import numpy as np<\/p>\ndef add_noise(data, noise_factor=0.5):<\/p>\n
noise = np.random.randn(*data.shape) * noise_factor<\/p>\n
return data + noise<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u4e8c\u3001\u5408\u6210\u5c11\u6570\u8fc7\u91c7\u6837\u6280\u672f\uff08SMOTE\uff09<\/p>\n<\/p>\n
SMOTE\u662f\u4e00\u79cd\u7528\u4e8e\u5904\u7406\u4e0d\u5e73\u8861\u6570\u636e\u96c6\u7684\u6280\u672f\u3002\u5b83\u901a\u8fc7\u5728\u7279\u5f81\u7a7a\u95f4\u4e2d\u5408\u6210\u65b0\u7684\u5c11\u6570\u7c7b\u6837\u672c\uff0c\u6765\u5e73\u8861\u6837\u672c\u5206\u5e03\u3002Python\u4e2d\u7684imbalanced-learn<\/code>\u5e93\u63d0\u4f9b\u4e86SMOTE\u7684\u5b9e\u73b0\u3002<\/p>\n<\/p>\n\n- SMOTE\u7684\u57fa\u672c\u539f\u7406<\/li>\n<\/ol>\n
SMOTE\u901a\u8fc7\u9009\u62e9\u5c11\u6570\u7c7b\u6837\u672c\uff0c\u5e76\u5728\u8fd9\u4e9b\u6837\u672c\u7684\u7279\u5f81\u7a7a\u95f4\u4e2d\u63d2\u503c\u751f\u6210\u65b0\u7684\u6837\u672c\uff0c\u589e\u52a0\u5c11\u6570\u7c7b\u6837\u672c\u7684\u6570\u91cf\u3002\u8fd9\u6837\u80fd\u591f\u6709\u6548\u5730\u5e73\u8861\u6570\u636e\u96c6\uff0c\u63d0\u5347\u5206\u7c7b\u5668\u5bf9\u5c11\u6570\u7c7b\u7684\u8bc6\u522b\u80fd\u529b\u3002<\/p>\n<\/p>\n
from imblearn.over_sampling import SMOTE<\/p>\n\u521b\u5efaSMOTE\u5bf9\u8c61<\/strong><\/h2>\nsmote = SMOTE(random_state=42)<\/p>\n
X_resampled, y_resampled = smote.fit_resample(X, y)<\/p>\n
<\/code><\/pre>\n<\/p>\n
\n- \u9002\u7528\u573a\u666f<\/li>\n<\/ol>\n
SMOTE\u7279\u522b\u9002\u7528\u4e8e\u5206\u7c7b\u95ee\u9898\u4e2d\u7684\u4e0d\u5e73\u8861\u6570\u636e\u96c6\uff0c\u4f8b\u5982\u5728\u533b\u7597\u8bca\u65ad\u4e2d\uff0c\u75c5\u60a3\u6837\u672c\u901a\u5e38\u8f83\u5c11\uff0c\u901a\u8fc7SMOTE\u53ef\u4ee5\u751f\u6210\u66f4\u591a\u7684\u75c5\u60a3\u6837\u672c\uff0c\u63d0\u5347\u8bca\u65ad\u6a21\u578b\u7684\u51c6\u786e\u6027\u3002<\/p>\n<\/p>\n
\u4e09\u3001\u6570\u636e\u751f\u6210\u5bf9\u6297\u7f51\u7edc\uff08GAN\uff09<\/p>\n<\/p>\n
GAN\u662f\u4e00\u79cd\u901a\u8fc7\u5bf9\u6297\u8bad\u7ec3\u751f\u6210\u65b0\u6837\u672c\u7684\u6280\u672f\u3002\u5b83\u7531\u751f\u6210\u5668\u548c\u5224\u522b\u5668\u4e24\u4e2a\u7f51\u7edc\u7ec4\u6210\uff0c\u5176\u4e2d\u751f\u6210\u5668\u8d1f\u8d23\u751f\u6210\u65b0\u6837\u672c\uff0c\u5224\u522b\u5668\u8d1f\u8d23\u533a\u5206\u771f\u5b9e\u6837\u672c\u548c\u751f\u6210\u6837\u672c\u3002<\/p>\n<\/p>\n
\n- GAN\u7684\u5de5\u4f5c\u539f\u7406<\/li>\n<\/ol>\n
GAN\u7684\u8bad\u7ec3\u8fc7\u7a0b\u662f\u751f\u6210\u5668\u548c\u5224\u522b\u5668\u4e4b\u95f4\u7684\u5bf9\u6297\u535a\u5f08\u3002\u751f\u6210\u5668\u8bd5\u56fe\u751f\u6210\u903c\u771f\u7684\u6837\u672c\u4ee5\u6b3a\u9a97\u5224\u522b\u5668\uff0c\u800c\u5224\u522b\u5668\u5219\u4e0d\u65ad\u63d0\u9ad8\u533a\u5206\u771f\u5b9e\u6837\u672c\u548c\u751f\u6210\u6837\u672c\u7684\u80fd\u529b\u3002\u6700\u7ec8\uff0c\u5f53\u751f\u6210\u5668\u751f\u6210\u7684\u6837\u672c\u8db3\u591f\u903c\u771f\uff0c\u4ee5\u81f3\u4e8e\u5224\u522b\u5668\u65e0\u6cd5\u533a\u5206\u65f6\uff0c\u8bad\u7ec3\u7ed3\u675f\u3002<\/p>\n<\/p>\n
