{"id":1850,"date":"2019-05-08T14:08:15","date_gmt":"2019-05-08T06:08:15","guid":{"rendered":"http:\/\/nick.txtcc.com\/?p=1850"},"modified":"2019-05-08T18:10:01","modified_gmt":"2019-05-08T10:10:01","slug":"%e5%b0%8f%e7%99%bd%e5%ad%a6cnn%e4%bb%a5%e5%8f%8akeras%e7%9a%84%e9%80%9f%e6%88%90","status":"publish","type":"post","link":"http:\/\/nick.txtcc.com\/index.php\/python\/1850","title":{"rendered":"\u5c0f\u767d\u5b66CNN\u4ee5\u53caKeras\u7684\u901f\u6210"},"content":{"rendered":"

\u4e00\u3001\u4e3a\u4f55\u8981\u7528Keras<\/h3>\n
\u5982\u4eca\u5728\u6df1\u5ea6\u5b66\u4e60\u5927\u706b\u7684\u65f6\u5019\uff0c\u7b2c\u4e09\u65b9\u5de5\u5177\u4e5f\u5c42\u51fa\u4e0d\u7a77\uff0c\u6bd4\u8f83\u51fa\u540d\u7684\u6709Tensorflow\uff0cCaffe\uff0cTheano\uff0cMXNet\uff0c\u5728\u5982\u6b64\u591a\u7684\u7b2c\u4e09\u65b9\u6846\u67b6\u4e2d\u9891\u7e41\u7684\u66f4\u6362\u65e0\u7591\u662f\u5f88\u4f4e\u6548\u7684\uff0c\u53ea\u8981\u4f60\u80fd\u591f\u597d\u597d\u638c\u63e1\u5176\u4e2d\u4e00\u4e2a\u6846\u67b6\uff0c\u719f\u6089\u5176\u539f\u7406\uff0c\u90a3\u4e48\u4e4b\u540e\u56e0\u4e3a\u5404\u79cd\u8981\u6c42\u4f60\u60f3\u8981\u66f4\u6362\u6846\u67b6\u4e5f\u662f\u5f88\u5bb9\u6613\u7684\u3002<\/p>\n
\u90a3\u4e48sherlock\u7528\u7684\u662f\u54ea\u4e2a\u6846\u67b6\u5462\uff1fsherlock\u4f7f\u7528\u7684\u662fGoogle\u7684\u5f00\u6e90\u6846\u67b6Tensorflow\uff0c\u56e0\u4e3aGoogle\u5f00\u6e90\u4e86tensorflow\u4e4b\u540e\u5176\u793e\u533a\u975e\u5e38\u6d3b\u8dc3\uff0c\u800c\u4e14\u7248\u672c\u66f4\u65b0\u4e5f\u975e\u5e38\u7a33\u5b9a\uff0c\u6240\u4ee5\u6211\u5c31\u9009\u62e9\u4e86\u8fd9\u4e2a\u6846\u67b6\u3002\u5bf9\u4e8e\u6846\u67b6\u4e4b\u4e89\uff0c\u5728\u77e5\u4e4e\u4e0a\u5df2\u7ecf\u6709\u5f88\u591a\u4eba\u5728\u6495\u903c\u4e86\uff0c\u8fd9\u4e2a\u5c31\u597d\u6bd4\u54ea\u79cd\u7f16\u7a0b\u8bed\u8a00\u597d\u8fd9\u4e2a\u95ee\u9898\u4e00\u6837\u3002\u5bf9\u4e8e\u6211\u4eec\u6765\u8bb2\uff0c\u9009\u62e9\u4e00\u4e2a\u7a33\u5b9a\u7684\u6846\u67b6\uff0c\u597d\u597d\u7684\u5b66\u4e60deep learning\u624d\u662f\u91cd\u4e2d\u4e4b\u91cd\uff0c\u5bf9\u4e8e\u54ea\u79cd\u6846\u67b6\u66f4\u597d\u7684\u95ee\u9898\uff0c\u6211\u4eec\u5b66\u597d\u4e4b\u540e\u81ea\u7136\u6709\u81ea\u5df1\u7684\u89c1\u89e3\uff0c\u6240\u4ee5\u524d\u671f\u5207\u5fcc\u5728\u5237\u77e5\u4e4e\u542c\u5b8c\u5927\u795e\u6495\u903c\u4e4b\u540e\u9891\u7e41\u66f4\u6362\u6846\u67b6\u3002<\/p>\n
\u5bf9\u4e8eTensorflow\u7684\u5b89\u88c5\uff0c\u4ee5\u53caCPU\u548cGPU\u7248\u672c\uff0c\u5404\u79cd\u7cfb\u7edf\u7684\u5b89\u88c5\u7f51\u4e0a\u5df2\u7ecf\u6709\u5f88\u591a\u4eba\u8be6\u7ec6\u7684\u5199\u8fc7\u653b\u7565\u4e86\uff0c\u53ef\u4ee5\u81ea\u5df1\u53bb\u7f51\u4e0a\u641c\u4e00\u641c\uff0c\u5f88\u5bb9\u6613\u5c31\u53ef\u4ee5\u5b89\u88c5\u6210\u529f\u3002<\/p>\n
\u9009\u62e9\u4e86Tensorflow\u4e4b\u540e\uff0c\u6211\u4eec\u5c31\u53ef\u4ee5\u6109\u5feb\u7684\u5f00\u59cb\u6211\u4eec\u7684\u6df1\u5ea6\u5b66\u4e60\u4e4b\u65c5\u4e86\u3002\u53bbTensorflow\u7684\u4e2d\u6587\u793e\u533a\uff0c\u53ef\u4ee5\u770b\u5230\u6709\u4e00\u4e9b\u65b0\u624b\u6559\u7a0b\uff0c\u7f51\u4e0a\u4e5f\u6709\u5f88\u591a\u5b66\u4e60\u6750\u6599\uff0c\u63a8\u8350\u770b\u770bstanford\u5927\u5b66cs224d\u7684\u8bfe\u4ef6\uff0c http:\/\/cs224d.stanford.edu\/lectures\/CS224d-Lecture7.pdf<\/a><\/a>\uff0c \u5f88\u8be6\u7ec6\u7684\u4ecb\u7ecd\u4e86tensorflow\u3002\u7136\u540e\u4f60\u5c31\u53ef\u4ee5\u5199tensorflow\u7684\u7a0b\u5e8f\u4e86\u3002\u867d\u7136\u8bf4tensorflow\u5df2\u7ecf\u662f\u4e00\u4e2a\u5c01\u88c5\u597d\u7684\u6846\u67b6\uff0c\u4f46\u662f\u4f60\u53d1\u73b0\u4f60\u5199\u4e00\u4e2a\u7b80\u5355\u7684\u795e\u7ecf\u7f51\u7edc\u4e5f\u9700\u8981\u5f88\u591a\u884c\u624d\u80fd\u591f\u5199\u5b8c\uff0c\u8fd9\u4e2a\u65f6\u5019\uff0c\u5c31\u6709\u5f88\u591a\u7684\u7b2c\u4e09\u65b9\u63d2\u67b6\u6765\u5e2e\u52a9\u4f60\u5199\u7f51\u7edc\uff0c\u4e5f\u5c31\u662f\u8bf4\u4f60\u7528tensorflow\u8981\u519910\u884c\uff0c\u7b2c\u4e09\u65b9\u63d2\u67b6\u5e2e\u4f60\u5c01\u88c5\u4e86\u4e00\u4e2a\u51fd\u6570\uff0c\u5c31\u662f\u628a\u8fd910\u884c\u96c6\u5408\u5728\u8fd9\u4e2a\u51fd\u6570\u91cc\u9762\uff0c\u90a3\u4e48\u4f60\u75281\u884c\uff0c\u4f20\u5165\u76f8\u540c\u7684\u53c2\u6570\uff0c\u5c31\u80fd\u591f\u8fbe\u523010\u884c\u76f8\u540c\u7684\u6548\u679c\uff0c\u5982\u6b64\u7b80\u4fbf\u5e76\u4e14\u8282\u7ea6\u65f6\u95f4\uff0c\u53ef\u4ee5\u5e2e\u52a9\u5f88\u5feb\u7684\u5b9e\u73b0\u6211\u4eec\u7684\u60f3\u6cd5\u3002<\/p>\n
Keras Documentation<\/a>\u00a0\u5c31\u662fKeras\u7684\u5b98\u65b9\u6587\u6863\uff0c\u91cc\u9762\u53ef\u4ee5\u67e5\u9605\u6240\u6709\u7684\u51fd\u6570\uff0c\u5e76\u4e14\u53ef\u4ee5\u5728github\u4e0a\u770b\u4ed6\u7684\u5f00\u6e90\u4ee3\u7801\uff0c\u975e\u5e38\u65b9\u4fbf\u3002\u5b89\u88c5\u4e5f\u5f88\u7b80\u5355\uff0c\u6253\u5f00\u7ec8\u7aef\uff0c\u8f93\u5165pip install keras \u5c31\u53ef\u4ee5\u7b49\u5f85\u5b89\u88c5\u4e86\u3002<\/p>\n
\u4e0b\u9762\u5c31\u7ed9\u4e00\u4e2a\u7b80\u5355\u7684\u4f8b\u5b50\uff0c\u6765\u770b\u4e00\u770bKeras\u5230\u5e95\u6709\u591a\u7b80\u5355\u3002<\/p>\n
from keras.models import Sequential\nmodel = Sequential()<\/blockquote>\n
\u5f15\u5165sequential\uff0c\u8fd9\u4e2a\u5c31\u662f\u4e00\u4e2a\u7a7a\u7684\u7f51\u7edc\u7ed3\u6784\uff0c\u5e76\u4e14\u8fd9\u4e2a\u7ed3\u6784\u662f\u4e00\u4e2a\u987a\u5e8f\u7684\u5e8f\u5217\uff0c\u6240\u4ee5\u53ebSequential\uff0cKeras\u91cc\u9762\u8fd8\u6709\u4e00\u4e9b\u5176\u4ed6\u7684\u7f51\u7edc\u7ed3\u6784\u3002<\/p>\n
from keras.layers import Dense, Activation\n\nmodel.add(Dense(units=64, input_dim=100))\n\nmodel.add(Activation(‘relu’))\n\nmodel.add(Dense(units=10))\n\nmodel.add(Activation(‘softmax’))<\/blockquote>\n
\u53ef\u4ee5\u770b\u5230\u52a0\u5165\u5c42\u5f88\u7b80\u5355\uff0c\u53ea\u9700\u8981\u5199.add\uff0c\u540e\u9762\u662f\u8981\u52a0\u7684\u5c42\u7684\u7c7b\u578b\u3002<\/p>\n
model.compile(loss=’categorical_crossentropy’,\n\noptimizer=’sgd’,\n\nmetrics=[‘accuracy’])<\/blockquote>\n
\u4e00\u65e6\u4f60\u5199\u597d\u4e86\u7f51\u7edc\u4e4b\u540e\uff0c\u5c31\u53ef\u4ee5\u7528compile\u7f16\u8bd1\u6574\u4e2a\u7f51\u7edc\uff0c\u770b\u53c2\u6570\u8bbe\u7f6e\u6709\u6ca1\u6709\u95ee\u9898<\/p>\n
model.compile(loss=keras.losses.categorical_crossentropy,\noptimizer=keras.optimizers.SGD(lr=0.01,\u00a0momentum=0.9,\u00a0nesterov=True))<\/blockquote>\n
\u4f60\u4e5f\u53ef\u4ee5\u81ea\u5b9a\u4e49\u5176\u4e2d\u7684\u4f18\u5316\u51fd\u6570\uff0c\u5c31\u50cf\u4e0a\u9762\u8fd9\u6837\uff0c\u2019sgd\u2019\u662fKeras\u5df2\u7ecf\u5199\u597d\u4e86\u4e00\u4e9b\u9ed8\u8ba4\u53c2\u6570\u7684\u4f18\u5316\u51fd\u6570\uff0c\u4f60\u53ef\u4ee5\u81ea\u5df1\u91cd\u65b0\u5b9a\u4e49\u53c2\u6570\uff0c\u5f97\u5230\u4e00\u4e2a\u4f18\u5316\u51fd\u6570\u3002<\/p>\n
model.fit(x_train,y_train,epochs=5,batch_size=32)<\/blockquote>\n
\u8fd9\u4e2a\u5c31\u50cfscikit-learn\u4e00\u6837\u8bad\u7ec3\u6a21\u578b\u3002<\/p>\n
loss_and_metrics=model.evaluate(x_test,y_test,batch_size=128)<\/blockquote>\n
\u8fd9\u4e2a\u5c31\u662f\u8bc4\u4f30\u8bad\u7ec3\u7ed3\u679c\u3002<\/p>\n
classes=model.predict(x_test,batch_size=128)<\/blockquote>\n
\u6216\u8005\u662f\u901a\u8fc7predict\u8fdb\u884c\u9884\u6d4b\u3002<\/p>\n
