{"id":253,"date":"2018-03-16T18:33:54","date_gmt":"2018-03-16T18:33:54","guid":{"rendered":""},"modified":"2018-03-16T18:33:54","modified_gmt":"2018-03-16T18:33:54","slug":"","status":"publish","type":"post","link":"http:\/\/weizn.net\/?p=253","title":{"rendered":"\u591a\u5c42\u611f\u77e5\u673a\uff08MLP\uff09\u539f\u7406\u7b80\u4ecb"},"content":{"rendered":"<div id=\"ez-toc-container\" class=\"ez-toc-v2_0_17 counter-hierarchy\">\n<div class=\"ez-toc-title-container\">\n<p class=\"ez-toc-title\">\u76ee\u5f55<\/p>\n<span class=\"ez-toc-title-toggle\"><a class=\"ez-toc-pull-right ez-toc-btn ez-toc-btn-xs ez-toc-btn-default ez-toc-toggle\" style=\"display: none;\"><i class=\"ez-toc-glyphicon ez-toc-icon-toggle\"><\/i><\/a><\/span><\/div>\n<nav><ul class=\"ez-toc-list ez-toc-list-level-1\"><li class=\"ez-toc-page-1 ez-toc-heading-level-1\"><a class=\"ez-toc-link ez-toc-heading-1\" href=\"http:\/\/weizn.net\/?p=253\/#sklearnneural_networkMLPClassifier\" title=\"\n\tsklearn.neural_network.MLPClassifier \n\">\n\tsklearn.neural_network.MLPClassifier \n<\/a><ul class=\"ez-toc-list-level-2\"><li class=\"ez-toc-heading-level-2\"><a class=\"ez-toc-link ez-toc-heading-2\" href=\"http:\/\/weizn.net\/?p=253\/#%E6%A6%82%E8%BF%B0\" title=\"\n\t\u6982\u8ff0 \n\">\n\t\u6982\u8ff0 \n<\/a><\/li><li class=\"ez-toc-page-1 ez-toc-heading-level-2\"><a class=\"ez-toc-link ez-toc-heading-3\" href=\"http:\/\/weizn.net\/?p=253\/#%E7%A5%9E%E7%BB%8F%E7%BD%91%E7%BB%9C%E6%A8%A1%E5%9E%8B\" title=\"\n\t\u795e\u7ecf\u7f51\u7edc\u6a21\u578b \n\">\n\t\u795e\u7ecf\u7f51\u7edc\u6a21\u578b \n<\/a><\/li><\/ul><\/li><\/ul><\/nav><\/div>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">\u591a\u5c42\u611f\u77e5\u673a\uff08MLP\uff0cMultilayer Perceptron\uff09\u4e5f\u53eb\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\uff08ANN\uff0cArtificial Neural Network\uff09\uff0c\u9664\u4e86\u8f93\u5165\u8f93\u51fa\u5c42\uff0c\u5b83\u4e2d\u95f4\u53ef\u4ee5\u6709\u591a\u4e2a\u9690\u5c42\uff0c\u6700\u7b80\u5355\u7684MLP\u53ea\u542b\u4e00\u4e2a\u9690\u5c42\uff0c\u5373\u4e09\u5c42\u7684\u7ed3\u6784\uff0c\u5982\u4e0b\u56fe\uff1a<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<img decoding=\"async\" src=\"https:\/\/img-blog.csdn.net\/20150128033221168?watermark\/2\/text\/aHR0cDovL2Jsb2cuY3Nkbi5uZXQvdTAxMjE2MjYxMw==\/font\/5a6L5L2T\/fontsize\/400\/fill\/I0JBQkFCMA==\/dissolve\/70\/gravity\/Center\" alt=\"\" style=\"max-width:900px;height:auto;\" \/>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">&nbsp;<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">\u4ece\u4e0a\u56fe\u53ef\u4ee5\u770b\u5230\uff0c\u591a\u5c42\u611f\u77e5\u673a\u5c42\u4e0e\u5c42\u4e4b\u95f4\u662f\u5168\u8fde\u63a5\u7684\uff08\u5168\u8fde\u63a5\u7684\u610f\u601d\u5c31\u662f\uff1a\u4e0a\u4e00\u5c42\u7684\u4efb\u4f55\u4e00\u4e2a\u795e\u7ecf\u5143\u4e0e\u4e0b\u4e00\u5c42\u7684\u6240\u6709\u795e\u7ecf\u5143\u90fd\u6709\u8fde\u63a5\uff09\u3002\u591a\u5c42\u611f\u77e5\u673a\u6700\u5e95\u5c42\u662f\u8f93\u5165\u5c42\uff0c\u4e2d\u95f4\u662f\u9690\u85cf\u5c42\uff0c\u6700\u540e\u662f\u8f93\u51fa\u5c42\u3002<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">&nbsp;<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">\u8f93\u5165\u5c42\u6ca1\u4ec0\u4e48\u597d\u8bf4\uff0c\u4f60\u8f93\u5165\u4ec0\u4e48\u5c31\u662f\u4ec0\u4e48\uff0c\u6bd4\u5982\u8f93\u5165\u662f\u4e00\u4e2an\u7ef4\u5411\u91cf\uff0c\u5c31\u6709n\u4e2a\u795e\u7ecf\u5143\u3002<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">\u9690\u85cf\u5c42\u7684\u795e\u7ecf\u5143\u600e\u4e48\u5f97\u6765\uff1f\u9996\u5148\u5b83\u4e0e\u8f93\u5165\u5c42\u662f\u5168\u8fde\u63a5\u7684\uff0c\u5047\u8bbe\u8f93\u5165\u5c42\u7528\u5411\u91cfX\u8868\u793a\uff0c\u5219\u9690\u85cf\u5c42\u7684\u8f93\u51fa\u5c31\u662f<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">f(W1X+b1)\uff0cW1\u662f\u6743\u91cd\uff08\u4e5f\u53eb\u8fde\u63a5\u7cfb\u6570\uff09\uff0cb1\u662f\u504f\u7f6e\uff0c\u51fd\u6570f \u53ef\u4ee5\u662f\u5e38\u7528\u7684sigmoid\u51fd\u6570\u6216\u8005tanh\u51fd\u6570\uff1a<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">&nbsp;<\/span>\n<\/p>\n<div>\n\t<img decoding=\"async\" src=\"https:\/\/img-blog.csdn.net\/20150128033821825?watermark\/2\/text\/aHR0cDovL2Jsb2cuY3Nkbi5uZXQvdTAxMjE2MjYxMw==\/font\/5a6L5L2T\/fontsize\/400\/fill\/I0JBQkFCMA==\/dissolve\/70\/gravity\/Center\" alt=\"\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp; &nbsp; &nbsp; &nbsp;<\/span><img decoding=\"async\" src=\"https:\/\/img-blog.csdn.net\/20150128033840093?watermark\/2\/text\/aHR0cDovL2Jsb2cuY3Nkbi5uZXQvdTAxMjE2MjYxMw==\/font\/5a6L5L2T\/fontsize\/400\/fill\/I0JBQkFCMA==\/dissolve\/70\/gravity\/Center\" alt=\"\" style=\"max-width:900px;height:auto;\" \/>\n<\/div>\n<div>\n\t<span style=\"font-size:16px;\">&nbsp;<\/span>\n<\/div>\n<div>\n\t<span style=\"font-size:16px;\">&nbsp;<\/span>\n<\/div>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">\u6700\u540e\u5c31\u662f\u8f93\u51fa\u5c42\uff0c\u8f93\u51fa\u5c42\u4e0e\u9690\u85cf\u5c42\u662f\u4ec0\u4e48\u5173\u7cfb\uff1f\u5176\u5b9e\u9690\u85cf\u5c42\u5230\u8f93\u51fa\u5c42\u53ef\u4ee5\u770b\u6210\u662f\u4e00\u4e2a\u591a\u7c7b\u522b\u7684\u903b\u8f91\u56de\u5f52\uff0c\u4e5f\u5373softmax\u56de\u5f52\uff0c\u6240\u4ee5\u8f93\u51fa\u5c42\u7684\u8f93\u51fa\u5c31\u662fsoftmax(W2X1+b2)\uff0cX1\u8868\u793a\u9690\u85cf\u5c42\u7684\u8f93\u51faf(W1X+b1)\u3002<\/span>\n<\/p>\n<h1 style=\"line-height:1.5;margin:10px 0px;\"><span class=\"ez-toc-section\" id=\"sklearnneural_networkMLPClassifier\"><\/span>\n\t<a class=\"reference internal\" title=\"sklearn.neural_network\" href=\"http:\/\/scikit-learn.org\/dev\/modules\/classes.html#module-sklearn.neural_network\"><code class=\"xref py py-mod docutils literal\"><span style=\"font-size:16px;\">sklearn.neural_network<\/span><\/code><\/a><span style=\"font-size:16px;\">.MLPClassifier<\/span><br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h1>\n<dl class=\"class\">\n<dt id=\"sklearn.neural_network.MLPClassifier\">\n\t\t<span style=\"font-size:16px;\">class&nbsp;<\/span><code class=\"descclassname\"><span style=\"font-size:16px;\">sklearn.neural_network.<\/span><\/code><code class=\"descname\"><span style=\"font-size:16px;\">MLPClassifier<\/span><\/code><span style=\"font-size:16px;\">(<\/span><span style=\"font-size:16px;\">hidden_layer_sizes=(100<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">)<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">activation=\u2019relu\u2019<\/span><span style=\"font-size:16px;\">,<\/span><span style=\"font-size:16px;\">&nbsp;<\/span><span style=\"font-size:16px;\">solver=\u2019adam\u2019<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">alpha=0.0001<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">batch_size=\u2019auto\u2019<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">learning_rate=\u2019constant\u2019<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">learning_rate_init=0.001<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">power_t=0.5<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">max_iter=200<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">shuffle=True<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">random_state=None<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">tol=0.0001<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">verbose=False<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">warm_start=False<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">momentum=0.9<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">nesterovs_momentum=True<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">early_stopping=False<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">validation_fraction=0.1<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">beta_1=0.9<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">beta_2=0.999<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">epsilon=1e-08<\/span><span style=\"font-size:16px;\">,&nbsp;<\/span><span style=\"font-size:16px;\">n_iter_no_change=10<\/span><span style=\"font-size:16px;\">)<\/span><a class=\"reference external\" href=\"https:\/\/github.com\/scikit-learn\/scikit-learn\/blob\/5fcf6f4\/sklearn\/neural_network\/multilayer_perceptron.py#L682\"><span style=\"font-size:16px;\">[source]<\/span><\/a>\n\t<\/dt>\n<dd>\n<p style=\"margin:10px auto;\">\n\t\t\t<span style=\"font-size:16px;\">Multi-layer Perceptron classifier.<\/span>\n\t\t<\/p>\n<p style=\"margin:10px auto;\">\n\t\t\t<span style=\"font-size:16px;\">This model optimizes the log-loss function using LBFGS or stochastic gradient descent.<\/span>\n\t\t<\/p>\n<div class=\"versionadded\">\n<p style=\"margin:10px auto;\">\n\t\t\t\t<span style=\"font-size:16px;\">New in version 0.18.<\/span>\n\t\t\t<\/p>\n<\/p><\/div>\n<table class=\"docutils field-list\" frame=\"void\" rules=\"none\" style=\"border:1px solid silver;word-break:break-word;\">\n<colgroup>\n<col class=\"field-name\" \/>\n<col class=\"field-body\" \/><\/colgroup>\n<tbody valign=\"top\">\n<tr class=\"field-odd field\">\n<th class=\"field-name\" style=\"border:1px solid silver;border-collapse:collapse;padding:8px 14px;\">\n\t\t\t\t\t\t<span style=\"font-size:16px;\">Parameters:<\/span>\n\t\t\t\t\t<\/th>\n<td class=\"field-body\" style=\"border:1px solid silver;border-collapse:collapse;padding:8px 14px;\">\n<dl class=\"first docutils\">\n<dt>\n\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">hidden_layer_sizes<\/span><span style=\"font-size:16px;\">&nbsp;<\/span><span style=\"font-size:16px;\">:&nbsp;<\/span><span style=\"font-size:16px;\">tuple, length = n_layers &#8211; 2, default (100,)<\/span>\n\t\t\t\t\t\t\t<\/dt>\n<dd>\n<p class=\"first last\" style=\"margin:10px auto;\">\n\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">The ith element represents the number of neurons in the ith hidden layer.