024-06-04T15:01:01","guid":{"rendered":""},"modified":"2024-06-04T23:01:01","modified_gmt":"2024-06-04T15:01:01","slug":"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b","status":"publish","type":"post","link":"https:\/\/mushiming.com\/8552.html","title":{"rendered":"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b"},"content":{"rendered":"<p><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" style=\"display: none;\"> \n <path stroke-linecap=\"round\" d=\"M5,0 0,2.5 5,5z\" id=\"raphael-marker-block\" style=\"-webkit-tap-highlight-color: rgba(0, 0, 0, 0);\"><\/path> \n<\/svg> <\/p>\n<div class=\"toc\">\n<h4>\u6587\u7ae0\u76ee\u5f55<\/h4>\n<ul>\n<li>2 \u56fe\u50cf\u5206\u7c7b<\/li>\n<li>\n<ul>\n<li>2.1 \u6570\u636e\u9a71\u52a8\u65b9\u6cd5<\/li>\n<li>\n<ul>\n<li>\u7b2c\u4e00\u4e2a\u5206\u7c7b\u5668:\u6700\u8fd1\u90bb\u5206\u7c7b\u5668<\/li>\n<li>L1\u8ddd\u79bb\uff08\u66fc\u54c8\u987f\u8ddd\u79bb\uff09<\/li>\n<li>\u6700\u8fd1\u90bb\u5206\u7c7b\u5668Python\u4ee3\u7801<\/li>\n<\/ul>\n<\/li>\n<li>2.2 K-\u6700\u90bb\u8fd1\u7b97\u6cd5<\/li>\n<li>\n<ul>\n<li>L2\u8ddd\u79bb\uff08\u6b27\u6c0f\u8ddd\u79bb\uff09<\/li>\n<li>k-Nearest Neighbor\u5206\u7c7b\u5668\uff08KNN\uff09<\/li>\n<li>\u5b9e\u9645\u5e94\u7528k-NN<\/li>\n<\/ul>\n<\/li>\n<li>2.3 \u7ebf\u6027\u5206\u7c7b\u5668<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<\/div>\n<h2>2 \u56fe\u50cf\u5206\u7c7b<\/h2>\n<h3>2.1 \u6570\u636e\u9a71\u52a8\u65b9\u6cd5<\/h3>\n<p><font size=\"5\" color=\"red\"><strong>\u6b65\u9aa4<\/strong><\/font><\/p>\n<ul>\n<li>step1\uff1a\u6536\u96c6\u56fe\u50cf\u548c\u6807\u7b7e\u7684\u6570\u636e\u96c6Collect a dataset of images and labels<\/li>\n<li>step2\uff1a\u4f7f\u7528\u673a\u5668\u5b66\u4e60\u6765\u8bad\u7ec3\u5206\u7c7b\u5668Use Machine Learning to train a classifier<\/li>\n<li>step3\uff1a\u8bc4\u4f30\u65b0\u56fe\u50cf\u4e0a\u7684\u5206\u7c7b\u5668 Evaluate the classifier on new images<\/li>\n<\/ul>\n<h4>\u7b2c\u4e00\u4e2a\u5206\u7c7b\u5668:\u6700\u8fd1\u90bb\u5206\u7c7b\u5668<\/h4>\n<p>\u7b97\u6cd5\u601d\u60f3\u5f88\u7b80\u5355\uff0c\u5206\u4e24\u6b65\uff1a<br \/> 1\u3001Memorize all data and labels<br \/> 2\u3001Predict the label of the most similar training image<\/p>\n<ul>\n<li>\n<p>\u8bad\u7ec3\u51fd\u6570\uff1a\u63a5\u6536\u56fe\u7247\u548c\u6807\u7b7e\uff0c\u8f93\u51fa\u6a21\u578b<\/p>\n<pre class=\"language-java prettyprint linenums\"><code class=\"prism language-python\"><span class=\"token keyword\">def<\/span> <span class=\"token function\">train<\/span><span class=\"token punctuation\">(<\/span>images<span class=\"token punctuation\">,<\/span>labels<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span>\n    <span class=\"token comment\"># Machine learing!<\/span>\n    <span class=\"token keyword\">return<\/span> model\n<\/code><\/pre>\n<\/li>\n<li>\n<p>\u9884\u6d4b\u51fd\u6570\uff1a\u63a5\u6536\u4e00\u4e2a\u6a21\u578b\uff0c\u5bf9\u56fe\u7247\u79cd\u7c7b\u8fdb\u884c\u9884\u6d4b\uff0c\u8fd4\u56de\u6807\u7b7e<\/p>\n<pre class=\"language-java prettyprint linenums\"><code class=\"prism language-python\"><span class=\"token keyword\">def<\/span> <span class=\"token function\">predict<\/span><span class=\"token punctuation\">(<\/span>model<span class=\"token punctuation\">,<\/span>test_images<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span>\n    <span class=\"token comment\"># Use model to predict labels<\/span>\n    <span class=\"token keyword\">return<\/span> test_labels\n<\/code><\/pre>\n<\/li>\n<\/ul>\n<p><strong>\u793a\u4f8b\u6570\u636e\u96c6\uff1a CIFAR10<\/strong><br \/> \u5305\u542b\uff1a10 classes 50,000 training images 10,000 testing images<\/p>\n<p>\u5c06\u6d4b\u8bd5\u56fe\u7247\u4e0e\u6240\u6709\u8bad\u7ec3\u56fe\u7247\u8fdb\u884c\u6bd4\u8f83\uff0c\u9700\u8981\u6bd4\u8f83\u56fe\u7247\u7684\u65b9\u6cd5\uff0c\u8ba1\u7b97\u76f8\u4f3c\u5ea6\u5c31\u9700\u8981\u5c31\u6709\u4e00\u4e2a\u5ea6\u91cf\u76f8\u4f3c\u5ea6\u7684\u4e1c\u897f\uff0c\u4f8b\u5982L1\u8ddd\u79bb<\/p>\n<h4>L1\u8ddd\u79bb\uff08\u66fc\u54c8\u987f\u8ddd\u79bb\uff09<\/h4>\n<p>L1 distance\uff1a<br \/> <span class=\"katex--display\"><span class=\"katex-display\"><span class=\"katex\"><span class=\"katex-mathml\"> d 1 ( I 1 , I 2 ) = \u2211 p \u2223 I 1 p \u2212 I 2 p \u2223 d_1(I_1,I_2)=\\sum_p\u2223I_1^p\u2212I_2^p| <\/span><span class=\"katex-html\"><span class=\"base\"><span class=\"strut\" style=\"height: 1em; vertical-align: -0.25em;\"><\/span><span class=\"mord\"><span class=\"mord mathdefault\">d<\/span><span class=\"msupsub\"><span class=\"vlist-t vlist-t2\"><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.301108em;\"><span class=\"\" style=\"top: -2.55em; margin-left: 0em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mtight\">1<\/span><\/span><\/span><\/span><span class=\"vlist-s\">\u200b<\/span><\/span><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.15em;\"><span class=\"\"><\/span><\/span><\/span><\/span><\/span><\/span><span class=\"mopen\">(<\/span><span class=\"mord\"><span style=\"margin-right: 0.07847em;\" class=\"mord mathdefault\">I<\/span><span class=\"msupsub\"><span class=\"vlist-t vlist-t2\"><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.301108em;\"><span