\u770b\u4e86\u4e0a\u9762\u7684\u4ee3\u7801\uff0c\u76f8\u4fe1\u5f88\u591a\u719f\u6089scikit-learn\u7684\u540c\u5b66\u90fd\u5f88\u4eb2\u5207\uff0c\u56e0\u4e3a\u786e\u5b9e\u5f88\u7b80\u4fbf\uff0c\u8ddfscikit-learn\u4e5f\u6709\u7740\u7c7b\u4f3c\u7684\u8bed\u6cd5\u3002<\/p>\n
\u4e8c\u3001\u5f00\u59cb\u5b66\u4e60CNN<\/h3>\n
\u5728\u7406\u89e3CNN\u4e4b\u524d\uff0c\u6211\u4eec\u6709\u5fc5\u8981\u5148\u7406\u89e3\u4e00\u4e0b\u4ec0\u4e48\u662f\u795e\u7ecf\u7f51\u7edc\uff0c\u8fd9\u6837\u624d\u80fd\u5f00\u59cb\u4e86\u89e3\u66f4\u9ad8\u7ea7\u7684\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u3002<\/p>\n
\u8981\u5b66\u4e60\u795e\u7ecf\u7f51\u7edc\u5f53\u7136\u6709\u5f88\u591a\u9014\u5f84\uff0c\u7f51\u4e0a\u4e0d\u5c11\u7684\u5927\u725b\u5199\u4e86\u5f88\u591a\u653b\u7565\uff0c\u6709\u7684\u63a8\u5d07\u4ece\u7406\u8bba\u5230\u5de5\u7a0b\u5b8c\u6210\u6df1\u5ea6\u5b66\u4e60\uff0c\u6709\u7684\u5e0c\u671b\u4ece\u5de5\u7a0b\u51fa\u53d1\u53d1\u73b0\u95ee\u9898\uff0c\u89e3\u51b3\u95ee\u9898\u3002\u5404\u79cd\u5404\u6837\u7684\u65b9\u5f0f\u90fd\u6709\u4e0d\u540c\u7684\u4eba\u53bb\u5c1d\u8bd5\uff0c\u653b\u7565\u4e5f\u662f\u4e00\u5927\u63a8\uff0c\u8fd9\u4f7f\u5f97\u4e0d\u5c11\u7684\u5c0f\u767d\u76f4\u63a5\u5012\u5728\u4e86\u9009\u62e9\u6750\u6599\u7684\u8def\u4e0a\uff0c\u4e00\u76f4\u5728\u8865\u5148\u4fee\u77e5\u8bc6\uff0c\u5f85\u5230\u70ed\u60c5\u7ed3\u675f\u5c31\u653e\u5f03\u4e86\u5b66\u4e60\uff0c\u8fde\u5377\u79ef\u7f51\u7edc\u90fd\u4e0d\u77e5\u9053\u662f\u4ec0\u4e48\uff0c\u5927\u5927\u5730\u6253\u51fb\u4e86\u5927\u5bb6\u7684\u5b66\u4e60\u70ed\u60c5\u3002\u4eca\u5929\uff0csherlock\u5728\u8fd9\u91cc\u7ed9\u5927\u5bb6\u63a8\u8350\u4e00\u4e2a\u5b66\u4e60\u6750\u6599\uff0c\u4fdd\u8bc1\u4f60\u80fd\u591f\u5feb\u901f\u5165\u95e8cnn\uff0c\u51fa\u53bb\u88c5\u903c\u4e5f\u80fd\u591f\u548c\u522b\u4eba\u804a\u51e0\u53e5\u3002<\/p>\n
\u8fd9\u4e2a\u6750\u6599\u662f\u4ec0\u4e48\u5462\uff0c\u5c31\u662f\u5927\u540d\u9f0e\u9f0e\u7684standford\u7684cs231n\u8fd9\u95e8\u8bfe\u7a0b\u3002\u00a0CS231n Convolutional Neural Networks for Visual Recognition<\/a>\u00a0\u00a0stanford\u5927\u5b66\u786e\u5b9e\u7b97\u662f\u6df1\u5ea6\u5b66\u4e60\u548c\u4eba\u5de5\u667a\u80fd\u9886\u57df\u975e\u5e38\u725b\u903c\u7684\u5b66\u6821\u3002<\/p>\n
\u795e\u7ecf\u7f51\u7edc<\/strong><\/p>\n
\u5e9f\u8bdd\u4e0d\u591a\u8bf4\uff0c\u5f00\u59cb\u5b66\u4e60\u6211\u4eec\u7684\u795e\u7ecf\u7f51\u7edc\u3002<\/p>\n $\"\u5c0f\u767d\u5b66CNN\u4ee5\u53caKeras\u7684\u901f\u6210\"$ \n
\u8fd9\u662f\u4e00\u5f20\u8111\u795e\u7ecf\u7684\u56fe\u7247\uff0c\u795e\u7ecf\u7f51\u7edc\u7684\u53d1\u660e\u4e5f\u662f\u7531\u6b64\u5f00\u59cb\u7684\uff0c\u8fd9\u5c31\u662f\u6240\u8c13\u7684\u4e00\u4e2a\u795e\u7ecf\u5143\uff0c\u4e0a\u9762\u6709\u5404\u79cd\u63a5\u53d7\u7a81\u89e6\uff0c\u7136\u540e\u901a\u8fc7\u4e00\u4e2a\u8111\u795e\u7ecf\u6765\u63a5\u53d7\uff0c\u6700\u540e\u5f97\u5230\u8f93\u51fa\u7684\u7ed3\u679c\u3002<\/p>\n
\u90a3\u4e48\u7531\u8fd9\u5f20\u8111\u795e\u7ecf\u56fe\u80fd\u591f\u62bd\u8c61\u51fa\u6765\u7684\u795e\u7ecf\u7f51\u7edc\u662f\u4ec0\u4e48\u5462?\u5c31\u662f\u4e0b\u9762\u8fd9\u4e2a\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\u3002<\/p>\n $\"\u5c0f\u767d\u5b66CNN\u4ee5\u53caKeras\u7684\u901f\u6210\"$ \n
\u8fd9\u4e2a\u600e\u4e48\u7406\u89e3\u5462\uff1f\u5c31\u662f\u8f93\u5165\u4e00\u4e2a\u5411\u91cf\uff0c\u7136\u540e\u7ed9\u5411\u91cf\u7684\u6bcf\u4e00\u4e2a\u5143\u7d20\u5206\u914d\u4e00\u4e2a\u6743\u91cd\uff0c\u7136\u540e\u901a\u8fc7\u6743\u91cd\u6c42\u548c\u5f97\u5230\u4e00\u4e2a\u7ed3\u679c\uff0c\u7136\u540e\u5c06\u8fd9\u4e2a\u7ed3\u679c\u8f93\u5165\u4e00\u4e2a\u6fc0\u6d3b\u51fd\u6570\uff0c\u5f97\u5230\u6700\u540e\u7684\u8f93\u51fa\u7ed3\u679c\u3002<\/p>\n