<\/span>\n\t\t\t\t\t\t\t\t<\/p>\n<\/dd>\n<dt>\n\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">activation<\/span><span style=\"font-size:16px;\">&nbsp;<\/span><span style=\"font-size:16px;\">:&nbsp;<\/span><span style=\"font-size:16px;\">{\u2018identity\u2019, \u2018logistic\u2019, \u2018tanh\u2019, \u2018relu\u2019}, default \u2018relu\u2019<\/span>\n\t\t\t\t\t\t\t<\/dt>\n<dd>\n<p class=\"first\" style=\"margin:10px auto;\">\n\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">Activation function for the hidden layer.<\/span>\n\t\t\t\t\t\t\t\t<\/p>\n<ul class=\"last simple\" style=\"list-style-type:circle;margin-left:30px;padding-left:0px;\">\n<li style=\"list-style-type:disc;margin-left:10px;\">\n\t\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">\u2018identity\u2019, no-op activation, useful to implement linear bottleneck, returns f(x) = x<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n<li style=\"list-style-type:disc;margin-left:10px;\">\n\t\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">\u2018logistic\u2019, the logistic sigmoid function, returns f(x) = 1 \/ (1 + exp(-x)).<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n<li style=\"list-style-type:disc;margin-left:10px;\">\n\t\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">\u2018tanh\u2019, the hyperbolic tan function, returns f(x) = tanh(x).<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n<li style=\"list-style-type:disc;margin-left:10px;\">\n\t\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">\u2018relu\u2019, the rectified linear unit function, returns f(x) = max(0, x)<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n<\/ul>\n<\/dd>\n<dt>\n\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">solver<\/span><span style=\"font-size:16px;\">&nbsp;<\/span><span style=\"font-size:16px;\">:&nbsp;<\/span><span style=\"font-size:16px;\">{\u2018lbfgs\u2019, \u2018sgd\u2019, \u2018adam\u2019}, default \u2018adam\u2019<\/span>\n\t\t\t\t\t\t\t<\/dt>\n<dd>\n<p class=\"first\" style=\"margin:10px auto;\">\n\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">The solver for weight optimization.<\/span>\n\t\t\t\t\t\t\t\t<\/p>\n<ul class=\"simple\" style=\"list-style-type:circle;margin-left:30px;padding-left:0px;\">\n<li style=\"list-style-type:disc;margin-left:10px;\">\n\t\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">\u2018lbfgs\u2019 is an optimizer in the family of quasi-Newton methods.<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n<li style=\"list-style-type:disc;margin-left:10px;\">\n\t\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">\u2018sgd\u2019 refers to stochastic gradient descent.<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n<li style=\"list-style-type:disc;margin-left:10px;\">\n\t\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">\u2018adam\u2019 refers to a stochastic gradient-based optimizer proposed by Kingma, Diederik, and Jimmy Ba<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n<\/ul>\n<p class=\"last\" style=\"margin:10px auto;\">\n\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">Note: The default solver \u2018adam\u2019 works pretty well on relatively large datasets (with thousands of training samples or more) in terms of both training time and validation score. For small datasets, however, \u2018lbfgs\u2019 can converge faster and perform better.<\/span>\n\t\t\t\t\t\t\t\t<\/p>\n<\/dd>\n<dt>\n\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">alpha<\/span><span style=\"font-size:16px;\">&nbsp;<\/span><span style=\"font-size:16px;\">:&nbsp;<\/span><span style=\"font-size:16px;\">float, optional, default 0.0001<\/span>\n\t\t\t\t\t\t\t<\/dt>\n<dd>\n<p class=\"first last\" style=\"margin:10px auto;\">\n\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">L2 penalty (regularization term) parameter.<\/span>\n\t\t\t\t\t\t\t\t<\/p>\n<\/dd>\n<dt>\n\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">batch_size<\/span><span style=\"font-size:16px;\">&nbsp;<\/span><span style=\"font-size:16px;\">:&nbsp;<\/span><span style=\"font-size:16px;\">int, optional, default \u2018auto\u2019<\/span>\n\t\t\t\t\t\t\t<\/dt>\n<dd>\n<p class=\"first last\" style=\"margin:10px auto;\">\n\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">Size of minibatches for stochastic optimizers. If the solver is \u2018lbfgs\u2019, the classifier will not use minibatch. When set to \u201cauto\u201d,&nbsp;<\/span><cite><span style=\"font-size:16px;\">batch_size=min(200, n_samples)<\/span><\/cite>\n\t\t\t\t\t\t\t\t<\/p>\n<\/dd>\n<dt>\n\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">learning_rate<\/span><span style=\"font-size:16px;\">&nbsp;<\/span><span style=\"font-size:16px;\">:&nbsp;<\/span><span style=\"font-size:16px;\">{\u2018constant\u2019, \u2018invscaling\u2019, \u2018adaptive\u2019}, default \u2018constant\u2019<\/span>\n\t\t\t\t\t\t\t<\/dt>\n<dd>\n<p class=\"first\" style=\"margin:10px auto;\">\n\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">Learning rate schedule for weight updates.<\/span>\n\t\t\t\t\t\t\t\t<\/p>\n<ul class=\"simple\" style=\"list-style-type:circle;margin-left:30px;padding-left:0px;\">\n<li style=\"list-style-type:disc;margin-left:10px;\">\n\t\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">\u2018constant\u2019 is a constant learning rate given by \u2018learning_rate_init\u2019.<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n<li style=\"list-style-type:disc;margin-left:10px;\">\n\t\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">\u2018invscaling\u2019 gradually decreases the learning rate&nbsp;<\/span><code class=\"docutils literal\"><span style=\"font-size:16px;\">learning_rate_<\/span><\/code><span style=\"font-size:16px;\">&nbsp;at each time step \u2018t\u2019 using an inverse scaling exponent of \u2018power_t\u2019. effective_learning_rate = learning_rate_init \/ pow(t, power_t)<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n<li style=\"list-style-type:disc;margin-left:10px;\">\n\t\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">\u2018adaptive\u2019 keeps the learning rate constant to \u2018learning_rate_init\u2019 as long as training loss keeps decreasing. Each time two consecutive epochs fail to decrease training loss by at least tol, or fail to increase validation score by at least tol if \u2018early_stopping\u2019 is on, the current learning rate is divided by 5.<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n<\/ul>\n<p class=\"last\" style=\"margin:10px auto;\">\n\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">Only used when&nbsp;<\/span><code class=\"docutils literal\"><span style=\"font-size:16px;\">solver='sgd'<\/span><\/code><span style=\"font-size:16px;\">.<\/span>\n\t\t\t\t\t\t\t\t<\/p>\n<\/dd>\n<dt>\n\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">learning_rate_init<\/span><span style=\"font-size:16px;\">&nbsp;<\/span><span style=\"font-size:16px;\">:&nbsp;<\/span><span style=\"font-size:16px;\">double, optional, default 0.001<\/span>\n\t\t\t\t\t\t\t<\/dt>\n<dd>\n<p class=\"first last\" style=\"margin:10px auto;\">\n\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">The initial learning rate used. It controls the step-size in updating the weights. Only used when solver=\u2019sgd\u2019 or \u2018adam\u2019.<\/span>\n\t\t\t\t\t\t\t\t<\/p>\n<\/dd>\n<dt>\n\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">power_t<\/span><span style=\"font-size:16px;\">&nbsp;<\/span><span style=\"font-size:16px;\">:&nbsp;<\/span><span style=\"font-size:16px;\">double, optional, default 0.5<\/span>\n\t\t\t\t\t\t\t<\/dt>\n<dd>\n<p class=\"first last\" style=\"margin:10px auto;\">\n\t\t\t\t\t\t\t\t\t<span style=\"font-size:16px;\">The exponent for inverse scaling learning rate. It is used in updating effective learning rate when the learning_rate is set to \u2018invscaling\u2019. Only used when solver=\u2019sgd\u2019.<\/span>\n\t\t\t\t\t\t\t\t<\/p>\n<\/dd>\n<\/dl>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/dd>\n<\/dl>\n<h2 style=\"line-height:1.5;margin:10px 0px;\"><span class=\"ez-toc-section\" id=\"%E6%A6%82%E8%BF%B0\"><\/span>\n\t<span style=\"font-size:16px;\">\u6982\u8ff0<\/span><br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">\u4ee5\u76d1\u7763\u5b66\u4e60\u4e3a\u4f8b\uff0c\u5047\u8bbe\u6211\u4eec\u6709\u8bad\u7ec3\u6837\u672c\u96c6&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/7\/0\/e\/70ebbf3d401302b5d148530b986f0602.png\" alt=\"\\textstyle (x(^ i),y(^ i))\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u90a3\u4e48\u795e\u7ecf\u7f51\u7edc\u7b97\u6cd5\u80fd\u591f\u63d0\u4f9b\u4e00\u79cd\u590d\u6742\u4e14\u975e\u7ebf\u6027\u7684\u5047\u8bbe\u6a21\u578b&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/5\/8\/d\/58d3a4fe4ad68b333b180071dd46db82.png\" alt=\"\\textstyle h_{W,b}(x)\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u5b83\u5177\u6709\u53c2\u6570&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/7\/c\/9\/7c9aa03f5258ecf79556ba374d7eb2cd.png\" alt=\"\\textstyle W, b\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u53ef\u4ee5\u4ee5\u6b64\u53c2\u6570\u6765\u62df\u5408\u6211\u4eec\u7684\u6570\u636e\u3002<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\"> \u4e3a\u4e86\u63cf\u8ff0\u795e\u7ecf\u7f51\u7edc\uff0c\u6211\u4eec\u5148\u4ece\u6700\u7b80\u5355\u7684\u795e\u7ecf\u7f51\u7edc\u8bb2\u8d77\uff0c\u8fd9\u4e2a\u795e\u7ecf\u7f51\u7edc\u4ec5\u7531\u4e00\u4e2a\u201c\u795e\u7ecf\u5143\u201d\u6784\u6210\uff0c\u4ee5\u4e0b\u5373\u662f\u8fd9\u4e2a\u201c\u795e\u7ecf\u5143\u201d\u7684\u56fe\u793a\uff1a<\/span>\n<\/p>\n<div class=\"center\">\n<div class=\"floatnone\">\n\t\t<a