class=\"\" style=\"top: -2.55em; margin-left: -0.07847em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mtight\">1<\/span><\/span><\/span><\/span><span class=\"vlist-s\">\u200b<\/span><\/span><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.15em;\"><span class=\"\"><\/span><\/span><\/span><\/span><\/span><\/span><span class=\"mpunct\">,<\/span><span class=\"mspace\" style=\"margin-right: 0.166667em;\"><\/span><span class=\"mord\"><span style=\"margin-right: 0.07847em;\" class=\"mord mathdefault\">I<\/span><span class=\"msupsub\"><span class=\"vlist-t vlist-t2\"><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.301108em;\"><span class=\"\" style=\"top: -2.55em; margin-left: -0.07847em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mtight\">2<\/span><\/span><\/span><\/span><span class=\"vlist-s\">\u200b<\/span><\/span><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.15em;\"><span class=\"\"><\/span><\/span><\/span><\/span><\/span><\/span><span class=\"mclose\">)<\/span><span class=\"mspace\" style=\"margin-right: 0.277778em;\"><\/span><span class=\"mrel\">=<\/span><span class=\"mspace\" style=\"margin-right: 0.277778em;\"><\/span><\/span><span class=\"base\"><span class=\"strut\" style=\"height: 2.43612em; vertical-align: -1.38611em;\"><\/span><span class=\"mop op-limits\"><span class=\"vlist-t vlist-t2\"><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 1.05001em;\"><span class=\"\" style=\"top: -1.9em; margin-left: 0em;\"><span class=\"pstrut\" style=\"height: 3.05em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mathdefault mtight\">p<\/span><\/span><\/span><span class=\"\" style=\"top: -3.05001em;\"><span class=\"pstrut\" style=\"height: 3.05em;\"><\/span><span class=\"\"><span class=\"mop op-symbol large-op\">\u2211<\/span><\/span><\/span><\/span><span class=\"vlist-s\">\u200b<\/span><\/span><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 1.38611em;\"><span class=\"\"><\/span><\/span><\/span><\/span><\/span><span class=\"mspace\" style=\"margin-right: 0.277778em;\"><\/span><span class=\"mrel\">\u2223<\/span><span class=\"mspace\" style=\"margin-right: 0.277778em;\"><\/span><\/span><span class=\"base\"><span class=\"strut\" style=\"height: 1.04861em; vertical-align: -0.266308em;\"><\/span><span class=\"mord\"><span style=\"margin-right: 0.07847em;\" class=\"mord mathdefault\">I<\/span><span class=\"msupsub\"><span class=\"vlist-t vlist-t2\"><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.7823em;\"><span class=\"\" style=\"top: -2.43369em; margin-left: -0.07847em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mtight\">1<\/span><\/span><\/span><span class=\"\" style=\"top: -3.18091em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mathdefault mtight\">p<\/span><\/span><\/span><\/span><span class=\"vlist-s\">\u200b<\/span><\/span><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.266308em;\"><span class=\"\"><\/span><\/span><\/span><\/span><\/span><\/span><span class=\"mord\">\u2212<\/span><span class=\"mord\"><span style=\"margin-right: 0.07847em;\" class=\"mord mathdefault\">I<\/span><span class=\"msupsub\"><span class=\"vlist-t vlist-t2\"><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.7823em;\"><span class=\"\" style=\"top: -2.43369em; margin-left: -0.07847em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mtight\">2<\/span><\/span><\/span><span class=\"\" style=\"top: -3.18091em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mathdefault mtight\">p<\/span><\/span><\/span><\/span><span class=\"vlist-s\">\u200b<\/span><\/span><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.266308em;\"><span class=\"\"><\/span><\/span><\/span><\/span><\/span><\/span><span class=\"mord\">\u2223<\/span><\/span><\/span><\/span><\/span><\/span><br \/> \u5982\u679c\u4e24\u5f20\u56fe\u7247\u4e00\u6a21\u4e00\u6837\uff0c\u90a3\u4e48L1\u8ddd\u79bb\u4e3a0\uff0c\u4f46\u662f\u5982\u679c\u4e24\u5f20\u56fe\u7247\u5f88\u662f\u4e0d\u540c\uff0c\u90a3L1\u503c\u5c06\u4f1a\u975e\u5e38\u5927\u3002<\/p>\n<p>\u5047\u8bbe\u56fe\u50cf\u662f4*4\u5927\u5c0f\uff0c\u90a3\u4e48\u53ea\u53d6\u5de6\u4e0a\u89d2\u7684\u5143\u7d20\u8fdb\u884c\u4ee5\u4e0b\u8ba1\u7b97\uff0c\u5f97\u5230456\u5904\u4e0d\u540c\uff1a<\/p>\n<p><img decoding=\"async\" src=\"\" =\"\" =\"image-20200217101243969\" alt=\"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b (https:\/\/mushiming.com\/)  \u7b2c1\u5f20\" title=\"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b  \u7b2c1\u5f20-\u7a46\u4e16\u660e\u535a\u5ba2\" > <\/p>\n<h4>\u6700\u8fd1\u90bb\u5206\u7c7b\u5668Python\u4ee3\u7801<\/h4>\n<pre class=\"language-java prettyprint linenums\"><code class=\"prism language-python\"><span class=\"token keyword\">import<\/span> numpy <span class=\"token keyword\">as<\/span> np\n\n<span class=\"token keyword\">class<\/span> <span class=\"token class-name\">NearestNeighbor<\/span><span class=\"token punctuation\">(<\/span><span class=\"token builtin\">object<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span>\n  <span class=\"token keyword\">def<\/span> <span class=\"token function\">__init__<\/span><span class=\"token punctuation\">(<\/span>self<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span>\n    <span class=\"token keyword\">pass<\/span>\n\n  <span class=\"token keyword\">def<\/span> <span class=\"token function\">train<\/span><span class=\"token punctuation\">(<\/span>self<span class=\"token punctuation\">,<\/span> X<span class=\"token punctuation\">,<\/span> y<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span>\n    <span class=\"token triple-quoted-string string\">\"\"\" X is N x D where each row is an example. Y is 1-dimension of size N \"\"\"<\/span>\n    <span class=\"token comment\"># the nearest neighbor classifier simply remembers all the training data<\/span>\n    self<span class=\"token punctuation\">.