import tensorflow as tf<\/p>\ndef build_generator():<\/p>\n
model = tf.keras.Sequential()<\/p>\n
model.add(tf.keras.layers.Dense(256, input_dim=100))<\/p>\n
model.add(tf.keras.layers.LeakyReLU(alpha=0.2))<\/p>\n
model.add(tf.keras.layers.BatchNormalization(momentum=0.8))<\/p>\n
model.add(tf.keras.layers.Dense(512))<\/p>\n
model.add(tf.keras.layers.LeakyReLU(alpha=0.2))<\/p>\n
model.add(tf.keras.layers.BatchNormalization(momentum=0.8))<\/p>\n
model.add(tf.keras.layers.Dense(1024))<\/p>\n
model.add(tf.keras.layers.LeakyReLU(alpha=0.2))<\/p>\n
model.add(tf.keras.layers.BatchNormalization(momentum=0.8))<\/p>\n
model.add(tf.keras.layers.Dense(28*28*1, activation='tanh'))<\/p>\n
model.add(tf.keras.layers.Reshape((28, 28, 1)))<\/p>\n
return model<\/p>\n
generator = build_generator()<\/p>\n
<\/code><\/pre>\n<\/p>\n
\n- GAN\u7684\u5e94\u7528<\/li>\n<\/ol>\n
GAN\u5728\u751f\u6210\u56fe\u50cf\u3001\u6587\u672c\u751f\u6210\u3001\u6570\u636e\u8865\u5168\u7b49\u9886\u57df\u6709\u5e7f\u6cdb\u5e94\u7528\u3002\u4f8b\u5982\uff0c\u5728\u533b\u5b66\u56fe\u50cf\u5904\u7406\u4e2d\uff0cGAN\u53ef\u4ee5\u7528\u4e8e\u751f\u6210\u7f3a\u5931\u6216\u4e0d\u5b8c\u6574\u7684\u56fe\u50cf\u6570\u636e\uff0c\u63d0\u9ad8\u6a21\u578b\u7684\u8bad\u7ec3\u6548\u679c\u3002<\/p>\n<\/p>\n
\u56db\u3001\u81ea\u52a8\u7f16\u7801\u5668\uff08Autoencoder\uff09\u7528\u4e8e\u6837\u672c\u6269\u5145<\/p>\n<\/p>\n
\u81ea\u52a8\u7f16\u7801\u5668\u662f\u4e00\u79cd\u65e0\u76d1\u7763\u5b66\u4e60\u6a21\u578b\uff0c\u53ef\u4ee5\u7528\u4e8e\u964d\u7ef4\u3001\u7279\u5f81\u5b66\u4e60\u548c\u6570\u636e\u751f\u6210\u3002\u5b83\u901a\u8fc7\u5c06\u8f93\u5165\u6570\u636e\u538b\u7f29\u5230\u4e00\u4e2a\u9690\u7a7a\u95f4\uff0c\u5e76\u4ece\u4e2d\u91cd\u5efa\u8f93\u5165\u6570\u636e\uff0c\u4ece\u800c\u53ef\u4ee5\u7528\u4e8e\u6837\u672c\u6269\u5145\u3002<\/p>\n<\/p>\n
\n- \u81ea\u52a8\u7f16\u7801\u5668\u7684\u57fa\u672c\u7ed3\u6784<\/li>\n<\/ol>\n
\u81ea\u52a8\u7f16\u7801\u5668\u7531\u7f16\u7801\u5668\u548c\u89e3\u7801\u5668\u4e24\u90e8\u5206\u7ec4\u6210\u3002\u7f16\u7801\u5668\u5c06\u8f93\u5165\u6570\u636e\u538b\u7f29\u4e3a\u4f4e\u7ef4\u8868\u793a\uff0c\u800c\u89e3\u7801\u5668\u5219\u4ece\u4f4e\u7ef4\u8868\u793a\u91cd\u5efa\u8f93\u5165\u6570\u636e\u3002\u901a\u8fc7\u5728\u4e2d\u95f4\u9690\u7a7a\u95f4\u4e2d\u8fdb\u884c\u91c7\u6837\uff0c\u53ef\u4ee5\u751f\u6210\u65b0\u7684\u6570\u636e\u6837\u672c\u3002<\/p>\n<\/p>\n
from keras.layers import Input, Dense<\/p>\nfrom keras.models import Model<\/p>\n
input_img = Input(shape=(784,))<\/p>\n
encoded = Dense(128, activation='relu')(input_img)<\/p>\n