\u6fc0\u6d3b\u51fd\u6570\u53c8\u662f\u4ec0\u4e48\u9b3c\uff1f\u6fc0\u6d3b\u51fd\u6570\u7684\u51fa\u73b0\u662f\u56e0\u4e3a\u4eba\u8111\u7684\u6784\u9020\uff0c\u4eba\u8111\u91cc\u9762\u63a5\u53d7\u4fe1\u606f\u5f97\u5230\u7ed3\u679c\u8fd9\u4e2a\u8fc7\u7a0b\u662f\u975e\u7ebf\u6027\u7684\uff0c\u6bd4\u5982\u4f60\u770b\u5230\u4e00\u6837\u4e1c\u897f\uff0c\u4f60\u4e0d\u53ef\u80fd\u4fdd\u7559\u8fd9\u4e2a\u4e1c\u897f\u7684\u5168\u90e8\u7279\u5f81\uff0c\u4f60\u4f1a\u91cd\u70b9\u89c2\u5bdf\u4f60\u611f\u5174\u8da3\u7684\u5730\u65b9\uff0c\u8fd9\u5c31\u662f\u975e\u7ebf\u6027\u7684\uff0c\u4e5f\u5c31\u662f\u8bf4\u9700\u8981\u4e00\u4e2a\u975e\u7ebf\u6027\u53d8\u5316\u5c06\u8f93\u5165\u7684\u7ed3\u679c\u53d8\u6362\u4e3a\u975e\u7ebf\u6027\u7684\u7ed3\u679c\u3002\u73b0\u5728\u5e38\u7528\u7684\u975e\u7ebf\u6027\u51fd\u6570\u5c31\u662fRelu(x) = max(x, 0)\uff0c\u5c31\u662f\u5c06\u5c0f\u4e8e0\u7684\u90e8\u5206\u53bb\u6389\uff0c\u53ea\u4fdd\u7559\u5927\u4e8e0\u7684\u90e8\u5206\u3002<\/p>\n
\u8fd9\u5c31\u662f\u5355\u5143\u7684\u8f93\u5165\u548c\u8f93\u51fa\uff0c\u5c06\u8fd9\u4e9b\u5355\u5143\u5408\u5728\u4e00\u8d77\u5c31\u662f\u4e00\u4e2a\u795e\u7ecf\u7f51\u7edc\u3002<\/p>\n $\"\u5c0f\u767d\u5b66CNN\u4ee5\u53caKeras\u7684\u901f\u6210\"$ \n
\u8fd9\u5c31\u662f\u7b80\u5355\u7684\u4e00\u5c42\u7f51\u7edc\uff0c\u4e5f\u53ef\u4ee5\u7531\u591a\u5c42\u7f51\u7edc<\/p>\n $\"\u5c0f\u767d\u5b66CNN\u4ee5\u53caKeras\u7684\u901f\u6210\"$ \n
\u8fd9\u91cc\u9762\u7684input layer\u5c31\u662f\u6240\u8c13\u7684\u5355\u4e2a\u8bad\u7ec3\u96c6\u7684\u7ef4\u6570\uff0c\u5c06\u6240\u6709\u7684\u8bad\u7ec3\u96c6\u8f93\u5165\u5c31\u53ef\u4ee5\u5f00\u59cb\u8bad\u7ec3\u4e00\u4e2a\u795e\u7ecf\u7f51\u7edc\u3002<\/p>\n
Keras\u5b9e\u73b0\u7b80\u5355\u7684\u795e\u7ecf\u7f51\u7edc<\/strong><\/p>\n
\u77e5\u9053\u4e86\u795e\u7ecf\u7f51\u7edc\u7684\u57fa\u672c\u7ed3\u6784\u548c\u539f\u7406\uff0c\u6211\u4eec\u5c31\u53ef\u4ee5\u5f00\u59cb\u4f7f\u7528keras\u53bb\u5b9e\u73b0\u4e00\u4e2a\u7b80\u5355\u7684\u795e\u7ecf\u7f51\u7edc\u3002<\/p>\n
import keras\n\nfrom keras.models import Sequential\n\nfrom keras.layers import Dense\n\nimport numpy as np<\/blockquote>\n
\u5bfc\u5165\u5fc5\u8981\u7684package<\/p>\n
x=np.array([[0,1,0],[0,0,1],[1,3,2],[3,2,1]])\n\ny=np.array([0,0,1,1]).T<\/blockquote>\n
\u8bbe\u5b9a\u8f93\u5165\u7684x\u548cy<\/p>\n
simple_model=Sequential()\n\nsimple_model.add(Dense(5,input_shape=(x.shape[1],),activation=’relu’,name=’layer1′))\n\nsimple_model.add(Dense(4,activation=’relu’,name=’layer2′))\n\nsimple_model.add(Dense(1,activation=’sigmoid’,name=’layer3′))<\/blockquote>\n
\u8f93\u5165\u4e00\u4e2a\u4e09\u5c42\u7684\u795e\u7ecf\u7f51\u7edc\uff0c\u4e2d\u95f4\u7684hidden layer\u7684\u5143\u7d20\u4e2a\u6570\u662f5\u548c4\uff0c\u6700\u540e\u4e00\u5c42\u8f93\u51fa\u4e00\u4e2a\u7ed3\u679c<\/p>\n
simple_model.compile(optimizer=’sgd’,loss=’mean_squared_error’)<\/blockquote>\n
complie\u8fd9\u4e2a\u7b80\u5355\u7684\u6a21\u578b<\/p>\n
simple_model.fit(x,y,epochs=20000)<\/blockquote>\n
\u8bad\u7ec320000\u6b21\u6a21\u578b<\/p>\n
simple_model.predict(x[0:1])<\/blockquote>\n
\u53ef\u4ee5\u9884\u6d4b\u4e00\u4e0b\u7b2c\u4e00\u4e2a\u8f93\u5165\u7684x\u7684\u7ed3\u679c\u4e0e\u5b9e\u9645\u7684\u662f\u5426\u76f8\u7b26\u3002<\/p>\n
\u4e0a\u9762\u5c31\u662f\u4e00\u4e2a\u7b80\u5355\u4e09\u5c42\u7f51\u7edc\u7684keras\u5b9e\u73b0\uff0c\u63a5\u4e0b\u6765\u6211\u4eec\u5c06\u6b63\u5f0f\u8fdb\u5165Convolutional Neural Network<\/p>\n