<\/span>Xtr <span class=\"token operator\">=<\/span> X\n    self<span class=\"token punctuation\">.<\/span>ytr <span class=\"token operator\">=<\/span> y\n\n  <span class=\"token keyword\">def<\/span> <span class=\"token function\">predict<\/span><span class=\"token punctuation\">(<\/span>self<span class=\"token punctuation\">,<\/span> X<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span>\n    <span class=\"token triple-quoted-string string\">\"\"\" X is N x D where each row is an example we wish to predict label for \"\"\"<\/span>\n    num_test <span class=\"token operator\">=<\/span> X<span class=\"token punctuation\">.<\/span>shape<span class=\"token punctuation\">[<\/span><span class=\"token number\">0<\/span><span class=\"token punctuation\">]<\/span>\n    <span class=\"token comment\"># lets make sure that the output type matches the input type<\/span>\n    Ypred <span class=\"token operator\">=<\/span> np<span class=\"token punctuation\">.<\/span>zeros<span class=\"token punctuation\">(<\/span>num_test<span class=\"token punctuation\">,<\/span> dtype <span class=\"token operator\">=<\/span> self<span class=\"token punctuation\">.<\/span>ytr<span class=\"token punctuation\">.<\/span>dtype<span class=\"token punctuation\">)<\/span>\n\n    <span class=\"token comment\"># loop over all test rows<\/span>\n    <span class=\"token keyword\">for<\/span> i <span class=\"token keyword\">in<\/span> <span class=\"token builtin\">xrange<\/span><span class=\"token punctuation\">(<\/span>num_test<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span>\n      <span class=\"token comment\"># find the nearest training image to the i'th test image<\/span>\n      <span class=\"token comment\"># using the L1 distance (sum of absolute value differences)<\/span>\n      distances <span class=\"token operator\">=<\/span> np<span class=\"token punctuation\">.<\/span><span class=\"token builtin\">sum<\/span><span class=\"token punctuation\">(<\/span>np<span class=\"token punctuation\">.<\/span><span class=\"token builtin\">abs<\/span><span class=\"token punctuation\">(<\/span>self<span class=\"token punctuation\">.<\/span>Xtr <span class=\"token operator\">-<\/span> X<span class=\"token punctuation\">[<\/span>i<span class=\"token punctuation\">,<\/span><span class=\"token punctuation\">:<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> axis <span class=\"token operator\">=<\/span> <span class=\"token number\">1<\/span><span class=\"token punctuation\">)<\/span>\n      min_index <span class=\"token operator\">=<\/span> np<span class=\"token punctuation\">.<\/span>argmin<span class=\"token punctuation\">(<\/span>distances<span class=\"token punctuation\">)<\/span> <span class=\"token comment\"># get the index with smallest distance<\/span>\n      Ypred<span class=\"token punctuation\">[<\/span>i<span class=\"token punctuation\">]<\/span> <span class=\"token operator\">=<\/span> self<span class=\"token punctuation\">.<\/span>ytr<span class=\"token punctuation\">[<\/span>min_index<span class=\"token punctuation\">]<\/span> <span class=\"token comment\"># predict the label of the nearest example<\/span>\n\n    <span class=\"token keyword\">return<\/span> Ypred\n\n<\/code><\/pre>\n<ul>\n<li>\u5c06CIFAR-10\u7684\u6570\u636e\u52a0\u8f7d\u5230\u5185\u5b58\u4e2d\uff0c\u5e76\u5206\u62104\u4e2a\u6570\u7ec4\uff1a\u8bad\u7ec3\u6570\u636e\u548c\u6807\u7b7e\uff0c\u6d4b\u8bd5\u6570\u636e\u548c\u6807\u7b7e\u3002\u5728\u540e\u9762\u7684\u4ee3\u7801\u4e2d\uff0c<strong>Xtr<\/strong>\uff08\u5927\u5c0f\u662f50000x32x32x3\uff09\u5b58\u6709\u8bad\u7ec3\u96c6\u4e2d\u6240\u6709\u7684\u56fe\u50cf\uff0c<strong>Ytr<\/strong>\u662f\u5bf9\u5e94\u7684\u957f\u5ea6\u4e3a50000\u76841\u7ef4\u6570\u7ec4\uff0c\u5b58\u6709\u56fe\u50cf\u5bf9\u5e94\u7684\u5206\u7c7b\u6807\u7b7e\uff08\u4ece0\u52309\uff09\uff1a<\/li>\n<li>\u5f15\u7528\u4e86numpy\u63d0\u4f9b\u7684\u5411\u91cf\u8fd0\u7b97\uff08\u5982shape\u51fd\u6570\uff09<\/li>\n<li>\u5728train\u65b9\u6cd5\u4e2d\uff0c\u4e3b\u8981\u662f\u8bb0\u5f55\u8bad\u7ec3\u6570\u636e\u96c6\uff0c\u8fd9\u91cc\u5176\u5b9e\u5c31\u662f\u4e00\u4e2a\u6307\u9488\u6307\u5411\u91cc\u9762\u7684\u67d0\u4e00\u4e2a\u56fe\u7247<\/li>\n<li>\u5728predict\u65b9\u6cd5\u91cc\u9762\uff0c\u7528L1\u8ddd\u79bb\u53bb\u9010\u4e00\u6bd4\u8f83\u6240\u6709\u8bad\u7ec3\u6570\u636e\u96c6\u7684\u56fe\u7247\uff0c\u627e\u5230\u8ddd\u79bb\u6700\u5c0f\u7684\u56fe\u7247\uff0c\u7136\u540e\u8f93\u51fa\u9884\u6d4b\u7ed3\u679c<\/li>\n<\/ul>\n<p><strong>\u7f3a\u70b9<\/strong>\uff1a\u8bad\u7ec3\u8fc7\u7a0b\u5feb\uff0c\u800c\u6d4b\u8bd5\u8fc7\u7a0b\u6162\uff0c\u4e0e\u671f\u671b\u4e0d\u7b26\uff0c\u8f83\u4e3a\u843d\u540e\u3002<\/p>\n<p>\u4ece\u56fe\u4e0a\u770b\u5b9e\u9645\u8fd0\u7528\u7684\u7f3a\u70b9\uff1a\u4f8b\u5982\u7eff\u8272\u533a\u57df\u91cc\u9762\u51fa\u73b0\u4e86\u9ec4\u70b9\uff0c\u7eff\u8272\u533a\u57df\u63d2\u5165\u4e86\u84dd\u8272\u548c\u7c89\u8272\u533a\u57df\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/img.mushiming.top\/app\/mushiming_com\/5ebe502439064bbda54ae91ad479518f.png\" =\"\" =\"\" alt=\"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b (https:\/\/mushiming.com\/)  \u7b2c2\u5f20\" title=\"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b  \u7b2c2\u5f20-\u7a46\u4e16\u660e\u535a\u5ba2\" ><\/p>\n<h3>2.2 K-\u6700\u90bb\u8fd1\u7b97\u6cd5<\/h3>\n<p><strong>\u539f\u7406\uff1a<\/strong><\/p>\n<p>\u627e\u6700\u76f8\u4f3c\u7684k\u4e2a\u56fe\u7247\u7684\u6807\u7b7e\uff0c\u7136\u540e\u8ba9\u4ed6\u4eec\u9488\u5bf9\u6d4b\u8bd5\u56fe\u7247\u8fdb\u884c\u6295\u7968\uff0c\u6700\u540e\u628a\u7968\u6570\u6700\u9ad8\u7684\u6807\u7b7e\u4f5c\u4e3a\u5bf9\u6d4b\u8bd5\u56fe\u7247\u7684\u9884\u6d4b\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/img.mushiming.top\/app\/mushiming_com\/5f826717a62e47209cdf88594fd812f9.png\" =\"\" =\"\" alt=\"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b (https:\/\/mushiming.com\/)  \u7b2c3\u5f20\" title=\"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b  \u7b2c3\u5f20-\u7a46\u4e16\u660e\u535a\u5ba2\" ><\/p>\n<p>\u53ef\u89c1k\u53d8\u5927\u65f6\u4e0d\u518d\u51fa\u73b0\u566a\u58f0\u70b9\uff0c\u8fb9\u754c\u4e5f\u66f4\u6e05\u6670\u5e73\u6ed1\u3002<\/p>\n<p><strong>\u7b97\u6cd5\u7684\u63cf\u8ff0:<\/strong><\/p>\n<p>1\uff09\u8ba1\u7b97\u6d4b\u8bd5\u6570\u636e\u4e0e\u5404\u4e2a\u8bad\u7ec3\u6570\u636e\u4e4b\u95f4\u7684\u8ddd\u79bb\uff1b<\/p>\n<p>2\uff09\u6309\u7167\u8ddd\u79bb\u7684\u9012\u589e\u5173\u7cfb\u8fdb\u884c\u6392\u5e8f\uff1b<\/p>\n<p>3\uff09\u9009\u53d6\u8ddd\u79bb\u6700\u5c0f\u7684K\u4e2a\u70b9\uff1b<\/p>\n<p>4\uff09\u786e\u5b9a\u524dK\u4e2a\u70b9\u6240\u5728\u7c7b\u522b\uff08\u6807\u7b7e\uff09\u7684\u51fa\u73b0\u9891\u7387\uff1b<\/p>\n<p>5\uff09\u8fd4\u56de\u524dK\u4e2a\u70b9\u4e2d\u51fa\u73b0\u9891\u7387\u6700\u9ad8\u7684\u7c7b\u522b\u4f5c\u4e3a\u6d4b\u8bd5\u6570\u636e\u7684\u9884\u6d4b\u5206\u7c7b\u3002<\/p>\n<blockquote>\n<p>\u53c2\u8003\uff1ahttps:\/\/www.cnblogs.com\/ybjourney\/p\/4702562.html<\/p>\n<\/blockquote>\n<h4>L2\u8ddd\u79bb\uff08\u6b27\u6c0f\u8ddd\u79bb\uff09<\/h4>\n<p><span class=\"katex--display\"><span class=\"katex-display\"><span class=\"katex\"><span class=\"katex-mathml\"> d 2 ( I 1 , I 2 ) = \u2211 p ( I 1 p \u2212 I 2 p ) 2 d_2(I_1,I_2)=\\sqrt{\\sum_p(I_1^p\u2212I_2^p)^2} <\/span><span class=\"katex-html\"><span class=\"base\"><span class=\"strut\" style=\"height: 1em; vertical-align: -0.25em;\"><\/span><span class=\"mord\"><span class=\"mord mathdefault\">d<\/span><span class=\"msupsub\"><span class=\"vlist-t vlist-t2\"><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.301108em;\"><span class=\"\" style=\"top: -2.55em; margin-left: 0em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mtight\">2<\/span><\/span><\/span><\/span><span class=\"vlist-s\">\u200b<\/span><\/span><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.15em;\"><span class=\"\"><\/span><\/span><\/span><\/span><\/span><\/span><span class=\"mopen\">(<\/span><span class=\"mord\"><span style=\"margin-right: 0.07847em;\" class=\"mord mathdefault\">I<\/span><span class=\"msupsub\"><span class=\"vlist-t vlist-t2\"><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.301108em;\"><span class=\"\" style=\"top: -2.55em; margin-left: -0.07847em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mtight\">1<\/span><\/span><\/span><\/span><span class=\"vlist-s\">\u200b<\/span><\/span><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.15em;\"><span class=\"\"><\/span><\/span><\/span><\/span><\/span><\/span><span class=\"mpunct\">,<\/span><span class=\"mspace\" style=\"margin-right: 0.166667em;\"><\/span><span class=\"mord\"><span style=\"margin-right: 0.07847em;\" class=\"mord mathdefault\">I<\/span><span class=\"msupsub\"><span class=\"vlist-t vlist-t2\"><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.301108em;\"><span class=\"\" style=\"top: -2.55em; margin-left: -0.07847em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mtight\">2<\/span><\/span><\/span><\/span><span class=\"vlist-s\">\u200b<\/span><\/span><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.15em;\"><span class=\"\"><\/span><\/span><\/span><\/span><\/span><\/span><span class=\"mclose\">)<\/span><span class=\"mspace\" style=\"margin-right: 0.277778em;\"><\/span><span class=\"mrel\">=<\/span><span class=\"mspace\" style=\"margin-right: 0.277778em;\"><\/span><\/span><span class=\"base\"><span class=\"strut\" style=\"height: 3.04em; vertical-align: -1.57418em;\"><\/span><span class=\"mord sqrt\"><span class=\"vlist-t vlist-t2\"><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 1.46582em;\"><span class=\"svg-align\" style=\"top: -5em;\"><span class=\"pstrut\" style=\"height: 5em;\"><\/span><span class=\"mord\" style=\"padding-left: 