decoded = Dense(784, activation='sigmoid')(encoded)<\/p>\n
autoencoder = Model(input_img, decoded)<\/p>\n
<\/code><\/pre>\n<\/p>\n
\n- \u5e94\u7528\u573a\u666f<\/li>\n<\/ol>\n
\u81ea\u52a8\u7f16\u7801\u5668\u53ef\u4ee5\u7528\u4e8e\u56fe\u50cf\u3001\u6587\u672c\u3001\u97f3\u9891\u7b49\u591a\u79cd\u6570\u636e\u7c7b\u578b\u7684\u6269\u5145\u3002\u7279\u522b\u662f\u5728\u7f3a\u4e4f\u5927\u91cf\u6807\u6ce8\u6570\u636e\u7684\u60c5\u51b5\u4e0b\uff0c\u81ea\u52a8\u7f16\u7801\u5668\u80fd\u591f\u751f\u6210\u5177\u6709\u591a\u6837\u6027\u7684\u65b0\u6837\u672c\uff0c\u63d0\u5347\u6a21\u578b\u7684\u6027\u80fd\u3002<\/p>\n<\/p>\n
\u4e94\u3001\u7ed3\u8bba<\/p>\n<\/p>\n
\u5728Python\u4e2d\uff0c\u6269\u5145\u6837\u672c\u7684\u65b9\u6cd5\u591a\u79cd\u591a\u6837\uff0c\u9009\u62e9\u5408\u9002\u7684\u65b9\u6cd5\u53d6\u51b3\u4e8e\u6570\u636e\u7684\u7c7b\u578b\u548c\u5177\u4f53\u7684\u5e94\u7528\u573a\u666f\u3002\u901a\u8fc7\u6570\u636e\u589e\u5f3a\u3001SMOTE\u3001GAN\u548c\u81ea\u52a8\u7f16\u7801\u5668\u7b49\u6280\u672f<\/strong>\uff0c\u53ef\u4ee5\u6709\u6548\u5730\u589e\u52a0\u6570\u636e\u96c6\u7684\u591a\u6837\u6027\uff0c\u63d0\u9ad8\u6a21\u578b\u7684\u8bad\u7ec3\u6548\u679c\u548c\u6cdb\u5316\u80fd\u529b\u3002\u5728\u5b9e\u9645\u5e94\u7528\u4e2d\uff0c\u9700\u8981\u6839\u636e\u6570\u636e\u7279\u70b9\u548c\u8ba1\u7b97\u8d44\u6e90\uff0c\u5408\u7406\u9009\u62e9\u548c\u7ec4\u5408\u8fd9\u4e9b\u65b9\u6cd5\uff0c\u4ee5\u8fbe\u5230\u6700\u4f73\u7684\u6837\u672c\u6269\u5145\u6548\u679c\u3002<\/p>\n<\/p>\n\u76f8\u5173\u95ee\u7b54FAQs\uff1a<\/strong><\/h2>\n \u5982\u4f55\u4f7f\u7528Python\u751f\u6210\u5408\u6210\u6570\u636e\u4ee5\u6269\u5145\u6837\u672c\uff1f<\/strong>
\u5728Python\u4e2d\uff0c\u53ef\u4ee5\u4f7f\u7528\u591a\u79cd\u65b9\u6cd5\u751f\u6210\u5408\u6210\u6570\u636e\u6765\u6269\u5145\u6837\u672c\u3002\u5e38\u89c1\u7684\u65b9\u6cd5\u5305\u62ec\u4f7f\u7528scikit-learn<\/code>\u5e93\u4e2d\u7684make_classification<\/code>\u51fd\u6570\uff0c\u6216\u901a\u8fc7numpy<\/code>\u548cpandas<\/code>\u5e93\u521b\u5efa\u968f\u673a\u6570\u636e\u3002\u6b64\u5916\uff0c\u6df1\u5ea6\u5b66\u4e60\u6846\u67b6\u5982TensorFlow<\/code>\u548cPyTorch<\/code>\u4e5f\u652f\u6301\u6570\u636e\u589e\u5f3a\u6280\u672f\uff0c\u53ef\u4ee5\u751f\u6210\u65b0\u7684\u6837\u672c\u3002\u8fd9\u4e9b\u65b9\u6cd5\u4f7f\u5f97\u6570\u636e\u96c6\u66f4\u52a0\u4e30\u5bcc\uff0c\u63d0\u9ad8\u6a21\u578b\u7684\u6cdb\u5316\u80fd\u529b\u3002<\/p>\n\u5728\u6269\u5145\u6837\u672c\u65f6\uff0c\u6709\u54ea\u4e9b\u5e38\u89c1\u7684\u6570\u636e\u589e\u5f3a\u6280\u672f\uff1f<\/strong>