\u4e09\u3001Convolutional Neural Network<\/strong><\/h3>\n
\u524d\u9762\u7ed9\u5927\u5bb6\u63a8\u8350\u4e86\u4e00\u95e8\u597d\u8bfecs231n\uff0c\u672c\u7bc7\u6587\u7ae0\u4e5f\u662f\u6309\u7167\u8fd9\u4e2a\u601d\u8def\u6765\u7684\u3002<\/p>\n
\u57fa\u672c\u7ed3\u6784<\/strong><\/p>\n
\u9996\u5148\u89e3\u91ca\u4e00\u4e0b\u4ec0\u4e48\u662f\u5377\u79ef\uff0c\u8fd9\u4e2a\u5377\u79ef\u5f53\u7136\u4e0d\u662f\u6570\u5b66\u4e0a\u7684\u5377\u79ef\uff0c\u8fd9\u91cc\u7684\u5377\u79ef\u5176\u5b9e\u8868\u793a\u7684\u662f\u4e00\u4e2a\u4e09\u7ef4\u7684\u6743\u91cd\uff0c\u8fd9\u4e48\u89e3\u91ca\u8d77\u6765\u53ef\u80fd\u4e0d\u592a\u7406\u89e3\uff0c\u6211\u4eec\u5148\u770b\u770b\u5377\u79ef\u7f51\u7edc\u7684\u57fa\u672c\u7ed3\u6784\u3002<\/p>\n $\"\u5c0f\u767d\u5b66CNN\u4ee5\u53caKeras\u7684\u901f\u6210\"$ \n
\u901a\u8fc7\u4e0a\u9762\u7684\u56fe\u6211\u4eec\u6e05\u695a\u5730\u4e86\u89e3\u5230\u5377\u79ef\u7f51\u7edc\u548c\u4e00\u822c\u7f51\u7edc\u7ed3\u6784\u4e0a\u9762\u7684\u5dee\u522b\uff0c\u4e5f\u53ef\u4ee5\u7406\u89e3\u4e3a\u5377\u79ef\u7f51\u7edc\u662f\u7acb\u4f53\u7684\uff0c\u800c\u4e00\u822c\u7684\u7f51\u7edc\u7ed3\u6784\u662f\u5e73\u9762\u7684\u3002<\/p>\n
\u5377\u79ef\u5c42<\/strong><\/p>\n
\u4e86\u89e3\u5b8c\u4e86\u57fa\u672c\u7684\u7ed3\u6784\u4e4b\u540e\uff0c\u6211\u4eec\u5c31\u8981\u4e86\u89e3cnn\u6700\u91cd\u8981\u7684\u4e00\u4e2a\u90e8\u5206\uff0c\u4e5f\u662f\u6700\u4e3a\u521b\u65b0\u7684\u4e00\u4e2a\u90e8\u5206\uff0c\u5377\u79ef\u5c42\u3002\u9996\u5148\u7528\u4e00\u5f20\u56fe\u7247\u6765\u6bd4\u8f83\u4e00\u4e0b\u5377\u79ef\u7f51\u7edc\u5230\u5e95\u521b\u65b0\u5728\u4ec0\u4e48\u5730\u65b9\u3002<\/p>\n $\"\u5c0f\u767d\u5b66CNN\u4ee5\u53caKeras\u7684\u901f\u6210\"$ \n
\u6211\u4eec\u901a\u8fc7\u8fd9\u4e2a\u7ed3\u6784\u5c31\u53ef\u4ee5\u6e05\u6670\u5730\u770b\u5230\u5377\u79ef\u7f51\u7edc\u5230\u5e95\u662f\u600e\u4e48\u5b9e\u73b0\u7684\u3002\u9996\u5148\u53f3\u8fb9\u662f\u4f20\u7edf\u7684\u7f51\u7edc\u7ed3\u6784\uff0c\u5728\u524d\u9762\u6211\u4eec\u5df2\u7ecf\u8be6\u7ec6\u7684\u89e3\u91ca\u8fc7\u4e86\u3002\u800c\u5de6\u8fb9\u7684\u56fe\u7247\uff0c\u6211\u4eec\u9996\u5148\u770b\u770b\u56fe\u4e2d\u6700\u5de6\u8fb9\u7684\u7ed3\u6784\uff0c\u4f60\u80af\u5b9a\u4f1a\u597d\u5947\u4e3a\u4ec0\u4e48\u662f32x32x3\u7684\u4e00\u5757\u7acb\u4f53\u65b9\u5757\u3002\u8fd9\u4e2a32×32\u4ee3\u8868\u7684\u662f\u50cf\u7d20\u70b9\uff0c\u8bf4\u767d\u4e86\u4e5f\u5c31\u662f\u56fe\u7247\u7684\u5927\u5c0f\uff0c\u8fd9\u4e2a\u5927\u5c0f\u662f\u4f60\u53ef\u4ee5\u8bbe\u7f6e\u7684\uff0c\u4f60\u53ef\u4ee5\u8bbe\u7f6e\u4e3a50×50\uff0c\u4e5f\u53ef\u4ee5\u662f256×256\uff0c\u8fd9\u90fd\u53d6\u51b3\u4e0e\u56fe\u7247\u7684\u5927\u5c0f\uff0c\u90a3\u4e483\u8868\u793a\u4ec0\u4e48\u5462\uff1f3\u5176\u5b9e\u8868\u793a\u7684\u662fRGB\u7684\u4e09\u4e2a\u901a\u9053\uff0cRGB\u4e5f\u662f\u4ec0\u4e48\uff1fRGB\u8868\u793ared\uff0cgreen\uff0cblue\uff0c\u8fd9\u4e09\u79cd\u989c\u8272\u7684\u5404\u79cd\u7ec4\u5408\u53e0\u52a0\u53ef\u4ee5\u5f62\u6210\u5404\u79cd\u5404\u6837\u7684\u989c\u8272\uff0c\u6240\u4ee5\u4efb\u4f55\u4e00\u5f20\u7167\u7247\u90fd\u53ef\u4ee5\u7528\u5de6\u8fb9\u8fd9\u79cd\u56fe\u5f62\u6765\u8868\u793a\u3002<\/p>\n