1em;\"><span class=\"mop op-limits\"><span class=\"vlist-t vlist-t2\"><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 1.05001em;\"><span class=\"\" style=\"top: -1.9em; margin-left: 0em;\"><span class=\"pstrut\" style=\"height: 3.05em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mathdefault mtight\">p<\/span><\/span><\/span><span class=\"\" style=\"top: -3.05001em;\"><span class=\"pstrut\" style=\"height: 3.05em;\"><\/span><span class=\"\"><span class=\"mop op-symbol large-op\">\u2211<\/span><\/span><\/span><\/span><span class=\"vlist-s\">\u200b<\/span><\/span><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 1.38611em;\"><span class=\"\"><\/span><\/span><\/span><\/span><\/span><span class=\"mopen\">(<\/span><span class=\"mord\"><span style=\"margin-right: 0.07847em;\" class=\"mord mathdefault\">I<\/span><span class=\"msupsub\"><span class=\"vlist-t vlist-t2\"><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.7823em;\"><span class=\"\" style=\"top: -2.43369em; margin-left: -0.07847em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mtight\">1<\/span><\/span><\/span><span class=\"\" style=\"top: -3.18091em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mathdefault mtight\">p<\/span><\/span><\/span><\/span><span class=\"vlist-s\">\u200b<\/span><\/span><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.266308em;\"><span class=\"\"><\/span><\/span><\/span><\/span><\/span><\/span><span class=\"mord\">\u2212<\/span><span class=\"mord\"><span style=\"margin-right: 0.07847em;\" class=\"mord mathdefault\">I<\/span><span class=\"msupsub\"><span class=\"vlist-t vlist-t2\"><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.7823em;\"><span class=\"\" style=\"top: -2.43369em; margin-left: -0.07847em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mtight\">2<\/span><\/span><\/span><span class=\"\" style=\"top: -3.18091em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mathdefault mtight\">p<\/span><\/span><\/span><\/span><span class=\"vlist-s\">\u200b<\/span><\/span><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.266308em;\"><span class=\"\"><\/span><\/span><\/span><\/span><\/span><\/span><span class=\"mclose\"><span class=\"mclose\">)<\/span><span class=\"msupsub\"><span class=\"vlist-t\"><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 0.740108em;\"><span class=\"\" style=\"top: -2.989em; margin-right: 0.05em;\"><span class=\"pstrut\" style=\"height: 2.7em;\"><\/span><span class=\"sizing reset-size6 size3 mtight\"><span class=\"mord mtight\">2<\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span><span class=\"\" style=\"top: -3.42582em;\"><span class=\"pstrut\" style=\"height: 5em;\"><\/span><span class=\"hide-tail\" style=\"min-width: 1.02em; height: 3.08em;\"><br \/>\n            <svg width=\"400em\" height=\"3.08em\" viewbox=\"0 0 400000 3240\" preserveaspectratio=\"xMinYMin slice\"> \n             <path d=\"M473,2793c339.3,-1799.3,509.3,-2700,510,-2702\nc3.3,-7.3,9.3,-11,18,-11H400000v40H1017.7s-90.5,478,-276.2,1466c-185.7,988,\n-279.5,1483,-281.5,1485c-2,6,-10,9,-24,9c-8,0,-12,-0.7,-12,-2c0,-1.3,-5.3,-32,\n-16,-92c-50.7,-293.3,-119.7,-693.3,-207,-1200c0,-1.3,-5.3,8.7,-16,30c-10.7,\n21.3,-21.3,42.7,-32,64s-16,33,-16,33s-26,-26,-26,-26s76,-153,76,-153s77,-151,\n77,-151c0.7,0.7,35.7,202,105,604c67.3,400.7,102,602.7,104,606z\nM1001 80H400000v40H1017z\"><\/path> \n            <\/svg><\/span><\/span><\/span><span class=\"vlist-s\">\u200b<\/span><\/span><span class=\"vlist-r\"><span class=\"vlist\" style=\"height: 1.57418em;\"><span class=\"\"><\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/p>\n<blockquote>\n<p><strong>\u91cd\u8981\u539f\u5219<\/strong>\uff1a\u53ef\u4ee5\u770b\u5230L1\u5728\u65cb\u8f6c\u5750\u6807\u540e\uff0c\u8ddd\u79bb\u662f\u53d8\u5316\u7684\uff0c\u800cL2\u4e0d\u4f1a\uff0c\u56e0\u6b64\u5982\u679c\u8f93\u5165\u7684\u7279\u5f81\u5411\u91cf\u4e2d\u7684\u4e00\u4e9b\u503c\u5bf9\u4e8e\u8f93\u51fa\u6709\u975e\u5e38\u91cd\u8981\u7684\u610f\u4e49\u6216\u5f71\u54cd\uff0c\u8fd9\u4e2a\u65f6\u5019\u5e94\u8be5\u4f7f\u7528L1\u8ddd\u79bb\uff0c\u4f46\u662f\u5982\u679c\u7279\u5f81\u5411\u91cf\u53ea\u662f\u67d0\u4e2a\u7a7a\u95f4\u4e2d\u7684\u4e00\u4e2a\u901a\u7528\u5411\u91cf\uff0c\u7528L2\u4f1a\u597d\u4e00\u70b9\u3002<\/p>\n<\/blockquote>\n<h4>k-Nearest Neighbor\u5206\u7c7b\u5668\uff08KNN\uff09<\/h4>\n<p>\u901a\u8fc7\u6307\u5b9a\u4e0d\u540c\u7684\u8ddd\u79bb\u5ea6\u91cf\uff0c\u53ef\u5c06KNN\u5206\u7c7b\u5668\u6cdb\u5316\u5230\u4e0d\u540c\u7c7b\u578b\u7684\u6570\u636e\u4e0a\u3002<\/p>\n<p>**\u8d85\u53c2\u6570\uff1a**\u8bf8\u5982\u5bf9K\u3001\u8ddd\u79bb\u5ea6\u91cf\u7684\u9009\u62e9\u3002 \u4e0d\u662f\u673a\u5668\u5b66\u4e60\u7684\u53c2\u6570\uff0c\u800c\u662f\u4eba\u4e3a\u4e3a\u7b97\u6cd5\u8bbe\u7f6e\u7684\u3002<\/p>\n<p><strong>\u8bbe\u7f6e\u8d85\u53c2\u6570\u7684\u65b9\u6cd5\u4e00\uff1a<\/strong><\/p>\n<ul>\n<li>\n<p>\u5c06\u6570\u636e\u5206\u4e3a<strong>\u8bad\u7ec3\u96c6<\/strong>\uff0c<strong>\u9a8c\u8bc1\u96c6<\/strong>\u548c<strong>\u6d4b\u8bd5\u96c6<\/strong>\uff08train\uff0cvalidation\uff0ctest\uff09<\/p>\n<\/li>\n<li>\n<p>\u5728train\u4e0a\u7528\u4e0d\u540c\u7684\u8d85\u53c2\u6570\u6765\u8bad\u7ec3\u7b97\u6cd5\uff1b\u518d\u5728val\u4e0a\u8fdb\u884c\u8bc4\u4f30\uff0c\u9009\u62e9\u5728val\u4e0a\u8868\u73b0\u6700\u597d\u7684\u4e00\u7ec4\u8d85\u53c2\u6570\uff1b\u6700\u540e\u628a\u8fd9\u7ec4\u5728val\u4e0a\u8868\u73b0\u6700\u597d\u7684\u5206\u7c7b\u5668\u8fd0\u7528\u5728\u6d4b\u8bd5\u96c6\u4e0a\uff0c\u5f97\u5230\u7ed3\u679c\u3002<\/p>\n<blockquote>\n<p>\u6ce8\u610f\uff1a\u5fc5\u987b\u5206\u5272\u9a8c\u8bc1\u96c6\u548c\u6d4b\u8bd5\u96c6\u3002<\/p>\n<\/blockquote>\n<\/li>\n<\/ul>\n<p><strong>\u65b9\u6cd5\u4e8c\uff1a\u4ea4\u53c9\u9a8c\u8bc1<\/strong><\/p>\n<blockquote>\n<p>\u5e38\u7528\u4e8e\u5c0f\u6570\u636e\u96c6\uff0c\u5728\u6df1\u5ea6\u5b66\u4e60\u4e2d\u4e0d\u5e38\u7528\u3002<\/p>\n<\/blockquote>\n<p>\u5148\u53d6\u90e8\u5206\u6570\u636e\u4f5c\u4e3a\u6d4b\u8bd5\u96c6\uff0c\u518d\u5c06\u5269\u4f59\u8bad\u7ec3\u6570\u636e\u5e73\u5747\u5206\u6210\u591a\u4efd\uff0c\u5176\u4e2d1\u4efd\u7528\u6765\u9a8c\u8bc1\uff0c\u5176\u4f59\u7528\u6765\u8bad\u7ec3\u3002\u7136\u540e\u8f6e\u6d41\u5c06\u6bcf1\u4efd\u5f53\u505a\u9a8c\u8bc1\u96c6\uff0c\u6700\u540e\u53d6\u6240\u6709\u9a8c\u8bc1\u7ed3\u679c\u7684\u5e73\u5747\u503c\u4f5c\u4e3a\u7b97\u6cd5\u9a8c\u8bc1\u7ed3\u679c\u3002<\/p>\n<hr>\n<p>\u793a\u4f8b\uff1a<br \/> <img decoding=\"async\" src=\"https:\/\/img.mushiming.top\/app\/mushiming_com\/d173b780b6724132a037afe20ec721c3.png\" =\"\" =\"\u5728\u8fd9\u91cc\u63d2\u5165\u56fe\u7247\u63cf\u8ff0\" alt=\"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b (https:\/\/mushiming.com\/)  \u7b2c4\u5f20\" title=\"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b  \u7b2c4\u5f20-\u7a46\u4e16\u660e\u535a\u5ba2\" ><\/p>\n<p>\u8fd9\u5c31\u662f5\u4efd\u4ea4\u53c9\u9a8c\u8bc1\u5bf9k\u503c\u8c03\u4f18\u7684\u4f8b\u5b50\u3002\u9488\u5bf9\u6bcf\u4e2ak\u503c\uff0c\u5f97\u52305\u4e2a\u51c6\u786e\u7387\u7ed3\u679c\uff0c\u53d6\u5176\u5e73\u5747\u503c\uff0c\u7136\u540e\u5bf9\u4e0d\u540ck\u503c\u7684\u5e73\u5747\u8868\u73b0\u753b\u7ebf\u8fde\u63a5\u3002\u672c\u4f8b\u4e2d\uff0c\u5f53k=7\u7684\u65f6\u7b97\u6cd5\u8868\u73b0\u6700\u597d\uff08\u5bf9\u5e94\u56fe\u4e2d\u7684\u51c6\u786e\u7387\u5cf0\u503c\uff09\u3002\u5982\u679c\u6211\u4eec\u5c06\u8bad\u7ec3\u96c6\u5206\u6210\u66f4\u591a\u4efd\u6570\uff0c\u76f4\u7ebf\u4e00\u822c\u4f1a\u66f4\u52a0\u5e73\u6ed1\uff08\u566a\u97f3\u66f4\u5c11\uff09\u3002<\/p>\n<p><strong>\u7f3a\u70b9:<\/strong><\/p>\n<p>1\u3001\u6d4b\u8bd5\u65f6\u95f4\u957f<\/p>\n<p>2\u3001L1\u3001L2\u8ddd\u79bb\u4e0d\u592a\u9002\u7528\u4e8e\u6bd4\u8f83\u56fe\u50cf\uff0c\u4e0d\u9002\u5408\u8868\u793a\u56fe\u50cf\u89c6\u89c9\u611f\u77e5\u4e0a\u7684\u5dee\u5f02\u3002<\/p>\n<p>3\u3001\u7ef4\u5ea6\u707e\u96be<\/p>\n<p><strong>\u5c0f\u7ed3<\/strong><\/p>\n<p>\u7b80\u8981\u8bf4\u6765\uff1a<\/p>\n<ul>\n<li>\u4ecb\u7ecd\u4e86<strong>\u56fe\u50cf\u5206\u7c7b<\/strong>\u95ee\u9898\u3002\u5728\u8be5\u95ee\u9898\u4e2d\uff0c\u7ed9\u51fa\u4e00\u4e2a\u7531\u88ab\u6807\u6ce8\u4e86\u5206\u7c7b\u6807\u7b7e\u7684\u56fe\u50cf\u7ec4\u6210\u7684\u96c6\u5408\uff0c\u8981\u6c42\u7b97\u6cd5\u80fd\u9884\u6d4b\u6ca1\u6709\u6807\u7b7e\u7684\u56fe\u50cf\u7684\u5206\u7c7b\u6807\u7b7e\uff0c\u5e76\u6839\u636e\u7b97\u6cd5\u9884\u6d4b\u51c6\u786e\u7387\u8fdb\u884c\u8bc4\u4ef7\u3002<\/li>\n<li>\u4ecb\u7ecd\u4e86\u4e00\u4e2a\u7b80\u5355\u7684\u56fe\u50cf\u5206\u7c7b\u5668\uff1a<strong>\u6700\u8fd1\u90bb\u5206\u7c7b\u5668(Nearest Neighbor classifier)<\/strong>\u3002\u5206\u7c7b\u5668\u4e2d\u5b58\u5728\u4e0d\u540c\u7684\u8d85\u53c2\u6570(\u6bd4\u5982k\u503c\u6216\u8ddd\u79bb\u7c7b\u578b\u7684\u9009\u53d6)\uff0c\u8981\u60f3\u9009\u53d6\u597d\u7684\u8d85\u53c2\u6570\u4e0d\u662f\u4e00\u4ef6\u8f7b\u800c\u6613\u4e3e\u7684\u4e8b\u3002<\/li>\n<li>\u9009\u53d6\u8d85\u53c2\u6570\u7684\u6b63\u786e\u65b9\u6cd5\u662f\uff1a\u5c06\u539f\u59cb\u8bad\u7ec3\u96c6\u5206\u4e3a\u8bad\u7ec3\u96c6\u548c<strong>\u9a8c\u8bc1\u96c6<\/strong>\uff0c\u6211\u4eec\u5728\u9a8c\u8bc1\u96c6\u4e0a\u5c1d\u8bd5\u4e0d\u540c\u7684\u8d85\u53c2\u6570\uff0c\u6700\u540e\u4fdd\u7559\u8868\u73b0\u6700\u597d\u90a3\u4e2a\u3002<\/li>\n<li>\u5982\u679c\u8bad\u7ec3\u6570\u636e\u91cf\u4e0d\u591f\uff0c\u4f7f\u7528<strong>\u4ea4\u53c9\u9a8c\u8bc1<\/strong>\u65b9\u6cd5\uff0c\u5b83\u80fd\u5e2e\u52a9\u6211\u4eec\u5728\u9009\u53d6\u6700\u4f18\u8d85\u53c2\u6570\u7684\u65f6\u5019\u51cf\u5c11\u566a\u97f3\u3002<\/li>\n<li>\u4e00\u65e6\u627e\u5230\u6700\u4f18\u7684\u8d85\u53c2\u6570\uff0c\u5c31\u8ba9\u7b97\u6cd5\u4ee5\u8be5\u53c2\u6570\u5728\u6d4b\u8bd5\u96c6\u8dd1\u4e14\u53ea\u8dd1\u4e00\u6b21\uff0c\u5e76\u6839\u636e\u6d4b\u8bd5\u7ed3\u679c\u8bc4\u4ef7\u7b97\u6cd5\u3002<\/li>\n<li>\u6700\u8fd1\u90bb\u5206\u7c7b\u5668\u80fd\u591f\u5728CIFAR-10\u4e0a\u5f97\u5230\u5c06\u8fd140%\u7684\u51c6\u786e\u7387\u3002\u8be5\u7b97\u6cd5\u7b80\u5355\u6613\u5b9e\u73b0\uff0c\u4f46\u9700\u8981\u5b58\u50a8\u6240\u6709\u8bad\u7ec3\u6570\u636e\uff0c\u5e76\u4e14\u5728\u6d4b\u8bd5\u7684\u65f6\u5019\u8fc7\u4e8e\u8017\u8d39\u8ba1\u7b97\u80fd\u529b\u3002<\/li>\n<li>\u6700\u540e\uff0c\u6211\u4eec\u77e5\u9053\u4e86\u4ec5\u4ec5\u4f7f\u7528L1\u548cL2\u8303\u6570\u6765\u8fdb\u884c\u50cf\u7d20\u6bd4\u8f83\u662f\u4e0d\u591f\u7684\uff0c\u56fe\u50cf\u66f4\u591a\u7684\u662f\u6309\u7167\u80cc\u666f\u548c\u989c\u8272\u88ab\u5206\u7c7b\uff0c\u800c\u4e0d\u662f\u8bed\u4e49\u4e3b\u4f53\u5206\u8eab\u3002<\/li>\n<\/ul>\n<h4>\u5b9e\u9645\u5e94\u7528k-NN<\/h4>\n<p>\u5982\u679c\u4f60\u5e0c\u671b\u5c06k-NN\u5206\u7c7b\u5668\u7528\u5230\u5b9e\u5904\uff08\u6700\u597d\u522b\u7528\u5230\u56fe\u50cf\u4e0a\uff0c\u82e5\u662f\u4ec5\u4ec5\u4f5c\u4e3a\u7ec3\u624b\u8fd8\u53ef\u4ee5\u63a5\u53d7\uff09\uff0c\u90a3\u4e48\u53ef\u4ee5\u6309\u7167\u4ee5\u4e0b\u6d41\u7a0b\uff1a<\/p>\n<ol>\n<li>\u9884\u5904\u7406\u4f60\u7684\u6570\u636e\uff1a\u5bf9\u4f60\u6570\u636e\u4e2d\u7684\u7279\u5f81\u8fdb\u884c\u5f52\u4e00\u5316\uff08normalize\uff09\uff0c\u8ba9\u5176\u5177\u6709\u96f6\u5e73\u5747\u503c\uff08zero