class=\"image\" href=\"http:\/\/ufldl.stanford.edu\/wiki\/index.php\/File:SingleNeuron.png\" data-rel=\"penci-gallery-image-content\" ><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/thumb\/3\/3d\/SingleNeuron.png\/300px-SingleNeuron.png\" alt=\"SingleNeuron.png\" width=\"300\" height=\"148\" style=\"max-width:900px;height:auto;\" \/><\/a>\n\t<\/div>\n<\/div>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\"> \u8fd9\u4e2a\u201c\u795e\u7ecf\u5143\u201d\u662f\u4e00\u4e2a\u4ee5&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/3\/c\/b\/3cb2ab026a8bb3279a30485c2220a5a4.png\" alt=\"\\textstyle x_1, x_2, x_3\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u53ca\u622a\u8ddd&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/d\/c\/b\/dcb8dd3d14a2c0aa9b06ec6ce4ec0d59.png\" alt=\"\\textstyle +1\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u4e3a\u8f93\u5165\u503c\u7684\u8fd0\u7b97\u5355\u5143\uff0c\u5176\u8f93\u51fa\u4e3a&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/8\/9\/f\/89f1f9e549b908834d9fedca36d07bd4.png\" alt=\"\\textstyle h_{W,b}(x) = f(W^Tx) = f(\\sum_{i=1}^3 W_{i}x_i +b)\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u5176\u4e2d\u51fd\u6570&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/5\/d\/f\/5df2a707a6b2421afcb345f96051297e.png\" alt=\"\\textstyle f&nbsp;: \\Re \\mapsto \\Re\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u88ab\u79f0\u4e3a\u201c\u6fc0\u6d3b\u51fd\u6570\u201d\u3002\u5728\u672c\u6559\u7a0b\u4e2d\uff0c\u6211\u4eec\u9009\u7528sigmoid\u51fd\u6570\u4f5c\u4e3a<\/span><span style=\"font-size:16px;\">\u6fc0\u6d3b\u51fd\u6570<\/span><span style=\"font-size:16px;\">&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/0\/3\/0\/0303dd697c0e1b72185d7939f9870784.png\" alt=\"\\textstyle f(\\cdot)\" style=\"max-width:900px;height:auto;\" \/>\n<\/p>\n<dl>\n<dd>\n\t\t<img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/c\/e\/5\/ce5df10952ab30aa868f44db2f77486b.png\" alt=\" f(z) = \\frac{1}{1+\\exp(-z)}. \" style=\"max-width:900px;height:auto;\" \/>\n\t<\/dd>\n<\/dl>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">\u53ef\u4ee5\u770b\u51fa\uff0c\u8fd9\u4e2a\u5355\u4e00\u201c\u795e\u7ecf\u5143\u201d\u7684\u8f93\u5165\uff0d\u8f93\u51fa\u6620\u5c04\u5173\u7cfb\u5176\u5b9e\u5c31\u662f\u4e00\u4e2a\u903b\u8f91\u56de\u5f52\uff08logistic regression\uff09\u3002<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\"> \u867d\u7136\u672c\u7cfb\u5217\u6559\u7a0b\u91c7\u7528sigmoid\u51fd\u6570\uff0c\u4f46\u4f60\u4e5f\u53ef\u4ee5\u9009\u62e9\u53cc\u66f2\u6b63\u5207\u51fd\u6570\uff08tanh\uff09\uff1a<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t\n<\/p>\n<dl>\n<dd>\n\t\t<img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/a\/9\/0\/a9025d0884453bd5898c9681e871b3fb.png\" alt=\" f(z) = \\tanh(z) = \\frac{e^z - e^{-z}}{e^z + e^{-z}}, \" style=\"max-width:900px;height:auto;\" \/>\n\t<\/dd>\n<\/dl>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">\u4ee5\u4e0b\u5206\u522b\u662fsigmoid\u53catanh\u7684\u51fd\u6570\u56fe\u50cf<\/span>\n<\/p>\n<div align=\"center\">\n<p style=\"margin:10px auto;\">\n\t\t<a class=\"image\" title=\"Sigmoid activation function.\" href=\"http:\/\/ufldl.stanford.edu\/wiki\/index.php\/File:Sigmoid_Function.png\" data-rel=\"penci-gallery-image-content\" ><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/thumb\/c\/ca\/Sigmoid_Function.png\/400px-Sigmoid_Function.png\" alt=\"Sigmoid activation function.\" width=\"400\" height=\"300\" style=\"max-width:900px;height:auto;vertical-align:top;\" \/><\/a><span style=\"font-size:16px;\">&nbsp;<\/span><a class=\"image\" title=\"Tanh activation function.\" href=\"http:\/\/ufldl.stanford.edu\/wiki\/index.php\/File:Tanh_Function.png\" data-rel=\"penci-gallery-image-content\" ><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/thumb\/a\/aa\/Tanh_Function.png\/400px-Tanh_Function.png\" alt=\"Tanh activation function.\" width=\"400\" height=\"300\" style=\"max-width:900px;height:auto;vertical-align:top;\" \/><\/a>\n\t<\/p>\n<\/div>\n<p style=\"margin:10px auto;\">\n\t<img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/8\/7\/e\/87e9b5fc0869fae518eed4b75536334f.png\" alt=\"\\textstyle \\tanh(z)\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u51fd\u6570\u662fsigmoid\u51fd\u6570\u7684\u4e00\u79cd\u53d8\u4f53\uff0c\u5b83\u7684\u53d6\u503c\u8303\u56f4\u4e3a&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/8\/5\/a\/85a1c5a07f21a9eebbfb1dca380f8d38.png\" alt=\"\\textstyle [-1,1]\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u800c\u4e0d\u662fsigmoid\u51fd\u6570\u7684&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/8\/4\/2\/84235d31ac83fe764546463aba7acc0e.png\" alt=\"\\textstyle [0,1]\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u3002<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\"> \u6ce8\u610f\uff0c\u4e0e\u5176\u5b83\u5730\u65b9\uff08\u5305\u62ecOpenClassroom\u516c\u5f00\u8bfe\u4ee5\u53ca\u65af\u5766\u798f\u5927\u5b66CS229\u8bfe\u7a0b\uff09\u4e0d\u540c\u7684\u662f\uff0c\u8fd9\u91cc\u6211\u4eec\u4e0d\u518d\u4ee4&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/c\/5\/8\/c582053ce9cb63d69ae80acb53ded0d3.png\" alt=\"\\textstyle x_0=1\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u3002\u53d6\u800c\u4ee3\u4e4b\uff0c\u6211\u4eec\u7528\u5355\u72ec\u7684\u53c2\u6570&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/5\/2\/5\/5254b90d248051980262672a1bbc2433.png\" alt=\"\\textstyle b\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u6765\u8868\u793a\u622a\u8ddd\u3002<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\"> \u6700\u540e\u8981\u8bf4\u660e\u7684\u662f\uff0c\u6709\u4e00\u4e2a\u7b49\u5f0f\u6211\u4eec\u4ee5\u540e\u4f1a\u7ecf\u5e38\u7528\u5230\uff1a\u5982\u679c\u9009\u62e9&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/e\/c\/6\/ec62a4df6800f8c9ea680a08003df5c3.png\" alt=\"\\textstyle f(z) = 1\/(1+\\exp(-z))\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u4e5f\u5c31\u662fsigmoid\u51fd\u6570\uff0c\u90a3\u4e48\u5b83\u7684\u5bfc\u6570\u5c31\u662f&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/9\/9\/4\/994ac235e9478c8f465a4acdd8aae017.png\" alt=\"\\textstyle f'(z) = f(z) (1-f(z))\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff08\u5982\u679c\u9009\u62e9tanh\u51fd\u6570\uff0c\u90a3\u5b83\u7684\u5bfc\u6570\u5c31\u662f&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/e\/7\/d\/e7deb0493f3858b59b86181afe368fec.png\" alt=\"\\textstyle f'(z) = 1- (f(z))^2\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u4f60\u53ef\u4ee5\u6839\u636esigmoid\uff08\u6216tanh\uff09\u51fd\u6570\u7684\u5b9a\u4e49\u81ea\u884c\u63a8\u5bfc\u8fd9\u4e2a\u7b49\u5f0f\u3002<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t\n<\/p>\n<h2 style=\"line-height:1.5;margin:10px 0px;\"><span class=\"ez-toc-section\" id=\"%E7%A5%9E%E7%BB%8F%E7%BD%91%E7%BB%9C%E6%A8%A1%E5%9E%8B\"><\/span>\n\t<span style=\"font-size:16px;\">\u795e\u7ecf\u7f51\u7edc\u6a21\u578b<\/span><br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">\u6240\u8c13\u795e\u7ecf\u7f51\u7edc\u5c31\u662f\u5c06\u8bb8\u591a\u4e2a\u5355\u4e00\u201c\u795e\u7ecf\u5143\u201d\u8054\u7ed3\u5728\u4e00\u8d77\uff0c\u8fd9\u6837\uff0c\u4e00\u4e2a\u201c\u795e\u7ecf\u5143\u201d\u7684\u8f93\u51fa\u5c31\u53ef\u4ee5\u662f\u53e6\u4e00\u4e2a\u201c\u795e\u7ecf\u5143\u201d\u7684\u8f93\u5165\u3002\u4f8b\u5982\uff0c\u4e0b\u56fe\u5c31\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u795e\u7ecf\u7f51\u7edc\uff1a<\/span>\n<\/p>\n<div class=\"center\">\n<div class=\"floatnone\">\n\t\t<a class=\"image\" href=\"http:\/\/ufldl.stanford.edu\/wiki\/index.php\/File:Network331.png\" data-rel=\"penci-gallery-image-content\" ><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/thumb\/9\/99\/Network331.png\/400px-Network331.png\" alt=\"Network331.png\" width=\"400\" height=\"282\" style=\"max-width:900px;height:auto;\" \/><\/a>\n\t<\/div>\n<\/div>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">\u6211\u4eec\u4f7f\u7528\u5706\u5708\u6765\u8868\u793a\u795e\u7ecf\u7f51\u7edc\u7684\u8f93\u5165\uff0c\u6807\u4e0a\u201c<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/d\/c\/b\/dcb8dd3d14a2c0aa9b06ec6ce4ec0d59.png\" alt=\"\\textstyle +1\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">\u201d\u7684\u5706\u5708\u88ab\u79f0\u4e3a<\/span><span style=\"font-size:16px;\">\u504f\u7f6e\u8282\u70b9<\/span><span style=\"font-size:16px;\">\uff0c\u4e5f\u5c31\u662f\u622a\u8ddd\u9879\u3002<\/span><span style=\"font-size:16px;\">\u795e\u7ecf\u7f51\u7edc\u6700\u5de6\u8fb9\u7684\u4e00\u5c42\u53eb\u505a<\/span><span style=\"font-size:16px;\">\u8f93\u5165\u5c42<\/span><span style=\"font-size:16px;\">\uff0c\u6700\u53f3\u7684\u4e00\u5c42\u53eb\u505a<\/span><span style=\"font-size:16px;\">\u8f93\u51fa\u5c42<\/span><span style=\"font-size:16px;\">\uff08\u672c\u4f8b\u4e2d\uff0c\u8f93\u51fa\u5c42\u53ea\u6709\u4e00\u4e2a\u8282\u70b9\uff09\u3002\u4e2d\u95f4\u6240\u6709\u8282\u70b9\u7ec4\u6210\u7684\u4e00\u5c42\u53eb\u505a<\/span><span style=\"font-size:16px;\">\u9690\u85cf\u5c42<\/span><span style=\"font-size:16px;\">\uff0c\u56e0\u4e3a\u6211\u4eec\u4e0d\u80fd\u5728\u8bad\u7ec3\u6837\u672c\u96c6\u4e2d\u89c2\u6d4b\u5230\u5b83\u4eec\u7684\u503c\u3002\u540c\u65f6\u53ef\u4ee5\u770b\u5230\uff0c\u4ee5\u4e0a\u795e\u7ecf\u7f51\u7edc\u7684\u4f8b\u5b50\u4e2d\u67093\u4e2a<\/span><span style=\"font-size:16px;\">\u8f93\u5165\u5355\u5143<\/span><span style=\"font-size:16px;\">\uff08\u504f\u7f6e\u5355\u5143\u4e0d\u8ba1\u5728\u5185\uff09\uff0c3\u4e2a<\/span><span style=\"font-size:16px;\">\u9690\u85cf\u5355\u5143<\/span><span style=\"font-size:16px;\">\u53ca\u4e00\u4e2a<\/span><span style=\"font-size:16px;\">\u8f93\u51fa\u5355\u5143<\/span><span