\u90a3\u4e48\u4e2d\u95f4\u8fd9\u4e2a\u5c0f\u65b9\u5757\u53c8\u8868\u793a\u4ec0\u4e48\u5462\uff1f\u8fd9\u4e2a\u5c31\u662f\u6211\u4eec\u8981\u91cd\u70b9\u8bb2\u7684\u5377\u79ef\u3002\u6240\u8c13\u7684\u5377\u79ef\uff0c\u5c31\u662f\u8fd9\u79cd\u5c0f\u65b9\u5757\uff0c\u6211\u4eec\u8bbe\u7f6e\u4e00\u4e2a\u5c0f\u65b9\u5757\u7684\u5927\u5c0f\uff0c\u4f46\u662f\u8fd9\u4e2a\u5c0f\u65b9\u5757\u7684\u539a\u5ea6\u5fc5\u987b\u548c\u5de6\u8fb9\u7684\u8fd9\u4e2a\u5927\u65b9\u5757\u7684\u539a\u5ea6\u662f\u4e00\u6837\u7684\uff0c\u5927\u65b9\u5757\u6bcf\u4e00\u4e2a\u50cf\u7d20\u70b9\u7531\u4e00\u4e2a0\u5230255\u7684\u6570\u5b57\u8868\u793a\uff0c\u8fd9\u6837\u6211\u4eec\u5c31\u53ef\u4ee5\u8d4b\u4e88\u5c0f\u65b9\u5757\u6743\u91cd\uff0c\u6bd4\u5982\u6211\u4eec\u53d6\u5c0f\u65b9\u5757\u7684\u5927\u5c0f\u662f3×3\uff0c\u6211\u4eec\u8981\u6c42\u539a\u5ea6\u5fc5\u987b\u8981\u548c\u5de6\u8fb9\u7684\u5927\u65b9\u5757\u539a\u5ea6\u4e00\u6837\uff0c\u90a3\u4e48\u5c0f\u65b9\u5757\u7684\u7684\u5927\u5c0f\u5c31\u4e3a3x3x3\uff0c\u6211\u4eec\u5c31\u53ef\u4ee5\u8d4b\u4e88\u51763x3x3\u4e2a\u6743\u91cd\uff0c\u7136\u540e\u6211\u4eec\u5c31\u53ef\u4ee5\u5f00\u59cb\u8ba1\u7b97\u5377\u79ef\u7684\u7ed3\u679c\uff0c\u5c06\u5c0f\u65b9\u5757\u4ece\u5927\u65b9\u5757\u7684\u5de6\u4e0a\u89d2\u5f00\u59cb\uff0c\u4e00\u4e2a\u5377\u79ef\u5c0f\u65b9\u5757\u6240\u8986\u76d6\u7684\u8303\u56f4\u662f3x3x3\uff0c\u7136\u540e\u6211\u4eec\u5c06\u5927\u65b9\u5757\u4e2d3x3x3\u7684\u6570\u5b57\u548c\u5c0f\u65b9\u5757\u4e2d\u7684\u6743\u91cd\u5206\u522b\u76f8\u4e58\u76f8\u52a0\uff0c\u518d\u52a0\u4e0a\u4e00\u4e2a\u504f\u5dee\uff0c\u5c31\u53ef\u4ee5\u5f97\u5230\u4e00\u4e2a\u5377\u79ef\u7684\u7ed3\u679c\uff0c\u53ef\u4ee5\u62bd\u8c61\u7684\u5199\u6210Wx+b\u8fd9\u79cd\u5f62\u5f0f\uff0c\u8fd9\u5c31\u662f\u56fe\u4e0a\u6240\u663e\u793a\u7684\u7ed3\u679c\uff0c\u7136\u540e\u6211\u4eec\u53ef\u4ee5\u8bbe\u7f6e\u5c0f\u65b9\u5757\u7684\u6ed1\u52a8\u8ddd\u79bb\uff0c\u6bcf\u6b21\u6ed1\u52a8\u5c31\u53ef\u4ee5\u5f62\u6210\u4e00\u4e2a\u5377\u79ef\u7684\u8ba1\u7b97\u7ed3\u679c\uff0c\u7136\u540e\u5c06\u6574\u5f20\u5927\u56fe\u7247\u6ed1\u52a8\u8986\u76d6\u4e4b\u540e\u5c31\u53ef\u4ee5\u5f62\u6210\u4e00\u5c42\u5377\u79ef\u7684\u7ed3\u679c\uff0c\u6211\u4eec\u770b\u5230\u56fe\u4e2d\u7684\u5377\u79ef\u7ed3\u679c\u662f\u5f88\u539a\u7684\uff0c\u4e5f\u5c31\u662f\u8bbe\u7f6e\u4e86\u5f88\u591a\u5c42\u5377\u79ef\u3002\u603b\u7ed3\u6765\u8bf4\uff0c\u6bcf\u5c42\u5377\u79ef\u5c31\u662f\u4e00\u4e2a\u5377\u79ef\u6838\u5728\u56fe\u7247\u4e0a\u6ed1\u52a8\u6c42\u503c\uff0c\u7136\u540e\u8bbe\u7f6e\u591a\u4e2a\u5377\u79ef\u6838\u5c31\u53ef\u4ee5\u5f62\u6210\u591a\u5c42\u7684\u5377\u79ef\u5c42\u3002<\/p>\n
\u6c60\u5316\u5c42<\/strong><\/p>\n
\u8bb2\u5b8c\u5377\u79ef\u5c42\uff0c\u63a5\u4e0b\u6765\u5c31\u8981\u8bb2\u4e00\u4e0b\u6c60\u5316\u5c42\u3002\u4e3a\u4ec0\u4e48\u4f1a\u6709\u6c60\u5316\u5c42\u7684\u51fa\u73b0\u5462\uff1f\u662f\u56e0\u4e3a\u4e0d\u65ad\u7684\u505a\u5377\u79ef\uff0c\u5f97\u5230\u7684\u4e2d\u95f4\u7ed3\u679c\u4f1a\u8d8a\u6765\u8d8a\u539a\uff0c\u5377\u79ef\u5c31\u76f8\u5f53\u4e8e\u63d0\u53d6\u56fe\u7247\u4e2d\u7684\u7279\u5f81\uff0c\u6240\u4ee5\u5377\u79ef\u5c42\u4e00\u822c\u4f1a\u8bbe\u7f6e\u5f97\u8d8a\u6765\u8d8a\u539a\uff0c\u4e0d\u7136\u4f60\u5c31\u65e0\u6cd5\u4ece\u524d\u9762\u7684\u7ed3\u679c\u6765\u63d0\u53d6\u66f4\u591a\u7684\u7279\u5f81\u3002\u8fd9\u6837\u5c31\u4f1a\u5bfc\u81f4\u4e2d\u95f4\u7684\u7ed3\u679c\u4f1a\u8d8a\u6765\u8d8a\u5927\uff0c\u8ba1\u7b97\u4f1a\u8d8a\u6765\u8d8a\u6162\uff0c\u6240\u4ee5\u63d0\u51fa\u4e86\u6c60\u5316\u5c42\u3002<\/p>\n
\u6240\u8c13\u7684\u6c60\u5316\u5c42\uff0c\u5c31\u662f\u5c06\u56fe\u7247\u7684\u5927\u5c0f\u7f29\u5c0f\u7684\u4e00\u79cd\u5904\u7406\u65b9\u5f0f\u3002\u6211\u4eec\u53ef\u4ee5\u5148\u770b\u770b\u4e0b\u9762\u7684\u56fe\u7247\u3002<\/p>\n $\"\u5c0f\u767d\u5b66CNN\u4ee5\u53caKeras\u7684\u901f\u6210\"$ \n