mean\uff09\u548c\u5355\u4f4d\u65b9\u5dee\uff08unit variance\uff09\u3002\u5728\u540e\u9762\u7684\u5c0f\u8282\u6211\u4eec\u4f1a\u8ba8\u8bba\u8fd9\u4e9b\u7ec6\u8282\u3002\u672c\u5c0f\u8282\u4e0d\u8ba8\u8bba\uff0c\u662f\u56e0\u4e3a\u56fe\u50cf\u4e2d\u7684\u50cf\u7d20\u90fd\u662f\u540c\u8d28\u7684\uff0c\u4e0d\u4f1a\u8868\u73b0\u51fa\u8f83\u5927\u7684\u5dee\u5f02\u5206\u5e03\uff0c\u4e5f\u5c31\u4e0d\u9700\u8981\u6807\u51c6\u5316\u5904\u7406\u4e86\u3002<\/li>\n<li>\u5982\u679c\u6570\u636e\u662f\u9ad8\u7ef4\u6570\u636e\uff0c\u8003\u8651\u4f7f\u7528\u964d\u7ef4\u65b9\u6cd5\uff0c\u6bd4\u5982PCA(wiki ref, CS229ref, blog ref)\u6216\u968f\u673a\u6295\u5f71\u3002<\/li>\n<li>\u5c06\u6570\u636e\u968f\u673a\u5206\u5165\u8bad\u7ec3\u96c6\u548c\u9a8c\u8bc1\u96c6\u3002\u6309\u7167\u4e00\u822c\u89c4\u5f8b\uff0c70%-90% \u6570\u636e\u4f5c\u4e3a\u8bad\u7ec3\u96c6\u3002\u8fd9\u4e2a\u6bd4\u4f8b\u6839\u636e\u7b97\u6cd5\u4e2d\u6709\u591a\u5c11\u8d85\u53c2\u6570\uff0c\u4ee5\u53ca\u8fd9\u4e9b\u8d85\u53c2\u6570\u5bf9\u4e8e\u7b97\u6cd5\u7684\u9884\u671f\u5f71\u54cd\u6765\u51b3\u5b9a\u3002\u5982\u679c\u9700\u8981\u9884\u6d4b\u7684\u8d85\u53c2\u6570\u5f88\u591a\uff0c\u90a3\u4e48\u5c31\u5e94\u8be5\u4f7f\u7528\u66f4\u5927\u7684\u9a8c\u8bc1\u96c6\u6765\u6709\u6548\u5730\u4f30\u8ba1\u5b83\u4eec\u3002\u5982\u679c\u62c5\u5fc3\u9a8c\u8bc1\u96c6\u6570\u91cf\u4e0d\u591f\uff0c\u90a3\u4e48\u5c31\u5c1d\u8bd5\u4ea4\u53c9\u9a8c\u8bc1\u65b9\u6cd5\u3002\u5982\u679c\u8ba1\u7b97\u8d44\u6e90\u8db3\u591f\uff0c\u4f7f\u7528\u4ea4\u53c9\u9a8c\u8bc1\u603b\u662f\u66f4\u52a0\u5b89\u5168\u7684\uff08\u4efd\u6570\u8d8a\u591a\uff0c\u6548\u679c\u8d8a\u597d\uff0c\u4e5f\u66f4\u8017\u8d39\u8ba1\u7b97\u8d44\u6e90\uff09\u3002<\/li>\n<li>\u5728\u9a8c\u8bc1\u96c6\u4e0a\u8c03\u4f18\uff0c\u5c1d\u8bd5\u8db3\u591f\u591a\u7684k\u503c\uff0c\u5c1d\u8bd5L1\u548cL2\u4e24\u79cd\u8303\u6570\u8ba1\u7b97\u65b9\u5f0f\u3002<\/li>\n<li>\u5982\u679c\u5206\u7c7b\u5668\u8dd1\u5f97\u592a\u6162\uff0c\u5c1d\u8bd5\u4f7f\u7528Approximate Nearest Neighbor\u5e93\uff08\u6bd4\u5982FLANN\uff09\u6765\u52a0\u901f\u8fd9\u4e2a\u8fc7\u7a0b\uff0c\u5176\u4ee3\u4ef7\u662f\u964d\u4f4e\u4e00\u4e9b\u51c6\u786e\u7387\u3002<\/li>\n<li>\u5bf9\u6700\u4f18\u7684\u8d85\u53c2\u6570\u505a\u8bb0\u5f55\u3002\u8bb0\u5f55\u6700\u4f18\u53c2\u6570\u540e\uff0c\u662f\u5426\u5e94\u8be5\u8ba9\u4f7f\u7528\u6700\u4f18\u53c2\u6570\u7684\u7b97\u6cd5\u5728\u5b8c\u6574\u7684\u8bad\u7ec3\u96c6\u4e0a\u8fd0\u884c\u5e76\u518d\u6b21\u8bad\u7ec3\u5462\uff1f\u56e0\u4e3a\u5982\u679c\u628a\u9a8c\u8bc1\u96c6\u91cd\u65b0\u653e\u56de\u5230\u8bad\u7ec3\u96c6\u4e2d\uff08\u81ea\u7136\u8bad\u7ec3\u96c6\u7684\u6570\u636e\u91cf\u5c31\u53c8\u53d8\u5927\u4e86\uff09\uff0c\u6709\u53ef\u80fd\u6700\u4f18\u53c2\u6570\u53c8\u4f1a\u6709\u6240\u53d8\u5316\u3002\u5728\u5b9e\u8df5\u4e2d\uff0c<strong>\u4e0d\u8981\u8fd9\u6837\u505a<\/strong>\u3002\u5343\u4e07\u4e0d\u8981\u5728\u6700\u7ec8\u7684\u5206\u7c7b\u5668\u4e2d\u4f7f\u7528\u9a8c\u8bc1\u96c6\u6570\u636e\uff0c\u8fd9\u6837\u505a\u4f1a\u7834\u574f\u5bf9\u4e8e\u6700\u4f18\u53c2\u6570\u7684\u4f30\u8ba1\u3002<strong>\u76f4\u63a5\u4f7f\u7528\u6d4b\u8bd5\u96c6\u6765\u6d4b\u8bd5\u7528\u6700\u4f18\u53c2\u6570\u8bbe\u7f6e\u597d\u7684\u6700\u4f18\u6a21\u578b<\/strong>\uff0c\u5f97\u5230\u6d4b\u8bd5\u96c6\u6570\u636e\u7684\u5206\u7c7b\u51c6\u786e\u7387\uff0c\u5e76\u4ee5\u6b64\u4f5c\u4e3a\u4f60\u7684kNN\u5206\u7c7b\u5668\u5728\u8be5\u6570\u636e\u4e0a\u7684\u6027\u80fd\u8868\u73b0\u3002<\/li>\n<\/ol>\n<h3>2.3 \u7ebf\u6027\u5206\u7c7b\u5668<\/h3>\n<p><strong>\u51fd\u6570\u5f62\u5f0f:<\/strong><\/p>\n<ul>\n<li>x\uff1a\u8f93\u5165\u6570\u636e\uff1bW\uff1a\u53c2\u6570\u6216\u8005\u6743\u91cd\uff1bb\uff1a\u504f\u597d\u5143\u7d20<\/li>\n<li>\u901a\u8fc7\u4e00\u4e2a\u53c2\u6570\u77e9\u9635<em>W<\/em>\u4e58\u4ee5\u6570\u636e\u77e9\u9635<em>X<\/em>\u518d\u52a0\u4e0a\u504f\u7f6e <em>b<\/em>b \u5f97\u5230\u9884\u6d4b\u7ed3\u679c\uff1a<\/li>\n<\/ul>\n<p><span class=\"katex--display\"><span class=\"katex-display\"><span class=\"katex\"><span class=\"katex-mathml\"> f ( x , W ) = W x + b f(x,W)=Wx+b <\/span><span class=\"katex-html\"><span class=\"base\"><span class=\"strut\" style=\"height: 1em; vertical-align: -0.25em;\"><\/span><span style=\"margin-right: 0.10764em;\" class=\"mord mathdefault\">f<\/span><span class=\"mopen\">(<\/span><span class=\"mord mathdefault\">x<\/span><span class=\"mpunct\">,<\/span><span class=\"mspace\" style=\"margin-right: 0.166667em;\"><\/span><span style=\"margin-right: 0.13889em;\" class=\"mord mathdefault\">W<\/span><span class=\"mclose\">)<\/span><span class=\"mspace\" style=\"margin-right: 