您好,登录后才能下订单哦!
小编给大家分享一下关于keras分类模型中输入数据与标签维度的案例,希望大家阅读完这篇文章后大所收获,下面让我们一起去探讨吧!
在《python深度学习》这本书中。
一、21页mnist十分类
导入数据集 from keras.datasets import mnist (train_images, train_labels), (test_images, test_labels) = mnist.load_data() 初始数据维度: >>> train_images.shape (60000, 28, 28) >>> len(train_labels) 60000 >>> train_labels array([5, 0, 4, ..., 5, 6, 8], dtype=uint8) 数据预处理: train_images = train_images.reshape((60000, 28 * 28)) train_images = train_images.astype('float32') / 255 train_labels = to_categorical(train_labels) 之后: print(train_images, type(train_images), train_images.shape, train_images.dtype) print(train_labels, type(train_labels), train_labels.shape, train_labels.dtype) 结果: [[0. 0. 0. ... 0. 0. 0.] [0. 0. 0. ... 0. 0. 0.] [0. 0. 0. ... 0. 0. 0.] ... [0. 0. 0. ... 0. 0. 0.] [0. 0. 0. ... 0. 0. 0.] [0. 0. 0. ... 0. 0. 0.]] <class 'numpy.ndarray'> (60000, 784) float32 [[0. 0. 0. ... 0. 0. 0.] [1. 0. 0. ... 0. 0. 0.] [0. 0. 0. ... 0. 0. 0.] ... [0. 0. 0. ... 0. 0. 0.] [0. 0. 0. ... 0. 0. 0.] [0. 0. 0. ... 0. 1. 0.]] <class 'numpy.ndarray'> (60000, 10) float32
二、51页IMDB二分类
导入数据:
from keras.datasets import imdb (train_data, train_labels), (test_data, test_labels) = imdb.load_data(num_words=10000)
参数 num_words=10000 的意思是仅保留训练数据中前 10 000 个最常出现的单词。
train_data和test_data都是numpy.ndarray类型,都是一维的(共25000个元素,相当于25000个list),其中每个list代表一条评论,每个list中的每个元素的值范围在0-9999 ,代表10000个最常见单词的每个单词的索引,每个list长度不一,因为每条评论的长度不一,例如train_data中的list最短的为11,最长的为189。
train_labels和test_labels都是含25000个元素(元素的值要不0或者1,代表两类)的list。
数据预处理:
# 将整数序列编码为二进制矩阵 def vectorize_sequences(sequences, dimension=10000): # Create an all-zero matrix of shape (len(sequences), dimension) results = np.zeros((len(sequences), dimension)) for i, sequence in enumerate(sequences): results[i, sequence] = 1. # set specific indices of results[i] to 1s return results x_train = vectorize_sequences(train_data) x_test = vectorize_sequences(test_data) 第一种方式:shape为(25000,) y_train = np.asarray(train_labels).astype('float32') #就用这种方式就行了 y_test = np.asarray(test_labels).astype('float32') 第二种方式:shape为(25000,1) y_train = np.asarray(train_labels).astype('float32').reshape(25000, 1) y_test = np.asarray(test_labels).astype('float32').reshape(25000, 1) 第三种方式:shape为(25000,2) y_train = to_categorical(train_labels) #变成one-hot向量 y_test = to_categorical(test_labels)
第三种方式,相当于把二分类看成了多分类,所以网络的结构同时需要更改,
最后输出的维度:1->2
最后的激活函数:sigmoid->softmax
损失函数:binary_crossentropy->categorical_crossentropy
预处理之后,train_data和test_data变成了shape为(25000,10000),dtype为float32的ndarray(one-hot向量),train_labels和test_labels变成了shape为(25000,)的一维ndarray,或者(25000,1)的二维ndarray,或者shape为(25000,2)的one-hot向量。
注:
1.sigmoid对应binary_crossentropy,softmax对应categorical_crossentropy
2.网络的所有输入和目标都必须是浮点数张量
补充知识:keras输入数据的方法:model.fit和model.fit_generator
1.第一种,普通的不用数据增强的
from keras.datasets import mnist,cifar10,cifar100 (X_train, y_train), (X_valid, Y_valid) = cifar10.load_data() model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch, shuffle=True, verbose=1, validation_data=(X_valid, Y_valid), )
2.第二种,带数据增强的 ImageDataGenerator,可以旋转角度、平移等操作。
from keras.preprocessing.image import ImageDataGenerator (trainX, trainY), (testX, testY) = cifar100.load_data() trainX = trainX.astype('float32') testX = testX.astype('float32') trainX /= 255. testX /= 255. Y_train = np_utils.to_categorical(trainY, nb_classes) Y_test = np_utils.to_categorical(testY, nb_classes) generator = ImageDataGenerator(rotation_range=15, width_shift_range=5./32, height_shift_range=5./32) generator.fit(trainX, seed=0) model.fit_generator(generator.flow(trainX, Y_train, batch_size=batch_size), steps_per_epoch=len(trainX) // batch_size, epochs=nb_epoch, callbacks=callbacks, validation_data=(testX, Y_test), validation_steps=testX.shape[0] // batch_size, verbose=1)
看完了这篇文章,相信你对关于keras分类模型中输入数据与标签维度的案例有了一定的了解,想了解更多相关知识,欢迎关注亿速云行业资讯频道,感谢各位的阅读!
免责声明:本站发布的内容(图片、视频和文字)以原创、转载和分享为主,文章观点不代表本网站立场,如果涉及侵权请联系站长邮箱:is@yisu.com进行举报,并提供相关证据,一经查实,将立刻删除涉嫌侵权内容。