\u901a\u8fc7\u8fd9\u4e2a\u56fe\u7247\uff0c\u6211\u4eec\u53ef\u4ee5\u6e05\u695a\u5730\u770b\u5230\u6c60\u5316\u5c42\u662f\u600e\u4e48\u5904\u7406\u7684\u3002\u6c60\u5316\u5c42\u4e5f\u662f\u9700\u8981\u5148\u8bbe\u7f6e\u4e00\u4e2a\u7a97\u53e3\uff0c\u4f46\u662f\u8fd9\u4e2a\u5c0f\u7a97\u53e3\u7684\u539a\u5ea6\u662f1\uff0c\u800c\u4e0d\u518d\u662f\u524d\u4e00\u5c42\u8f93\u51fa\u7684\u7ed3\u679c\u7684\u539a\u5ea6\u3002\u7136\u540e\u6709\u4e24\u79cd\u5904\u7406\u65b9\u5f0f\uff0c\u4e00\u79cd\u662f\u53d6\u8fd9\u4e2a\u5c0f\u7a97\u53e3\u91cc\u9762\u6240\u6709\u5143\u7d20\u7684\u6700\u5927\u503c\u6765\u4ee3\u8868\u8fd9\u4e2a\u5c0f\u7a97\u53e3\uff0c\u4e00\u79cd\u662f\u53d6\u5e73\u5747\u503c\uff0c\u7136\u540e\u5c06\u5c0f\u7a97\u53e3\u6ed1\u52a8\uff0c\u5728\u7b2c\u4e8c\u7684\u4f4d\u7f6e\u518d\u505a\u540c\u6837\u7684\u5904\u7406\uff0c\u4e0a\u5c42\u7f51\u7edc\u8f93\u51fa\u65b9\u5757\u7684\u6bcf\u4e00\u5c42\u505a\u5b8c\u4e4b\u540e\u5c31\u8fdb\u5165\u8fd9\u4e2a\u5927\u65b9\u5757\u7684\u4e0b\u4e00\u5c42\u505a\u540c\u6837\u7684\u64cd\u4f5c\uff0c\u8fd9\u4e2a\u5904\u7406\u529e\u6cd5\u5c31\u53ef\u4ee5\u8ba9\u6574\u4e2a\u5927\u65b9\u5757\u7684\u5927\u5c0f\u53d8\u5c0f\uff0c\u53ef\u4ee5\u770b\u770b\u4e0a\u9762\u7684\u56fe\u7247\u7684\u5de6\u8fb9\u3002\u53f3\u8fb9\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u4e00\u5c42\u539a\u5ea6\uff0c\u53d6\u6700\u5927\u503c\u7684\u4f8b\u5b50\u3002<\/p>\n
\u5b9e\u73b0Lenet<\/strong><\/p>\n
\u8bb2\u5b8c\u4e86\u5377\u79ef\u7f51\u7edc\u7684\u57fa\u672c\u7ed3\u6784\u4e4b\u540e\uff0c\u4f60\u662f\u4e0d\u662f\u5df2\u7ecf\u8feb\u4e0d\u53ca\u5f85\u5e0c\u671b\u80fd\u591f\u5b9e\u73b0\u4e00\u4e2a\u7b80\u5355\u7684\u795e\u7ecf\u7f51\u7edc\u4e86\u5462\uff1f\u5377\u79ef\u7f51\u7edc\u53d1\u5c55\u7684\u7279\u522b\u8fc5\u901f\uff0c\u6700\u65e9\u662f\u7531Lecun\u63d0\u51fa\u6765\u7684\uff0cLenet\u6210\u4e3acnn\u7684\u9f3b\u7956\uff0c\u63a5\u4e0b\u6765\u4ed6\u7684\u5b66\u751fAlex\u63d0\u51fa\u4e86\u5c42\u6570\u66f4\u6df1\u7684Alexnet\uff0c\u7136\u540e2013\u5e74\u53c8\u63d0\u51fa\u4e86VGGnet\uff0c\u670916\u5c42\u548c19\u5c42\u4e24\u79cd\uff0c\u8fd9\u4e9b\u90fd\u53ea\u662f\u5728\u5c42\u6570\u4e0a\u9762\u7684\u52a0\u6df1\uff0c\u5e76\u6ca1\u6709\u4ec0\u4e48\u5176\u4ed6\u7684\u521b\u65b0\uff0c\u800c\u4e4b\u540egoogle\u63d0\u51fa\u4e86inception net\u5728\u7f51\u7edc\u7ed3\u6784\u4e0a\u5b9e\u73b0\u4e86\u521b\u65b0\uff0c\u63d0\u51fa\u4e86\u4e00\u79cdinception\u7684\u673a\u6784\uff0cfacebook ai \u5b9e\u9a8c\u5ba4\u53c8\u63d0\u51fa\u4e86resnet\uff0c\u6b8b\u5dee\u7f51\u7edc\uff0c\u5b9e\u73b0\u4e86150\u5c42\u7684\u7f51\u7edc\u7ed3\u6784\u53ef\u8bad\u7ec3\u5316<\/strong>\uff0c\u8fd9\u4e9b\u6211\u4eec\u4e4b\u540e\u4f1a\u6162\u6162\u8bb2\u5230\u3002<\/p>\n
\u63a5\u4e0b\u6765\u6211\u4eec\u5c31\u6765\u5b9e\u73b0\u4e00\u4e0b\u6700\u7b80\u5355\u7684Lenet\uff0c\u4f7f\u7528mnist\u624b\u5199\u5b50\u4f53\u4f5c\u4e3a\u8bad\u7ec3\u96c6\u3002<\/p>\n
import keras\n\nfrom keras.datasets import mnist\n\n(x_train, y_train), (x_test,y_test) =mnist.load_data()<\/blockquote>\n
\u5bfc\u5165\u5fc5\u8981\u7684\u5e93\u548c\u6570\u636e\u96c6<\/p>\n
x_train=x_train.reshape(-1,28,28,1)\n\nx_test=x_test.reshape(-1,28,28,1)\n\nx_train=x_train\/255.\n\nx_test=x_test\/255.\n\ny_train=keras.utils.to_categorical(y_train)\n\ny_test=keras.utils.to_categorical(y_test)<\/blockquote>\n
\u5904\u7406\u6570\u636e\uff0c\u8ba9\u6570\u636e\u7684shape\u662f(28, 28, 1)\uff0c\u7136\u540elabel\u505a\u4e00\u4e2aone-hot encoding\u5904\u7406\uff0c\u6bd4\u5982\u7c7b\u522b\u662f3\uff0c\u90a3\u4e48\u53d8\u6210[0, 0, 1 ,0, 0, 0, 0, 0, 0, 0]\u3002<\/p>\n