style=\"font-size:16px;\">\u3002<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\"> \u6211\u4eec\u7528&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/5\/4\/6\/546158a6d0082614d47e7f8a63225b0b.png\" alt=\"\\textstyle {n}_l\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u6765\u8868\u793a\u7f51\u7edc\u7684\u5c42\u6570\uff0c\u672c\u4f8b\u4e2d&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/3\/c\/8\/3c89b5db1e49221343428af57c90e44a.png\" alt=\"\\textstyle n_l=3\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u6211\u4eec\u5c06\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/b\/a\/0\/ba0593b3db2fa8535b077516f4b0d70b.png\" alt=\"\\textstyle l\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u8bb0\u4e3a&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/5\/5\/e\/55ea36127aa64b92b071c269cd1e3990.png\" alt=\"\\textstyle L_l\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u4e8e\u662f&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/1\/3\/e\/13e0887b9e716279d9a7b8bc8e6ad63b.png\" alt=\"\\textstyle L_1\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u662f\u8f93\u5165\u5c42\uff0c\u8f93\u51fa\u5c42\u662f&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/2\/2\/1\/221a7296664022427d488fdb9b14b19b.png\" alt=\"\\textstyle L_{n_l}\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u3002\u672c\u4f8b\u795e\u7ecf\u7f51\u7edc\u6709\u53c2\u6570&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/a\/a\/3\/aa3d6ed3c577d41a791324008558efbe.png\" alt=\"\\textstyle (W,b) = (W^{(1)}, b^{(1)}, W^{(2)}, b^{(2)})\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u5176\u4e2d&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/d\/f\/e\/dfe43c64e3c42ea4ff1774fc82b87805.png\" alt=\"\\textstyle W^{(l)}_{ij}\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff08\u4e0b\u9762\u7684\u5f0f\u5b50\u4e2d\u7528\u5230\uff09\u662f\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/b\/a\/0\/ba0593b3db2fa8535b077516f4b0d70b.png\" alt=\"\\textstyle l\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/2\/3\/5\/235c5146ab110558897640c34dad7d97.png\" alt=\"\\textstyle j\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5355\u5143\u4e0e\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/9\/0\/6\/9068105ec8ebb97277c937bfa61b606d.png\" alt=\"\\textstyle l+1\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/0\/b\/3\/0b36ee693126b34b58f77dba7ed23987.png\" alt=\"\\textstyle i\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5355\u5143\u4e4b\u95f4\u7684\u8054\u63a5\u53c2\u6570\uff08\u5176\u5b9e\u5c31\u662f\u8fde\u63a5\u7ebf\u4e0a\u7684\u6743\u91cd\uff0c\u6ce8\u610f\u6807\u53f7\u987a\u5e8f\uff09\uff0c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/4\/c\/7\/4c786c16575b63bbb554254725b6b648.png\" alt=\"\\textstyle b^{(l)}_i\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u662f\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/9\/0\/6\/9068105ec8ebb97277c937bfa61b606d.png\" alt=\"\\textstyle l+1\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/0\/b\/3\/0b36ee693126b34b58f77dba7ed23987.png\" alt=\"\\textstyle i\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5355\u5143\u7684\u504f\u7f6e\u9879\u3002\u56e0\u6b64\u5728\u672c\u4f8b\u4e2d\uff0c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/5\/c\/a\/5ca0efbb17e86cb00091f6a528e0ab0e.png\" alt=\"\\textstyle W^{(1)} \\in \\Re^{3\\times 3}\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/4\/3\/1\/431cf6f298e4106efb5bff4495aa3c6d.png\" alt=\"\\textstyle W^{(2)} \\in \\Re^{1\\times 3}\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u3002\u6ce8\u610f\uff0c\u6ca1\u6709\u5176\u4ed6\u5355\u5143\u8fde\u5411\u504f\u7f6e\u5355\u5143(\u5373\u504f\u7f6e\u5355\u5143\u6ca1\u6709\u8f93\u5165)\uff0c\u56e0\u4e3a\u5b83\u4eec\u603b\u662f\u8f93\u51fa&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/d\/c\/b\/dcb8dd3d14a2c0aa9b06ec6ce4ec0d59.png\" alt=\"\\textstyle +1\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">\u3002\u540c\u65f6\uff0c\u6211\u4eec\u7528&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/8\/a\/f\/8afb62ac69ccb2911bb24795ff052a07.png\" alt=\"\\textstyle s_l\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u8868\u793a\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/b\/a\/0\/ba0593b3db2fa8535b077516f4b0d70b.png\" alt=\"\\textstyle l\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u7684\u8282\u70b9\u6570\uff08\u504f\u7f6e\u5355\u5143\u4e0d\u8ba1\u5728\u5185\uff09\u3002<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\"> \u6211\u4eec\u7528&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/c\/9\/b\/c9b144e0a6735fafb01b3615a2a0dc05.png\" alt=\"\\textstyle a^{(l)}_i\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u8868\u793a\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/b\/a\/0\/ba0593b3db2fa8535b077516f4b0d70b.png\" alt=\"\\textstyle l\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/0\/b\/3\/0b36ee693126b34b58f77dba7ed23987.png\" alt=\"\\textstyle i\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5355\u5143\u7684<\/span><span style=\"font-size:16px;\">\u6fc0\u6d3b\u503c<\/span><span style=\"font-size:16px;\">\uff08\u8f93\u51fa\u503c\uff09\u3002\u5f53&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/4\/a\/4\/4a4725e295806f22b26342fe3cd3338f.png\" alt=\"\\textstyle l=1\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u65f6\uff0c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/f\/5\/c\/f5c1979e94318aee674de68348b96557.png\" alt=\"\\textstyle a^{(1)}_i = x_i\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u4e5f\u5c31\u662f\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/0\/b\/3\/0b36ee693126b34b58f77dba7ed23987.png\" alt=\"\\textstyle i\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u4e2a\u8f93\u5165\u503c\uff08\u8f93\u5165\u503c\u7684\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/0\/b\/3\/0b36ee693126b34b58f77dba7ed23987.png\" alt=\"\\textstyle i\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u4e2a\u7279\u5f81\uff09\u3002\u5bf9\u4e8e\u7ed9\u5b9a\u53c2\u6570\u96c6\u5408&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/7\/c\/9\/7c9aa03f5258ecf79556ba374d7eb2cd.png\" alt=\"\\textstyle W,b\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u6211\u4eec\u7684\u795e\u7ecf\u7f51\u7edc\u5c31\u53ef\u4ee5\u6309\u7167\u51fd\u6570&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/5\/8\/d\/58d3a4fe4ad68b333b180071dd46db82.png\" alt=\"\\textstyle h_{W,b}(x)\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u6765\u8ba1\u7b97\u8f93\u51fa\u7ed3\u679c\u3002\u672c\u4f8b\u795e\u7ecf\u7f51\u7edc\u7684\u8ba1\u7b97\u6b65\u9aa4\u5982\u4e0b\uff1a<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t\n<\/p>\n<dl>\n<dd>\n\t\t<img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/f\/d\/e\/fde22a388f607f526f03644c71a72f92.png\" alt=\" \\begin{align} a_1^{(2)} &amp;= f(W_{11}^{(1)}x_1 + W_{12}^{(1)} x_2 + W_{13}^{(1)} x_3 + b_1^{(1)}) \\\\ a_2^{(2)} &amp;= f(W_{21}^{(1)}x_1 + W_{22}^{(1)} x_2 + W_{23}^{(1)} x_3 + b_2^{(1)}) \\\\ a_3^{(2)} &amp;= f(W_{31}^{(1)}x_1 + W_{32}^{(1)} x_2 + W_{33}^{(1)} x_3 + b_3^{(1)}) \\\\ h_{W,b}(x) &amp;= a_1^{(3)} = f(W_{11}^{(2)}a_1^{(2)} + W_{12}^{(2)} a_2^{(2)} + W_{13}^{(2)} a_3^{(2)} + b_1^{(2)}) \\end{align} \" style=\"max-width:900px;height:auto;\" \/>\n\t<\/dd>\n<\/dl>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\"> \u6211\u4eec\u7528&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/3\/d\/d\/3dd5c56e0949e76de86690e1b868cdcf.png\" alt=\"\\textstyle z^{(l)}_i\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u8868\u793a\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/b\/a\/0\/ba0593b3db2fa8535b077516f4b0d70b.png\" alt=\"\\textstyle l\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/0\/b\/3\/0b36ee693126b34b58f77dba7ed23987.png\" alt=\"\\textstyle i\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5355\u5143\u8f93\u5165\u52a0\u6743\u548c\uff08\u5305\u62ec\u504f\u7f6e\u5355\u5143\uff09\uff0c\u6bd4\u5982\uff0c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/a\/a\/e\/aae7340fe1eb75c824b8abc107c3db27.png\" alt=\"\\textstyle z_i^{(2)} = \\sum_{j=1}^n W^{(1)}_{ij} x_j + b^{(1)}_i\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u5219&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/c\/f\/8\/cf8cb56750f5aaca7dc59480a53d9676.png\" alt=\"\\textstyle a^{(l)}_i = f(z^{(l)}_i)\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u3002<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\"> \u8fd9\u6837\u6211\u4eec\u5c31\u53ef\u4ee5\u5f97\u5230\u4e00\u79cd\u66f4\u7b80\u6d01\u7684\u8868\u793a\u6cd5\u3002\u8fd9\u91cc\u6211\u4eec\u5c06\u6fc0\u6d3b\u51fd\u6570&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/0\/3\/0\/0303dd697c0e1b72185d7939f9870784.png\" alt=\"\\textstyle f(\\cdot)\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u6269\u5c55\u4e3a\u7528\u5411\u91cf\uff08\u5206\u91cf\u7684\u5f62\u5f0f\uff09\u6765\u8868\u793a\uff0c\u5373&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/d\/b\/8\/db84346dcd6187f0fbb0f6c1a72eecf8.png\" alt=\"\\textstyle f([z_1, z_2, z_3]) = [f(z_1), f(z_2), f(z_3)]\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u90a3\u4e48\uff0c\u4e0a\u9762\u7684\u7b49\u5f0f\u53ef\u4ee5\u66f4\u7b80\u6d01\u5730\u8868\u793a\u4e3a\uff1a<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t\n<\/p>\n<dl>\n<dd>\n\t\t<img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/9\/6\/9\/9690acc03c1e5133b0509257b532b4f7.png\" alt=\"\\begin{align} z^{(2)} &amp;= W^{(1)} x + b^{(1)} \\\\ a^{(2)} &amp;= f(z^{(2)}) \\\\ z^{(3)} &amp;= W^{(2)} a^{(2)} + b^{(2)} \\\\ h_{W,b}(x) &amp;= a^{(3)} = f(z^{(3)}) \\end{align}\" style=\"max-width:900px;height:auto;\" \/>\n\t<\/dd>\n<\/dl>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\"> \u6211\u4eec\u5c06\u4e0a\u9762\u7684\u8ba1\u7b97\u6b65\u9aa4\u53eb\u4f5c<\/span><span