\u6570\u636e\u589e\u5f3a\u6280\u672f\u6709\u5f88\u591a\u79cd\uff0c\u5e38\u89c1\u7684\u5305\u62ec\u65cb\u8f6c\u3001\u7ffb\u8f6c\u3001\u7f29\u653e\u3001\u88c1\u526a\u4ee5\u53ca\u989c\u8272\u8c03\u6574\u7b49\u3002\u8fd9\u4e9b\u6280\u672f\u4e0d\u4ec5\u53ef\u4ee5\u7528\u4e8e\u56fe\u50cf\u6570\u636e\uff0c\u4e5f\u9002\u7528\u4e8e\u6587\u672c\u3001\u97f3\u9891\u7b49\u5176\u4ed6\u6570\u636e\u7c7b\u578b\u3002\u5728Python\u4e2d\uff0cimgaug<\/code>\u548cAlbumentations<\/code>\u662f\u4e13\u95e8\u7528\u4e8e\u56fe\u50cf\u6570\u636e\u589e\u5f3a\u7684\u5e93\uff0c\u800cnltk<\/code>\u548cTextAugment<\/code>\u5219\u53ef\u4ee5\u5e2e\u52a9\u5904\u7406\u6587\u672c\u6570\u636e\u3002\u7ed3\u5408\u8fd9\u4e9b\u6280\u672f\uff0c\u53ef\u4ee5\u6709\u6548\u63d0\u9ad8\u6837\u672c\u7684\u591a\u6837\u6027\u3002<\/p>\n\u6269\u5145\u6837\u672c\u65f6\uff0c\u5982\u4f55\u907f\u514d\u8fc7\u62df\u5408\uff1f<\/strong>
\u5728\u6269\u5145\u6837\u672c\u7684\u8fc7\u7a0b\u4e2d\uff0c\u4fdd\u6301\u6570\u636e\u7684\u591a\u6837\u6027\u662f\u5173\u952e\u3002\u91c7\u7528\u6570\u636e\u589e\u5f3a\u6280\u672f\u65f6\uff0c\u5e94\u786e\u4fdd\u751f\u6210\u7684\u65b0\u6837\u672c\u5177\u6709\u4ee3\u8868\u6027\uff0c\u907f\u514d\u7b80\u5355\u590d\u5236\u73b0\u6709\u6837\u672c\u3002\u6b64\u5916\uff0c\u53ef\u4ee5\u4f7f\u7528\u4ea4\u53c9\u9a8c\u8bc1\u548c\u6b63\u5219\u5316\u7b49\u65b9\u6cd5\u6765\u76d1\u63a7\u6a21\u578b\u7684\u6027\u80fd\uff0c\u4ece\u800c\u9632\u6b62\u8fc7\u62df\u5408\u3002\u786e\u4fdd\u6a21\u578b\u5728\u9a8c\u8bc1\u96c6\u4e0a\u7684\u8868\u73b0\u826f\u597d\u662f\u8bc4\u4f30\u6570\u636e\u6269\u5145\u6548\u679c\u7684\u91cd\u8981\u6307\u6807\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"\u5f00\u5934\u6bb5\u843d\uff1a \u5728\u6570\u636e\u79d1\u5b66\u548c\u673a\u5668\u5b66\u4e60\u4e2d\uff0c\u6269\u5145\u6837\u672c\u662f\u63d0\u9ad8\u6a21\u578b\u6027\u80fd\u548c\u6cdb\u5316\u80fd\u529b\u7684\u91cd\u8981\u6b65\u9aa4\u3002\u4f7f\u7528Python\u6269\u5145\u6837\u672c\u7684\u65b9\u6cd5 […]","protected":false},"author":3,"featured_media":986264,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[37],"tags":[],"acf":[],"_links":{"self":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/986256"}],"collection":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/comments?post=986256"}],"version-history":[{"count":"1","href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/986256\/revisions"}],"predecessor-version":[{"id":986268,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/986256\/revisions\/986268"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/media\/986264"}],"wp:attachment":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/media?parent=986256"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/categories?post=986256"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/tags?post=986256"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
![\"\u7528python\u5982\u4f55\u6269\u5145\u6837\u672c\"](\"https:\/\/cdn-kb.worktile.com\/kb\/wp-content\/uploads\/2024\/04\/25062923\/5b909bd0-c423-43ec-ab7e-3ff515890f16.webp\")
<\/p>\n