from keras.layers import Conv2D,MaxPool2D,Dense,Flatten\n\nfrom keras.models import Sequential\n\nlenet=Sequential()\n\nlenet.add(Conv2D(6,kernel_size=3,strides=1,padding=’same’,input_shape=(28, 28, 1)))\n\nlenet.add(MaxPool2D(pool_size=2,strides=2))\n\nlenet.add(Conv2D(16,kernel_size=5,strides=1,padding=’valid’))\n\nlenet.add(MaxPool2D(pool_size=2,strides=2))\n\nlenet.add(Flatten())\n\nlenet.add(Dense(120))\n\nlenet.add(Dense(84))\n\nlenet.add(Dense(10,activation=’softmax’))<\/blockquote>\n
\u6784\u5efalenet<\/strong><\/p>\n $\"\u5c0f\u767d\u5b66CNN\u4ee5\u53caKeras\u7684\u901f\u6210\"$ \n
lenet.compile(‘sgd’,loss=’categorical_crossentropy’,metrics=[‘accuracy’])<\/blockquote>\n
\u7f16\u8bd1<\/p>\n
lenet.fit(x_train,y_train,batch_size=64,epochs=50,validation_data=[x_test,y_test])<\/blockquote>\n
\u8bad\u7ec350\u6b21\uff0c\u5f97\u5230\u7ed3\u679c\u5982\u4e0b<\/p>\n $\"\u5c0f\u767d\u5b66CNN\u4ee5\u53caKeras\u7684\u901f\u6210\"$ \n
lenet.save(‘myletnet.h5’)<\/blockquote>\n
\u53ef\u4ee5\u4fdd\u5b58\u8bad\u7ec3\u597d\u7684\u6a21\u578b<\/p>\n
\u603b\u7ed3<\/strong><\/p>\n
OK\uff0c \u8fd9\u5c31\u662f\u6211\u4eec\u5199\u7684\u4e00\u4e2a\u8d85\u7ea7\u7b80\u5355\u7684Lenet\uff0c\u8bad\u7ec350\u6b21\u5f97\u5230\u7684\u8bad\u7ec3\u51c6\u786e\u7387\u5df2\u7ecf\u8fbe\u52300.9939\uff0c\u6d4b\u8bd5\u51c6\u786e\u7387\u8fbe\u52300.9852\u3002<\/p>\n
\u6b22\u8fce\u5173\u6ce8\u6211\u7684\u77e5\u4e4e\u4e13\u680f\uff0c\u6df1\u5ea6\u70bc\u4e39<\/a><\/p>\n
\u6b22\u8fce\u5173\u6ce8\u6211\u7684 github<\/a>\u4e3b\u9875<\/p>\n
\u6b22\u8fce\u8bbf\u95ee\u6211\u7684 \u535a\u5ba2<\/a><\/p>\n
\u8fd9\u7bc7\u6587\u7ae0\u7684\u4ee3\u7801\u90fd\u5df2\u4f20\u5230\u4e86github\u4e0a\n SherlockLiao\/lenet<\/a><\/p>","protected":false},"excerpt":{"rendered":"
\u4e00\u3001\u4e3a\u4f55\u8981\u7528Keras \u5982\u4eca\u5728\u6df1\u5ea6\u5b66\u4e60\u5927\u706b\u7684\u65f6\u5019\uff0c\u7b2c\u4e09\u65b9\u5de5\u5177\u4e5f\u5c42\u51fa\u4e0d\u7a77\uff0c\u6bd4\u8f83\u51fa\u540d\u7684\u6709Tensorflow\uff0cCaffe\uff0cTheano\uff0cMXNet\uff0c\u5728\u5982\u6b64\u591a\u7684\u7b2c\u4e09\u65b9\u6846\u67b6\u4e2d\u9891\u7e41\u7684\u66f4\u6362\u65e0\u7591\u662f\u5f88\u4f4e\u6548\u7684\uff0c\u53ea\u8981\u4f60\u80fd...<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[274],"tags":[501,500],"class_list":["post-1850","post","type-post","status-publish","format-standard","hentry","category-python","tag-cnn","tag-keras"],"_links":{"self":[{"href":"http:\/\/nick.txtcc.com\/index.php\/wp-json\/wp\/v2\/posts\/1850","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/nick.txtcc.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/nick.txtcc.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/nick.txtcc.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/nick.txtcc.com\/index.php\/wp-json\/wp\/v2\/comments?post=1850"}],"version-history":[{"count":10,"href":"http:\/\/nick.txtcc.com\/index.php\/wp-json\/wp\/v2\/posts\/1850\/revisions"}],"predecessor-version":[{"id":1939,"href":"http:\/\/nick.txtcc.com\/index.php\/wp-json\/wp\/v2\/posts\/1850\/revisions\/1939"}],"wp:attachment":[{"href":"http:\/\/nick.txtcc.com\/index.php\/wp-json\/wp\/v2\/media?parent=1850"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/nick.txtcc.com\/index.php\/wp-json\/wp\/v2\/categories?post=1850"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/nick.txtcc.com\/index.php\/wp-json\/wp\/v2\/tags?post=1850"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}