style=\"font-size:16px;\">\u524d\u5411\u4f20\u64ad<\/span><span style=\"font-size:16px;\">\u3002\u56de\u60f3\u4e00\u4e0b\uff0c\u4e4b\u524d\u6211\u4eec\u7528&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/d\/e\/0\/de0b51a7e4a2b2047d52a165419ac048.png\" alt=\"\\textstyle a^{(1)} = x\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u8868\u793a\u8f93\u5165\u5c42\u7684\u6fc0\u6d3b\u503c\uff0c\u90a3\u4e48\u7ed9\u5b9a\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/b\/a\/0\/ba0593b3db2fa8535b077516f4b0d70b.png\" alt=\"\\textstyle l\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u7684\u6fc0\u6d3b\u503c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/b\/d\/2\/bd2728b5337ccec5b5729756d5796b20.png\" alt=\"\\textstyle a^{(l)}\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u540e\uff0c\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/9\/0\/6\/9068105ec8ebb97277c937bfa61b606d.png\" alt=\"\\textstyle l+1\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u7684\u6fc0\u6d3b\u503c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/e\/b\/8\/eb8a863a7b57397bf06a0532d4f1daf1.png\" alt=\"\\textstyle a^{(l+1)}\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c31\u53ef\u4ee5\u6309\u7167\u4e0b\u9762\u6b65\u9aa4\u8ba1\u7b97\u5f97\u5230\uff1a<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t\n<\/p>\n<dl>\n<dd>\n\t\t<img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/5\/c\/f\/5cfcbbe6d55b6c882f56a85a57eafe6e.png\" alt=\" \\begin{align} z^{(l+1)} &amp;= W^{(l)} a^{(l)} + b^{(l)} \\\\ a^{(l+1)} &amp;= f(z^{(l+1)}) \\end{align}\" style=\"max-width:900px;height:auto;\" \/>\n\t<\/dd>\n<\/dl>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\"> \u5c06\u53c2\u6570\u77e9\u9635\u5316\uff0c\u4f7f\u7528\u77e9\u9635\uff0d\u5411\u91cf\u8fd0\u7b97\u65b9\u5f0f\uff0c\u6211\u4eec\u5c31\u53ef\u4ee5\u5229\u7528\u7ebf\u6027\u4ee3\u6570\u7684\u4f18\u52bf\u5bf9\u795e\u7ecf\u7f51\u7edc\u8fdb\u884c\u5feb\u901f\u6c42\u89e3\u3002<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\"> \u76ee\u524d\u4e3a\u6b62\uff0c\u6211\u4eec\u8ba8\u8bba\u4e86\u4e00\u79cd\u795e\u7ecf\u7f51\u7edc\uff0c\u6211\u4eec\u4e5f\u53ef\u4ee5\u6784\u5efa\u53e6\u4e00\u79cd<\/span><span style=\"font-size:16px;\">\u7ed3\u6784<\/span><span style=\"font-size:16px;\">\u7684\u795e\u7ecf\u7f51\u7edc\uff08\u8fd9\u91cc\u7ed3\u6784\u6307\u7684\u662f\u795e\u7ecf\u5143\u4e4b\u95f4\u7684\u8054\u63a5\u6a21\u5f0f\uff09\uff0c\u4e5f\u5c31\u662f\u5305\u542b\u591a\u4e2a\u9690\u85cf\u5c42\u7684\u795e\u7ecf\u7f51\u7edc\u3002\u6700\u5e38\u89c1\u7684\u4e00\u4e2a\u4f8b\u5b50\u662f&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/5\/b\/7\/5b7a0657fdea25f29866c8e1d6e884ac.png\" alt=\"\\textstyle n_l\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u7684\u795e\u7ecf\u7f51\u7edc\uff0c\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/6\/e\/9\/6e924e04b5c9d4c5be131609a038b821.png\" alt=\"\\textstyle 1\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u662f\u8f93\u5165\u5c42\uff0c\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/5\/b\/7\/5b7a0657fdea25f29866c8e1d6e884ac.png\" alt=\"\\textstyle n_l\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u662f\u8f93\u51fa\u5c42\uff0c\u4e2d\u95f4\u7684\u6bcf\u4e2a\u5c42&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/b\/a\/0\/ba0593b3db2fa8535b077516f4b0d70b.png\" alt=\"\\textstyle l\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">\u4e0e\u5c42&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/9\/0\/6\/9068105ec8ebb97277c937bfa61b606d.png\" alt=\"\\textstyle l+1\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u7d27\u5bc6\u76f8\u8054\u3002\u8fd9\u79cd\u6a21\u5f0f\u4e0b\uff0c\u8981\u8ba1\u7b97\u795e\u7ecf\u7f51\u7edc\u7684\u8f93\u51fa\u7ed3\u679c\uff0c\u6211\u4eec\u53ef\u4ee5\u6309\u7167\u4e4b\u524d\u63cf\u8ff0\u7684\u7b49\u5f0f\uff0c\u6309\u90e8\u5c31\u73ed\uff0c\u8fdb\u884c\u524d\u5411\u4f20\u64ad\uff0c\u9010\u4e00\u8ba1\u7b97\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/c\/f\/7\/cf7d186efd913f4fb9ceb939bf5135c4.png\" alt=\"\\textstyle L_2\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u7684\u6240\u6709\u6fc0\u6d3b\u503c\uff0c\u7136\u540e\u662f\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/d\/9\/b\/d9b949d768ca8bab18830d9efc3fa441.png\" alt=\"\\textstyle L_3\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u7684\u6fc0\u6d3b\u503c\uff0c\u4ee5\u6b64\u7c7b\u63a8\uff0c\u76f4\u5230\u7b2c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/2\/2\/1\/221a7296664022427d488fdb9b14b19b.png\" alt=\"\\textstyle L_{n_l}\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u5c42\u3002\u8fd9\u662f\u4e00\u4e2a<\/span><span style=\"font-size:16px;\">\u524d\u9988<\/span><span style=\"font-size:16px;\">\u795e\u7ecf\u7f51\u7edc\u7684\u4f8b\u5b50\uff0c\u56e0\u4e3a\u8fd9\u79cd\u8054\u63a5\u56fe\u6ca1\u6709\u95ed\u73af\u6216\u56de\u8def\u3002<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\"> \u795e\u7ecf\u7f51\u7edc\u4e5f\u53ef\u4ee5\u6709\u591a\u4e2a\u8f93\u51fa\u5355\u5143\u3002\u6bd4\u5982\uff0c\u4e0b\u9762\u7684\u795e\u7ecf\u7f51\u7edc\u6709\u4e24\u5c42\u9690\u85cf\u5c42\uff1a&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/c\/f\/7\/cf7d186efd913f4fb9ceb939bf5135c4.png\" alt=\"\\textstyle L_2\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u53ca&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/d\/9\/b\/d9b949d768ca8bab18830d9efc3fa441.png\" alt=\"\\textstyle L_3\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u8f93\u51fa\u5c42&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/a\/b\/0\/ab05e0667abe37f2e3cbc05735573034.png\" alt=\"\\textstyle L_4\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u6709\u4e24\u4e2a\u8f93\u51fa\u5355\u5143\u3002<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t\n<\/p>\n<div class=\"center\">\n<div class=\"floatnone\">\n\t\t<a class=\"image\" href=\"http:\/\/ufldl.stanford.edu\/wiki\/index.php\/File:Network3322.png\" data-rel=\"penci-gallery-image-content\" ><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/thumb\/4\/40\/Network3322.png\/500px-Network3322.png\" alt=\"Network3322.png\" width=\"500\" height=\"274\" style=\"max-width:900px;height:auto;\" \/><\/a>\n\t<\/div>\n<\/div>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\"> \u8981\u6c42\u89e3\u8fd9\u6837\u7684\u795e\u7ecf\u7f51\u7edc\uff0c\u9700\u8981\u6837\u672c\u96c6&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/f\/1\/7\/f178249571382c3921d2c46f7abd47da.png\" alt=\"\\textstyle (x^{(i)}, y^{(i)})\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\uff0c\u5176\u4e2d&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/9\/e\/d\/9edce3bff2898e4b7f084ad3a2bbf494.png\" alt=\"\\textstyle y^{(i)} \\in \\Re^2\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u3002\u5982\u679c\u4f60\u60f3\u9884\u6d4b\u7684\u8f93\u51fa\u662f\u591a\u4e2a\u7684\uff0c\u90a3\u8fd9\u79cd\u795e\u7ecf\u7f51\u7edc\u5f88\u9002\u7528\u3002\uff08\u6bd4\u5982\uff0c\u5728\u533b\u7597\u8bca\u65ad\u5e94\u7528\u4e2d\uff0c\u60a3\u8005\u7684\u4f53\u5f81\u6307\u6807\u5c31\u53ef\u4ee5\u4f5c\u4e3a\u5411\u91cf\u7684\u8f93\u5165\u503c\uff0c\u800c\u4e0d\u540c\u7684\u8f93\u51fa\u503c&nbsp;<\/span><img decoding=\"async\" class=\"tex\" src=\"http:\/\/ufldl.stanford.edu\/wiki\/images\/math\/7\/a\/5\/7a5d164f3df0329a8032cda67d95d9d4.png\" alt=\"\\textstyle y_i\" style=\"max-width:900px;height:auto;\" \/><span style=\"font-size:16px;\">&nbsp;\u53ef\u4ee5\u8868\u793a\u4e0d\u540c\u7684\u75be\u75c5\u5b58\u5728\u4e0e\u5426\u3002\uff09<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">\u539f\u6587\uff1a<\/span><span style=\"font-size:16px;\">https:\/\/www.cnblogs.com\/bonelee\/p\/9092080.html<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">MLP\u662f\u4e00\u4e2a\u76d1\u7763\u5b66\u4e60\u7b97\u6cd5\uff0c\u56fe1\u662f\u5e26\u4e00\u4e2a\u9690\u85cf\u5c42\u7684MLP\u6a21\u578b&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">\u5de6\u8fb9\u5c42\u662f\u8f93\u5165\u5c42\uff0c\u7531\u795e\u7ecf\u5143\u96c6\u5408{xi|x1,x2,\u2026,xm},\u4ee3\u8868\u8f93\u5165\u7279\u5f81\uff0c\u9690\u85cf\u5c42\u7684\u6bcf\u4e2a\u795e\u7ecf\u5143\u5c06\u524d\u4e00\u5c42\u7684\u7684\u503c\u901a\u8fc7\u7ebf\u6027\u52a0\u6743\u6c42\u548c\u7684\u65b9\u5f0f\u8868\u793a\uff0c\u5373w1x1+w2x2+\u2026+wmxm,\u5176\u6b21\u662f\u4e00\u4e2a\u975e\u7ebf\u6027\u6fc0\u6d3b\u51fd\u6570g\uff08.):R-&gt;R,\u6bd4\u5982\u53cc\u66f2\u51fd\u6570\uff0c\u8f93\u51fa\u5c42\u63a5\u53d7\u4ece\u6700\u540e\u4e00\u4e2a\u9690\u85cf\u5c42\u8f93\u51fa\u7684\u503c\u5e76\u5c06\u4ed6\u4eec\u8f6c\u6362\u6210\u503c\u3002&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">\u8fd9\u4e2a\u6a21\u5757\u5305\u542b\u516c\u5171\u5c5e\u6027coefs_ \u548cintercepts_ \u3002coefs_ \u662f\u6743\u91cd\u77e9\u9635\u5217\u8868\uff0c\u4e0b\u6807\u4e3ai\u7684\u6743\u91cd\u77e9\u9635\u4ee3\u8868i\u548ci+1\u5c42\u7684\u6743\u91cd\u3002intercepts_ \u662f\u4e00\u4e2a\u504f\u5dee\u5411\u91cf\u5217\u8868\uff0c\u5176\u4e2d\u7b2ci\u4e2a\u504f\u5dee\u5411\u91cf\u4ee3\u8868\u52a0\u5230i+1\u5c42\u4e0a\u7684\u504f\u5dee\u503c\u3002&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">MLP\u7684\u4f18\u70b9\u662f\uff1a&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; \u53ef\u4ee5\u5b66\u4e60\u975e\u7ebf\u6027\u6a21\u578b&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; \u4f7f\u7528partial_fit \u5b9e\u65f6\u5b66\u4e60&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">MLP\u7684\u7f3a\u70b9\u662f\uff1a&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; \u6709\u9690\u85cf\u5c42\u7684MLP\u5305\u542b\u4e00\u4e2a\u975e\u51f8\u6027\u635f\u5931\u51fd\u6570\uff0c\u5b58\u5728\u8d85\u8fc7\u4e00\u4e2a\u6700\u5c0f\u503c\uff0c\u6240\u4ee5\u4e0d\u540c\u7684\u968f\u673a\u521d\u59cb\u6743\u91cd\u53ef\u80fd\u5bfc\u81f4\u4e0d\u540c\u9a8c\u8bc1\u7cbe\u786e\u5ea6&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; MLP\u8981\u6c42\u8c03\u6574\u4e00\u7cfb\u5217\u8d85\u53c2\u6570\uff0c\u6bd4\u5982\u9690\u85cf\u795e\u7ecf\u5143\uff0c\u9690\u85cf\u5c42\u7684\u4e2a\u6570\u4ee5\u53ca\u8fed\u4ee3\u7684\u6b21\u6570&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; MLP\u5bf9\u7279\u5f81\u7f29\u653e\u6bd4\u8f83\u654f\u611f<\/span><\/p>\n<p><span style=\"font-size:16px;\">2.