0.277778em;\"><\/span><span class=\"mrel\">=<\/span><span class=\"mspace\" style=\"margin-right: 0.277778em;\"><\/span><\/span><span class=\"base\"><span class=\"strut\" style=\"height: 0.76666em; vertical-align: -0.08333em;\"><\/span><span style=\"margin-right: 0.13889em;\" class=\"mord mathdefault\">W<\/span><span class=\"mord mathdefault\">x<\/span><span class=\"mspace\" style=\"margin-right: 0.222222em;\"><\/span><span class=\"mbin\">+<\/span><span class=\"mspace\" style=\"margin-right: 0.222222em;\"><\/span><\/span><span class=\"base\"><span class=\"strut\" style=\"height: 0.69444em; vertical-align: 0em;\"><\/span><span class=\"mord mathdefault\">b<\/span><\/span><\/span><\/span><\/span><\/span><\/p>\n<p>\u53c2\u6570\u5316\u65b9\u6cd5\uff0c\u603b\u7ed3\u5bf9\u8bad\u7ec3\u6570\u636e\u7684\u8ba4\u8bc6\u5e76\u7528\u5230W\u4e2d\uff0c\u6d4b\u8bd5\u65f6\u4e0d\u518d\u9700\u8981\u8bad\u7ec3\u6570\u636e\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/img.mushiming.top\/app\/mushiming_com\/c557f9ee41964519b912168aa82d6ac5.png\" =\"\" =\"\" alt=\"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b (https:\/\/mushiming.com\/)  \u7b2c5\u5f20\" title=\"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b  \u7b2c5\u5f20-\u7a46\u4e16\u660e\u535a\u5ba2\" ><\/p>\n<p><strong>\u91cd\u70b9<\/strong>\u8fd9\u91cc\u9762\u6ce8\u610f\u884c\u7684\u989c\u8272\uff0cW\u7684\u6bcf\u4e00\u884c\u90fd\u5bf9\u5e94\u4e86\u4e00\u4e2a\u5206\u7c7b\u3002\u628a\u8bad\u7ec3\u597d\u7684\u5355\u72ec\u7684\u6bcf\u4e00\u884c\u7684\u63d0\u53d6\u51fa\u6765\uff0c\u8fd8\u539f\u4e3a\u56fe\u50cf\uff0c\u5c31\u53ef\u4ee5\u770b\u5230\u7ebf\u6027\u5206\u7c7b\u5668\u662f\u5982\u4f55\u6839\u636e\u4ec0\u4e48\u7279\u5f81\u8fdb\u884c\u5206\u7c7b\u7684\u3002<br \/> \u8fd9\u91cc\u8981\u6ce8\u610f\u4e00\u4e2a\u6982\u5ff5\uff1a<strong>\u56fe\u50cf\u5206\u7c7b\u6a21\u677f<\/strong>\uff0c\u8fd9\u4e2a\u6a21\u677f\u5728CNN\u91cc\u9762\u548c\u6bcf\u4e2a\u56fe\u7247\u8fdb\u884c\u70b9\u4e58\u540e\uff0c\u76f8\u5e94\u5206\u7c7b\u5c31\u4f1a\u5f97\u5230\u6bd4\u8f83\u5927\u7684\u7ed3\u679c\u3002<\/p>\n<blockquote>\n<p>\u7ebf\u6027\u5206\u7c7b\u7a76\u7adf\u5728\u505a\u4e9b\u4ec0\u4e48, \u5982\u4f55\u89e3\u91ca\u7ebf\u6027\u5206\u7c7b\u5668\u5462\uff1f<\/p>\n<p>\u5176\u5b9e\u53c2\u6570 W \u7684\u6bcf\u4e00\u884c\u5bf9\u5e94\u7684\u662f\u4e00\u4e2a\u7c7b\u522b\u7684\u5206\u7c7b\u5668\uff0c\u6539\u53d8 W \u7684\u4e00\u884c\u76f8\u5f53\u4e8e\u5411\u4e0d\u540c\u65b9\u5411\u65cb\u8f6c\u5728\u50cf\u7d20\u7a7a\u95f4\u7684\u7ebf\uff0c\u6539\u53d8 b \u76f8\u5f53\u4e8e\u5bf9\u8fd9\u6761\u7ebf\u505a\u5e73\u79fb\u3002<\/p>\n<ul>\n<li>\u53ef\u4ee5\u628a W \u770b\u6210\u662f\u6a21\u677f\u5339\u914d\uff0cW \u7684\u6bcf\u4e00\u884c\u76f8\u5f53\u4e8e\u5bf9\u5e94\u7c7b\u7684\u4e00\u4e2a\u6a21\u677f\uff0c\u901a\u8fc7\u5185\u79ef\u7684\u5f62\u5f0f\u53bb\u627e\u5230\u6700\u9ad8\u7684\u5206\u6570(\u5bf9\u5e94\u7684\u7c7b\u522b)\u3002<\/li>\n<li>\u4e5f\u53ef\u4ee5\u8ba4\u4e3a\u7ebf\u6027\u5206\u7c7b\u4ecd\u7136\u5728\u6c42\u89e3\u6700\u90bb\u8fd1\uff0c\u53ea\u662f\u8fd9\u6b21\u4e0d\u662f\u5c06\u6d4b\u8bd5\u56fe\u50cf\u4e0e\u8bad\u7ec3\u96c6\u505a\u5185\u79ef\uff0c\u800c\u662f\u4e0e W \u7684\u4e00\u884c\u505a\u5185\u79ef\uff0c\u4ee5\u5185\u79ef\u6765\u4f5c\u4e3a\u7c7b\u6a21\u677f\u4e0e\u6d4b\u8bd5\u56fe\u7247\u7684\u8ddd\u79bb\uff0c\u800c\u4e0d\u662f\u4f7f\u7528 L1 \u6216 L2 distance\u3002<\/li>\n<\/ul>\n<\/blockquote>\n<p><font size=\"5\"><strong>\u7ebf\u6027\u5206\u7c7b\u5668\u7684\u9650\u5236<\/strong><\/font><\/p>\n<p>\u7ebf\u6027\u5206\u7c7b\u5668\u65e0\u6cd5\u5212\u5206multi model data\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/img.mushiming.top\/app\/mushiming_com\/d430d2dd3b2a4961b247a092325ac4c3.png\" =\"\" =\"\" alt=\"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b (https:\/\/mushiming.com\/)  \u7b2c6\u5f20\" title=\"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b  \u7b2c6\u5f20-\u7a46\u4e16\u660e\u535a\u5ba2\" ><\/p>\n","protected":false},"excerpt":{"rendered":"2017CS231n\u5b66\u4e60\u7b14\u8bb01-\u56fe\u50cf\u5206\u7c7b\u6587\u7ae0\u76ee\u5f552\u56fe\u50cf\u5206\u7c7b2.1\u6570\u636e\u9a71\u52a8\u65b9\u6cd5\u7b2c\u4e00\u4e2a\u5206\u7c7b\u5668:\u6700\u8fd1\u90bb\u5206\u7c7b\u5668L1\u8ddd\u79bb\uff08\u66fc\u54c8\u987f\u8ddd\u79bb\uff09\u6700\u8fd1\u90bb\u5206\u7c7b\u5668...","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[],"tags":[],"_links":{"self":[{"href":"https:\/\/mushiming.com\/wp-json\/wp\/v2\/posts\/8552"}],"collection":[{"href":"https:\/\/mushiming.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/mushiming.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/mushiming.com\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/mushiming.com\/wp-json\/wp\/v2\/comments?post=8552"}],"version-history":[{"count":0,"href":"https:\/\/mushiming.com\/wp-json\/wp\/v2\/posts\/8552\/revisions"}],"wp:attachment":[{"href":"https:\/\/mushiming.com\/wp-json\/wp\/v2\/media?parent=8552"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/mushiming.com\/wp-json\/wp\/v2\/categories?post=8552"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/mushiming.com\/wp-json\/wp\/v2\/tags?post=8552"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}