\u5206\u7c7b<\/span><br \/>\n<span style=\"font-size:16px;\">MLPClassifier\u4f7f\u7528BP\u7b97\u6cd5\u6765\u5b9e\u73b0\u4e00\u4e2a\u591a\u5c42\u611f\u77e5\u673a\u3002&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">\u793a\u4f8b1\uff1a<\/span><\/p>\n<p><span style=\"font-size:16px;\">&gt;&gt;&gt; from sklearn.neural_network import MLPClassifier<\/span><br \/>\n<span style=\"font-size:16px;\">&gt;&gt;&gt; X=[[0.,0.],[1.,1.]]<\/span><br \/>\n<span style=\"font-size:16px;\">&gt;&gt;&gt; y=[0,1]<\/span><br \/>\n<span style=\"font-size:16px;\">&gt;&gt;&gt; clf = MLPClassifier(solver=&#8217;lbfgs&#8217;,alpha=1e-5,hidden_layer_sizes=(5,2),random_state=1)<\/span><br \/>\n<span style=\"font-size:16px;\">&gt;&gt;&gt; clf.fit(X,y)<\/span><br \/>\n<span style=\"font-size:16px;\">MLPClassifier(activation=&#8217;relu&#8217;, alpha=1e-05, batch_size=&#8217;auto&#8217;, beta_1=0.9,<\/span><br \/>\n<span style=\"font-size:16px;\">&nbsp; &nbsp; &nbsp; &nbsp;beta_2=0.999, early_stopping=False, epsilon=1e-08,<\/span><br \/>\n<span style=\"font-size:16px;\">&nbsp; &nbsp; &nbsp; &nbsp;hidden_layer_sizes=(5, 2), learning_rate=&#8217;constant&#8217;,<\/span><br \/>\n<span style=\"font-size:16px;\">&nbsp; &nbsp; &nbsp; &nbsp;learning_rate_init=0.001, max_iter=200, momentum=0.9,<\/span><br \/>\n<span style=\"font-size:16px;\">&nbsp; &nbsp; &nbsp; &nbsp;nesterovs_momentum=True, power_t=0.5, random_state=1, shuffle=True,<\/span><br \/>\n<span style=\"font-size:16px;\">&nbsp; &nbsp; &nbsp; &nbsp;solver=&#8217;lbfgs&#8217;, tol=0.0001, validation_fraction=0.1, verbose=False,<\/span><br \/>\n<span style=\"font-size:16px;\">&nbsp; &nbsp; &nbsp; &nbsp;warm_start=False)<\/span><br \/>\n<span style=\"font-size:16px;\">&gt;&gt;&gt; clf.predict([[2.,2.],[-1.,-2.]])<\/span><br \/>\n<span style=\"font-size:16px;\">array([1, 0])<\/span><br \/>\n<span style=\"font-size:16px;\">&gt;&gt;&gt; [coef.shape for coef in clf.coefs_]<\/span><br \/>\n<span style=\"font-size:16px;\">[(2L, 5L), (5L, 2L), (2L, 1L)]<\/span><br \/>\n<span style=\"font-size:16px;\">&gt;&gt;&gt; clf.predict_proba([[2.,2.],[1.,2.]])<\/span><br \/>\n<span style=\"font-size:16px;\">array([[&nbsp; 1.96718015e-04,&nbsp; &nbsp;9.99803282e-01],<\/span><br \/>\n<span style=\"font-size:16px;\">&nbsp; &nbsp; &nbsp; &nbsp;[&nbsp; 1.96718015e-04,&nbsp; &nbsp;9.99803282e-01]])<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<br \/>\n<span style=\"font-size:16px;\">\u53c2\u6570\u8bf4\u660e\uff08\u7ea2\u8272\u4e3a\u4f18\u5148\u4f18\u5316\u53c2\u6570\uff09\uff1a&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">1. <span style=\"color:#E53333;\">hidden_laye<\/span><span style=\"color:#E53333;\">r_sizes <\/span>:\u5143\u7956\u683c\u5f0f\uff0c\u957f\u5ea6=n_layers-2, \u9ed8\u8ba4(100\uff0c\uff09\uff0c\u7b2ci\u4e2a\u5143\u7d20\u8868\u793a\u7b2ci\u4e2a\u9690\u85cf\u5c42\u7684\u795e\u7ecf\u5143\u7684\u4e2a\u6570\u3002&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">2. <span style=\"color:#E53333;\">activation <\/span>:{\u2018identity\u2019, \u2018logistic\u2019, \u2018tanh\u2019, \u2018relu\u2019}, \u9ed8\u8ba4\u2018relu&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; \u2018identity\u2019\uff1a no-op activation, useful to implement linear bottleneck\uff0c&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">\u8fd4\u56def(x) = x&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; \u2018logistic\u2019\uff1athe logistic sigmoid function, returns f(x) = 1 \/ (1 + exp(-x)).&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; \u2018tanh\u2019\uff1athe hyperbolic tan function, returns f(x) = tanh(x).&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; \u2018relu\u2019\uff1athe rectified linear unit function, returns f(x) = max(0, x)&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">4. <span style=\"color:#E53333;\">solver<\/span>\uff1a {\u2018lbfgs\u2019, \u2018sgd\u2019, \u2018adam\u2019}, \u9ed8\u8ba4 \u2018adam\u2019\uff0c\u7528\u6765\u4f18\u5316\u6743\u91cd&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; lbfgs\uff1aquasi-Newton\u65b9\u6cd5\u7684\u4f18\u5316\u5668&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; sgd\uff1a\u968f\u673a\u68af\u5ea6\u4e0b\u964d&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; adam\uff1a Kingma, Diederik, and Jimmy Ba\u63d0\u51fa\u7684\u673a\u9047\u968f\u673a\u68af\u5ea6\u7684\u4f18\u5316\u5668&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">\u6ce8\u610f\uff1a\u9ed8\u8ba4solver \u2018adam\u2019\u5728\u76f8\u5bf9\u8f83\u5927\u7684\u6570\u636e\u96c6\u4e0a\u6548\u679c\u6bd4\u8f83\u597d\uff08\u51e0\u5343\u4e2a\u6837\u672c\u6216\u8005\u66f4\u591a\uff09\uff0c\u5bf9\u5c0f\u6570\u636e\u96c6\u6765\u8bf4\uff0clbfgs\u6536\u655b\u66f4\u5feb\u6548\u679c\u4e5f\u66f4\u597d\u3002&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">5. alpha :float,\u53ef\u9009\u7684\uff0c\u9ed8\u8ba40.0001,\u6b63\u5219\u5316\u9879\u53c2\u6570&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">6. batch_size : int , \u53ef\u9009\u7684\uff0c\u9ed8\u8ba4\u2018auto\u2019,\u968f\u673a\u4f18\u5316\u7684minibatches\u7684\u5927\u5c0f\uff0c\u5982\u679csolver\u662f\u2018lbfgs\u2019\uff0c\u5206\u7c7b\u5668\u5c06\u4e0d\u4f7f\u7528minibatch\uff0c\u5f53\u8bbe\u7f6e\u6210\u2018auto\u2019\uff0cbatch_size=min(200,n_samples)&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">7. learning_rate :{\u2018constant\u2019\uff0c\u2018invscaling\u2019, \u2018adaptive\u2019},\u9ed8\u8ba4\u2018constant\u2019\uff0c\u7528\u4e8e\u6743\u91cd\u66f4\u65b0\uff0c\u53ea\u6709\u5f53solver\u4e3a\u2019sgd\u2018\u65f6\u4f7f\u7528&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; \u2018constant\u2019: \u6709\u2018learning_rate_init\u2019\u7ed9\u5b9a\u7684\u6052\u5b9a\u5b66\u4e60\u7387&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; \u2018incscaling\u2019\uff1a\u968f\u7740\u65f6\u95f4t\u4f7f\u7528\u2019power_t\u2019\u7684\u9006\u6807\u5ea6\u6307\u6570\u4e0d\u65ad\u964d\u4f4e\u5b66\u4e60\u7387learning_rate_ \uff0ceffective_learning_rate = learning_rate_init \/ pow(t, power_t)&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; \u2018adaptive\u2019\uff1a\u53ea\u8981\u8bad\u7ec3\u635f\u8017\u5728\u4e0b\u964d\uff0c\u5c31\u4fdd\u6301\u5b66\u4e60\u7387\u4e3a\u2019learning_rate_init\u2019\u4e0d\u53d8\uff0c\u5f53\u8fde\u7eed\u4e24\u6b21\u4e0d\u80fd\u964d\u4f4e\u8bad\u7ec3\u635f\u8017\u6216\u9a8c\u8bc1\u5206\u6570\u505c\u6b62\u5347\u9ad8\u81f3\u5c11tol\u65f6\uff0c\u5c06\u5f53\u524d\u5b66\u4e60\u7387\u9664\u4ee55.&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">8. <span style=\"color:#E53333;\">max_iter<\/span>: int\uff0c\u53ef\u9009\uff0c\u9ed8\u8ba4200\uff0c\u6700\u5927\u8fed\u4ee3\u6b21\u6570\u3002&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">9. random_state:int \u6216RandomState\uff0c\u53ef\u9009\uff0c\u9ed8\u8ba4None\uff0c\u968f\u673a\u6570\u751f\u6210\u5668\u7684\u72b6\u6001\u6216\u79cd\u5b50\u3002&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">10. shuffle: bool\uff0c\u53ef\u9009\uff0c\u9ed8\u8ba4True,\u53ea\u6709\u5f53solver=\u2019sgd\u2019\u6216\u8005\u2018adam\u2019\u65f6\u4f7f\u7528\uff0c\u5224\u65ad\u662f\u5426\u5728\u6bcf\u6b21\u8fed\u4ee3\u65f6\u5bf9\u6837\u672c\u8fdb\u884c\u6e05\u6d17\u3002&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">11. tol\uff1afloat, \u53ef\u9009\uff0c\u9ed8\u8ba41e-4\uff0c\u4f18\u5316\u7684\u5bb9\u5fcd\u5ea6&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">12. learning_rate_int:double,\u53ef\u9009\uff0c\u9ed8\u8ba40.001\uff0c\u521d\u59cb\u5b66\u4e60\u7387\uff0c\u63a7\u5236\u66f4\u65b0\u6743\u91cd\u7684\u8865\u507f\uff0c\u53ea\u6709\u5f53solver=\u2019sgd\u2019 \u6216\u2019adam\u2019\u65f6\u4f7f\u7528\u3002&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">13. power_t: double, optional, default 0.5\uff0c\u53ea\u6709solver=\u2019sgd\u2019\u65f6\u4f7f\u7528\uff0c\u662f\u9006\u6269\u5c55\u5b66\u4e60\u7387\u7684\u6307\u6570.\u5f53learning_rate=\u2019invscaling\u2019\uff0c\u7528\u6765\u66f4\u65b0\u6709\u6548\u5b66\u4e60\u7387\u3002&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">14. verbose : bool, optional, default False,\u662f\u5426\u5c06\u8fc7\u7a0b\u6253\u5370\u5230stdout&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">15. warm_start : bool, optional, default False,\u5f53\u8bbe\u7f6e\u6210True\uff0c\u4f7f\u7528\u4e4b\u524d\u7684\u89e3\u51b3\u65b9\u6cd5\u4f5c\u4e3a\u521d\u59cb\u62df\u5408\uff0c\u5426\u5219\u91ca\u653e\u4e4b\u524d\u7684\u89e3\u51b3\u65b9\u6cd5\u3002&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">16. momentum : float, default 0.9,Momentum(\u52a8\u91cf\uff09 for gradient descent update. Should be between 0 and 1. Only used when solver=\u2019sgd\u2019.&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">17. nesterovs_momentum : boolean, default True, Whether to use Nesterov\u2019s momentum. Only used when solver=\u2019sgd\u2019 and momentum &gt; 0.&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">18. early_stopping : bool, default False,Only effective when solver=\u2019sgd\u2019 or \u2018adam\u2019,\u5224\u65ad\u5f53\u9a8c\u8bc1\u6548\u679c\u4e0d\u518d\u6539\u5584\u7684\u65f6\u5019\u662f\u5426\u7ec8\u6b62\u8bad\u7ec3\uff0c\u5f53\u4e3aTrue\u65f6\uff0c\u81ea\u52a8\u9009\u51fa10%\u7684\u8bad\u7ec3\u6570\u636e\u7528\u4e8e\u9a8c\u8bc1\u5e76\u5728\u4e24\u6b65\u8fde\u7eed\u7239\u8fed\u4ee3\u6539\u5584\u4f4e\u4e8etol\u65f6\u7ec8\u6b62\u8bad\u7ec3\u3002&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">19. validation_fraction : float, optional, default 0.1,\u7528\u4f5c\u65e9\u671f\u505c\u6b62\u9a8c\u8bc1\u7684\u9884\u7559\u8bad\u7ec3\u6570\u636e\u96c6\u7684\u6bd4\u4f8b\uff0c\u65e90-1\u4e4b\u95f4\uff0c\u53ea\u5f53early_stopping=True\u6709\u7528&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">20. beta_1 : float, optional, default 0.9\uff0cOnly used when solver=\u2019adam\u2019\uff0c\u4f30\u8ba1\u4e00\u9636\u77e9\u5411\u91cf\u7684\u6307\u6570\u8870\u51cf\u901f\u7387\uff0c[0,1)\u4e4b\u95f4&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">21. beta_2 : float, optional, default 0.999,Only used when solver=\u2019adam\u2019\u4f30\u8ba1\u4e8c\u9636\u77e9\u5411\u91cf\u7684\u6307\u6570\u8870\u51cf\u901f\u7387[0,1)\u4e4b\u95f4&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">22. epsilon : float, optional, default 1e-8,Only used when solver=\u2019adam\u2019\u6570\u503c\u7a33\u5b9a\u503c\u3002&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">\u5c5e\u6027\u8bf4\u660e\uff1a&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; classes_:\u6bcf\u4e2a\u8f93\u51fa\u7684\u7c7b\u6807\u7b7e&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; loss_:\u635f\u5931\u51fd\u6570\u8ba1\u7b97\u51fa\u6765\u7684\u5f53\u524d\u635f\u5931\u503c&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; coefs_:\u5217\u8868\u4e2d\u7684\u7b2ci\u4e2a\u5143\u7d20\u8868\u793ai\u5c42\u7684\u6743\u91cd\u77e9\u9635&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; intercepts_:\u5217\u8868\u4e2d\u7b2ci\u4e2a\u5143\u7d20\u4ee3\u8868i+1\u5c42\u7684\u504f\u5dee\u5411\u91cf&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; n_iter_ \uff1a\u8fed\u4ee3\u6b21\u6570&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; n_layers_:\u5c42\u6570&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; n_outputs_:\u8f93\u51fa\u7684\u4e2a\u6570&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; out_activation_:\u8f93\u51fa\u6fc0\u6d3b\u51fd\u6570\u7684\u540d\u79f0\u3002&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">\u65b9\u6cd5\u8bf4\u660e\uff1a&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; fit(X,y):\u62df\u5408&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; get_params([deep]):\u83b7\u53d6\u53c2\u6570&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; predict(X):\u4f7f\u7528MLP\u8fdb\u884c\u9884\u6d4b&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; predic_log_proba(X):\u8fd4\u56de\u5bf9\u6570\u6982\u7387\u4f30\u8ba1&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; predic_proba(X)\uff1a\u6982\u7387\u4f30\u8ba1&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">&#8211; score(X,y[,sample_weight]):\u8fd4\u56de\u7ed9\u5b9a\u6d4b\u8bd5\u6570\u636e\u548c\u6807\u7b7e\u4e0a\u7684\u5e73\u5747\u51c6\u786e\u5ea6&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">-set_params(**params):\u8bbe\u7f6e\u53c2\u6570\u3002<\/span><br \/>\n<span style=\"font-size:16px;\">&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">\u4f5c\u8005\uff1ahaiyu94&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">\u6765\u6e90\uff1aCSDN&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">\u539f\u6587\uff1ahttps:\/\/blog.csdn.net\/haiyu94\/article\/details\/53001726&nbsp;<\/span><br \/>\n<span style=\"font-size:16px;\">\u7248\u6743\u58f0\u660e\uff1a\u672c\u6587\u4e3a\u535a\u4e3b\u539f\u521b\u6587\u7ae0\uff0c\u8f6c\u8f7d\u8bf7\u9644\u4e0a\u535a\u6587\u94fe\u63a5\uff01<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t\n<\/p>\n<p style=\"margin:10px auto;\">\n\t<span style=\"font-size:16px;\">MLP\u56de\u5f52<\/span><span style=\"font-size:16px;\">\u6848\u4f8b\uff1a<\/span>\n<\/p>\n<p style=\"margin:10px auto;\">\n\t\n<\/p>\n<pre class=\"prettyprint lang-py linenums\"># -*- coding:utf-8 -*-\nimport matplotlib.pyplot as plt\nfrom sklearn.neural_network import MLPRegressor\n\nx_train = []\ny_train = []\n\nfor i in range(100):\n    if i in range(50, 70):\n        continue\n    y = 8 * i ** 4 + 7\n    #y = 8 * i + 7\n    x_train.append([i])\n    y_train.append(y)\n\nforestReg=MLPRegressor(solver='lbfgs', hidden_layer_sizes=(2000,), random_state=1)\nforestReg.fit(x_train, y_train)\n\nx_test = [[i] for i in range(50, 70)]\nprint x_test\np = list(forestReg.predict(x_test))\nprint p\n\nplt.scatter(map(lambda x: x[0], x_train), y_train, c=\"green\")\nplt.scatter(map(lambda x: x[0], x_test), p, marker=\"s\", c=\"yellow\")\nplt.show()<\/pre>\n<pre class=\"prettyprint lang-py linenums\"><\/pre>\n<pre class=\"prettyprint lang-py linenums\"> <a target=\"_blank\" href=\"http:\/\/www.weizn.net\/content\/uploadfile\/201811\/7a281542393569.png\" data-rel=\"penci-gallery-image-content\"  id=\"ematt:612\"><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/www.weizn.net\/content\/uploadfile\/201811\/7a281542393569.png\" title=\"\u70b9\u51fb\u67e5\u770b\u539f\u56fe\" alt=\"BaiduShurufa_2018-11-17_2-41-3.png\" border=\"0\" width=\"640\" height=\"480\" \/><\/a> <\/pre>\n<p>\n\t\n<\/p>\n<p>\n\t\n<\/p>\n<p>\n\t\n<\/p>\n<pre class=\"prettyprint lang-py linenums\"># -*- coding:utf-8 -*-\n\nimport time\nimport math\nimport matplotlib.pyplot as plt\nfrom sklearn.neural_network import MLPRegressor\nimport copy\n\nx_train = []\ny_train = []\n\nfor i in range(5000):\n    if i in range(5000, 5500):\n        continue\n    y = 8 * i ** 3 + 7\n    x_train.append([i])\n    y_train.append(y)\n\nforestReg=MLPRegressor(solver='lbfgs', max_iter=3000, activation=\"relu\",hidden_layer_sizes=(2000,), random_state=int(time.time()))\nforestReg.fit(copy.deepcopy(x_train), copy.deepcopy(y_train))\n\nx_test = [[i] for i in range(5000, 5500)]\nprint x_test\np = list(forestReg.predict(copy.deepcopy(x_test)))\nprint p\n\nplt.scatter(map(lambda x: x[0], x_train), y_train, c=\"green\")\nplt.scatter(map(lambda x: x[0], x_test), p, marker=\"s\", c=\"yellow\")\nplt.show()\n\n\n<\/pre>\n<p><a target=\"_blank\" href=\"http:\/\/www.weizn.net\/content\/uploadfile\/201811\/061a1542440083.png\" data-rel=\"penci-gallery-image-content\"  id=\"ematt:614\"><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/www.weizn.net\/content\/uploadfile\/201811\/061a1542440083.png\" title=\"\u70b9\u51fb\u67e5\u770b\u539f\u56fe\" alt=\"BaiduShurufa_2018-11-17_15-37-23.png\" border=\"0\" width=\"640\" height=\"480\" \/><\/a> <\/p>\n<p>\n\t\n<\/p>\n<p>\n\t\n<\/p>\n<p>\n\t<\/p>\n","protected":false},"excerpt":{"rendered":"<p style=\"margin:10px auto;font-family:&quot;font-size:14px;white-space:normal;background-color:#FFFFFF;\">\n\t<span style=\"font-size:18px;\">\u591a\u5c42\u611f\u77e5\u673a\uff08MLP\uff0cMultilayer Perceptron\uff09\u4e5f\u53eb\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\uff08ANN\uff0cArtificial Neural Network\uff09\uff0c\u9664\u4e86\u8f93\u5165\u8f93\u51fa\u5c42\uff0c\u5b83\u4e2d\u95f4\u53ef\u4ee5\u6709\u591a\u4e2a\u9690\u5c42\uff0c\u6700\u7b80\u5355\u7684MLP\u53ea\u542b\u4e00\u4e2a\u9690\u5c42\uff0c\u5373\u4e09\u5c42\u7684\u7ed3\u6784\uff0c\u5982\u4e0b\u56fe\uff1a<\/span>\n<\/p>\n<p style=\"margin:10px auto;font-family:&quot;font-size:14px;white-space:normal;background-color:#FFFFFF;text-align:center;\">\n\t<span style=\"font-size:18px;\"><img decoding=\"async\" src=\"https:\/\/img-blog.csdn.net\/20150128033221168?watermark\/2\/text\/aHR0cDovL2Jsb2cuY3Nkbi5uZXQvdTAxMjE2MjYxMw==\/font\/5a6L5L2T\/fontsize\/400\/fill\/I0JBQkFCMA==\/dissolve\/70\/gravity\/Center\" alt=\"\" style=\"max-width:900px;height:auto;\" \/><br \/>\n<\/span>\n<\/p>\n<p style=\"margin:10px auto;font-family:&quot;font-size:14px;white-space:normal;background-color:#FFFFFF;\">\n\t<span style=\"font-size:18px;\">&nbsp;<\/span>\n<\/p>\n<p style=\"margin:10px auto;font-family:&quot;font-size:14px;white-space:normal;background-color:#FFFFFF;\">\n\t<span style=\"font-size:18px;\">\u4ece\u4e0a\u56fe\u53ef\u4ee5\u770b\u5230\uff0c\u591a\u5c42\u611f\u77e5\u673a\u5c42\u4e0e\u5c42\u4e4b\u95f4\u662f\u5168\u8fde\u63a5\u7684\uff08\u5168\u8fde\u63a5\u7684\u610f\u601d\u5c31\u662f\uff1a\u4e0a\u4e00\u5c42\u7684\u4efb\u4f55\u4e00\u4e2a\u795e\u7ecf\u5143\u4e0e\u4e0b\u4e00\u5c42\u7684\u6240\u6709\u795e\u7ecf\u5143\u90fd\u6709\u8fde\u63a5\uff09\u3002\u591a\u5c42\u611f\u77e5\u673a\u6700\u5e95\u5c42\u662f\u8f93\u5165\u5c42\uff0c\u4e2d\u95f4\u662f\u9690\u85cf&#8230;<\/span>\n<\/p>\n<dl class=\"class\" style=\"font-family:&quot;font-size:14px;white-space:normal;background-color:#FFFFFF;\">\n<dd>\n<table class=\"docutils field-list\" frame=\"void\" rules=\"none\" style=\"border:1px solid silver;word-break:break-word;\">\n<colgroup>\n<col class=\"field-name\" \/>\n<col class=\"field-body\" \/><\/colgroup>\n<\/table>\n<\/dd>\n<\/dl>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[337],"tags":[],"class_list":["post-253","post","type-post","status-publish","format-standard","hentry"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v16.9 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>\u591a\u5c42\u611f\u77e5\u673a\uff08MLP\uff09\u539f\u7406\u7b80\u4ecb - Wayne&#039;s Blog<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"http:\/\/weizn.net\/?p=253\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"\u591a\u5c42\u611f\u77e5\u673a\uff08MLP\uff09\u539f\u7406\u7b80\u4ecb - Wayne&#039;s Blog\" \/>\n<meta property=\"og:description\" content=\"\u591a\u5c42\u611f\u77e5\u673a\uff08MLP\uff0cMultilayer Perceptron\uff09\u4e5f\u53eb\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\uff08ANN\uff0cArtificial Neural Network\uff09\uff0c\u9664\u4e86\u8f93\u5165\u8f93\u51fa\u5c42\uff0c\u5b83\u4e2d\u95f4\u53ef\u4ee5\u6709\u591a\u4e2a\u9690\u5c42\uff0c\u6700\u7b80\u5355\u7684MLP\u53ea\u542b\u4e00\u4e2a\u9690\u5c42\uff0c\u5373\u4e09\u5c42\u7684\u7ed3\u6784\uff0c\u5982\u4e0b\u56fe\uff1a      &nbsp;   \u4ece\u4e0a\u56fe\u53ef\u4ee5\u770b\u5230\uff0c\u591a\u5c42\u611f\u77e5\u673a\u5c42\u4e0e\u5c42\u4e4b\u95f4\u662f\u5168\u8fde\u63a5\u7684\uff08\u5168\u8fde\u63a5\u7684\u610f\u601d\u5c31\u662f\uff1a\u4e0a\u4e00\u5c42\u7684\u4efb\u4f55\u4e00\u4e2a\u795e\u7ecf\u5143\u4e0e\u4e0b\u4e00\u5c42\u7684\u6240\u6709\u795e\u7ecf\u5143\u90fd\u6709\u8fde\u63a5\uff09\u3002\u591a\u5c42\u611f\u77e5\u673a\u6700\u5e95\u5c42\u662f\u8f93\u5165\u5c42\uff0c\u4e2d\u95f4\u662f\u9690\u85cf...\" \/>\n<meta property=\"og:url\" content=\"http:\/\/weizn.net\/?p=253\" \/>\n<meta property=\"og:site_name\" content=\"Wayne&#039;s Blog\" \/>\n<meta property=\"article:published_time\" content=\"2018-03-16T18:33:54+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/img-blog.csdn.net\/20150128033221168?watermark\/2\/text\/aHR0cDovL2Jsb2cuY3Nkbi5uZXQvdTAxMjE2MjYxMw==\/font\/5a6L5L2T\/fontsize\/400\/fill\/I0JBQkFCMA==\/dissolve\/70\/gravity\/Center\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"zinan\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"8 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebSite\",\"@id\":\"http:\/\/weizn.net\/#website\",\"url\":\"http:\/\/weizn.net\/\",\"name\":\"Wayne&#039;s Blog\",\"description\":\"\",\"publisher\":{\"@id\":\"http:\/\/weizn.net\/#\/schema\/person\/e88bc12c590502d8b6249326f960b264\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"http:\/\/weizn.net\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"ImageObject\",\"@id\":\"http:\/\/weizn.net\/?p=253#primaryimage\",\"inLanguage\":\"zh-Hans\",\"url\":\"https:\/\/img-blog.csdn.net\/20150128033221168?watermark\/2\/text\/aHR0cDovL2Jsb2cuY3Nkbi5uZXQvdTAxMjE2MjYxMw==\/font\/5a6L5L2T\/fontsize\/400\/fill\/I0JBQkFCMA==\/dissolve\/70\/gravity\/Center\",\"contentUrl\":\"https:\/\/img-blog.csdn.net\/20150128033221168?watermark\/2\/text\/aHR0cDovL2Jsb2cuY3Nkbi5uZXQvdTAxMjE2MjYxMw==\/font\/5a6L5L2T\/fontsize\/400\/fill\/I0JBQkFCMA==\/dissolve\/70\/gravity\/Center\"},{\"@type\":\"WebPage\",\"@id\":\"http:\/\/weizn.net\/?p=253#webpage\",\"url\":\"http:\/\/weizn.net\/?p=253\",\"name\":\"\\u591a\\u5c42\\u611f\\u77e5\\u673a\\uff08MLP\\uff09\\u539f\\u7406\\u7b80\\u4ecb - Wayne&#039;s Blog\",\"isPartOf\":{\"@id\":\"http:\/\/weizn.net\/#website\"},\"primaryImageOfPage\":{\"@id\":\"http:\/\/weizn.net\/?p=253#primaryimage\"},\"datePublished\":\"2018-03-16T18:33:54+00:00\",\"dateModified\":\"2018-03-16T18:33:54+00:00\",\"breadcrumb\":{\"@id\":\"http:\/\/weizn.net\/?p=253#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"http:\/\/weizn.net\/?p=253\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"http:\/\/weizn.net\/?p=253#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\\u9996\\u9875\",\"item\":\"http:\/\/weizn.net\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\\u591a\\u5c42\\u611f\\u77e5\\u673a\\uff08MLP\\uff09\\u539f\\u7406\\u7b80\\u4ecb\"}]},{\"@type\":\"Article\",\"@id\":\"http:\/\/weizn.net\/?p=253#article\",\"isPartOf\":{\"@id\":\"http:\/\/weizn.net\/?p=253#webpage\"},\"author\":{\"@id\":\"http:\/\/weizn.net\/#\/schema\/person\/e88bc12c590502d8b6249326f960b264\"},\"headline\":\"\\u591a\\u5c42\\u611f\\u77e5\\u673a\\uff08MLP\\uff09\\u539f\\u7406\\u7b80\\u4ecb\",\"datePublished\":\"2018-03-16T18:33:54+00:00\",\"dateModified\":\"2018-03-16T18:33:54+00:00\",\"mainEntityOfPage\":{\"@id\":\"http:\/\/weizn.net\/?p=253#webpage\"},\"wordCount\":1269,\"commentCount\":0,\"publisher\":{\"@id\":\"http:\/\/weizn.net\/#\/schema\/person\/e88bc12c590502d8b6249326f960b264\"},\"image\":{\"@id\":\"http:\/\/weizn.net\/?p=253#primaryimage\"},\"thumbnailUrl\":\"https:\/\/img-blog.csdn.net\/20150128033221168?watermark\/2\/text\/aHR0cDovL2Jsb2cuY3Nkbi5uZXQvdTAxMjE2MjYxMw==\/font\/5a6L5L2T\/fontsize\/400\/fill\/I0JBQkFCMA==\/dissolve\/70\/gravity\/Center\",\"articleSection\":[\"\\u6570\\u636e\\u5206\\u6790\"],\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"http:\/\/weizn.net\/?p=253#respond\"]}]},{\"@type\":[\"Person\",\"Organization\"],\"@id\":\"http:\/\/weizn.net\/#\/schema\/person\/e88bc12c590502d8b6249326f960b264\",\"name\":\"zinan\",\"logo\":{\"@id\":\"http:\/\/weizn.net\/#personlogo\"}}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"\u591a\u5c42\u611f\u77e5\u673a\uff08MLP\uff09\u539f\u7406\u7b80\u4ecb - Wayne&#039;s Blog","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":"http:\/\/weizn.net\/?p=253","og_locale":"zh_CN","og_type":"article","og_title":"\u591a\u5c42\u611f\u77e5\u673a\uff08MLP\uff09\u539f\u7406\u7b80\u4ecb - Wayne&#039;s Blog","og_description":"\u591a\u5c42\u611f\u77e5\u673a\uff08MLP\uff0cMultilayer Perceptron\uff09\u4e5f\u53eb\u4eba\u5de5\u795e\u7ecf\u7f51\u7edc\uff08ANN\uff0cArtificial Neural Network\uff09\uff0c\u9664\u4e86\u8f93\u5165\u8f93\u51fa\u5c42\uff0c\u5b83\u4e2d\u95f4\u53ef\u4ee5\u6709\u591a\u4e2a\u9690\u5c42\uff0c\u6700\u7b80\u5355\u7684MLP\u53ea\u542b\u4e00\u4e2a\u9690\u5c42\uff0c\u5373\u4e09\u5c42\u7684\u7ed3\u6784\uff0c\u5982\u4e0b\u56fe\uff1a      &nbsp;   \u4ece\u4e0a\u56fe\u53ef\u4ee5\u770b\u5230\uff0c\u591a\u5c42\u611f\u77e5\u673a\u5c42\u4e0e\u5c42\u4e4b\u95f4\u662f\u5168\u8fde\u63a5\u7684\uff08\u5168\u8fde\u63a5\u7684\u610f\u601d\u5c31\u662f\uff1a\u4e0a\u4e00\u5c42\u7684\u4efb\u4f55\u4e00\u4e2a\u795e\u7ecf\u5143\u4e0e\u4e0b\u4e00\u5c42\u7684\u6240\u6709\u795e\u7ecf\u5143\u90fd\u6709\u8fde\u63a5\uff09\u3002\u591a\u5c42\u611f\u77e5\u673a\u6700\u5e95\u5c42\u662f\u8f93\u5165\u5c42\uff0c\u4e2d\u95f4\u662f\u9690\u85cf...","og_url":"http:\/\/weizn.net\/?p=253","og_site_name":"Wayne&#039;s Blog","article_published_time":"2018-03-16T18:33:54+00:00","og_image":[{"url":"https:\/\/img-blog.csdn.net\/20150128033221168?watermark\/2\/text\/aHR0cDovL2Jsb2cuY3Nkbi5uZXQvdTAxMjE2MjYxMw==\/font\/5a6L5L2T\/fontsize\/400\/fill\/I0JBQkFCMA==\/dissolve\/70\/gravity\/Center"}],"twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"zinan","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"8 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebSite","@id":"http:\/\/weizn.net\/#website","url":"http:\/\/weizn.net\/","name":"Wayne&#039;s Blog","description":"","publisher":{"@id":"http:\/\/weizn.net\/#\/schema\/person\/e88bc12c590502d8b6249326f960b264"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"http:\/\/weizn.net\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"zh-Hans"},{"@type":"ImageObject","@id":"http:\/\/weizn.net\/?p=253#primaryimage","inLanguage":"zh-Hans","url":"https:\/\/img-blog.csdn.net\/20150128033221168?watermark\/2\/text\/aHR0cDovL2Jsb2cuY3Nkbi5uZXQvdTAxMjE2MjYxMw==\/font\/5a6L5L2T\/fontsize\/400\/fill\/I0JBQkFCMA==\/dissolve\/70\/gravity\/Center","contentUrl":"https:\/\/img-blog.csdn.net\/20150128033221168?watermark\/2\/text\/aHR0cDovL2Jsb2cuY3Nkbi5uZXQvdTAxMjE2MjYxMw==\/font\/5a6L5L2T\/fontsize\/400\/fill\/I0JBQkFCMA==\/dissolve\/70\/gravity\/Center"},{"@type":"WebPage","@id":"http:\/\/weizn.net\/?p=253#webpage","url":"http:\/\/weizn.net\/?p=253","name":"\u591a\u5c42\u611f\u77e5\u673a\uff08MLP\uff09\u539f\u7406\u7b80\u4ecb - Wayne&#039;s Blog","isPartOf":{"@id":"http:\/\/weizn.net\/#website"},"primaryImageOfPage":{"@id":"http:\/\/weizn.net\/?p=253#primaryimage"},"datePublished":"2018-03-16T18:33:54+00:00","dateModified":"2018-03-16T18:33:54+00:00","breadcrumb":{"@id":"http:\/\/weizn.net\/?p=253#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["http:\/\/weizn.net\/?p=253"]}]},{"@type":"BreadcrumbList","@id":"http:\/\/weizn.net\/?p=253#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"http:\/\/weizn.net\/"},{"@type":"ListItem","position":2,"name":"\u591a\u5c42\u611f\u77e5\u673a\uff08MLP\uff09\u539f\u7406\u7b80\u4ecb"}]},{"@type":"Article","@id":"http:\/\/weizn.net\/?p=253#article","isPartOf":{"@id":"http:\/\/weizn.net\/?p=253#webpage"},"author":{"@id":"http:\/\/weizn.net\/#\/schema\/person\/e88bc12c590502d8b6249326f960b264"},"headline":"\u591a\u5c42\u611f\u77e5\u673a\uff08MLP\uff09\u539f\u7406\u7b80\u4ecb","datePublished":"2018-03-16T18:33:54+00:00","dateModified":"2018-03-16T18:33:54+00:00","mainEntityOfPage":{"@id":"http:\/\/weizn.net\/?p=253#webpage"},"wordCount":1269,"commentCount":0,"publisher":{"@id":"http:\/\/weizn.net\/#\/schema\/person\/e88bc12c590502d8b6249326f960b264"},"image":{"@id":"http:\/\/weizn.net\/?p=253#primaryimage"},"thumbnailUrl":"https:\/\/img-blog.csdn.net\/20150128033221168?watermark\/2\/text\/aHR0cDovL2Jsb2cuY3Nkbi5uZXQvdTAxMjE2MjYxMw==\/font\/5a6L5L2T\/fontsize\/400\/fill\/I0JBQkFCMA==\/dissolve\/70\/gravity\/Center","articleSection":["\u6570\u636e\u5206\u6790"],"inLanguage":"zh-Hans","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["http:\/\/weizn.net\/?p=253#respond"]}]},{"@type":["Person","Organization"],"@id":"http:\/\/weizn.net\/#\/schema\/person\/e88bc12c590502d8b6249326f960b264","name":"zinan","logo":{"@id":"http:\/\/weizn.net\/#personlogo"}}]}},"_links":{"self":[{"href":"http:\/\/weizn.net\/index.php?rest_route=\/wp\/v2\/posts\/253","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/weizn.net\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/weizn.net\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/weizn.net\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/weizn.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=253"}],"version-history":[{"count":0,"href":"http:\/\/weizn.net\/index.php?rest_route=\/wp\/v2\/posts\/253\/revisions"}],"wp:attachment":[{"href":"http:\/\/weizn.net\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=253"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/weizn.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=253"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/weizn.net\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=253"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}