tf.keras学习总结（三）：keras的函数式编程（Text-CNN实现）

首先回归一下之前写的一篇博客：keras函数式编程的官方文档示例

---

本篇文章以Text-CNN模型的实现为例，展示一下keras函数式编程的使用。建议先了解一下text-CNN模型。

---

首先贴上keras函数式编程的第一句语句：

keras.layers.Input(shape=None,batch_shape=None,name=None,dtype=K.floatx(),sparse=False,tensor=None)

Input(shape=None,batch_shape=None,name=None,dtype=K.floatx(),sparse=False,tensor=None)

Input():用来实例化一个keras张量

#参数：

• shape: 形状元组（整型），不包括batch size。for instance, shape=(32,) 表示了预期的输入将是一批32维的向量。
• batch_shape: 形状元组（整型），包括了batch size。for instance, batch_shape=(10,32)表示了预期的输入将是10个32维向量的批次。
• name: 对于该层是可选的名字字符串。在一个模型中是独一无二的（同一个名字不能复用2次）。如果name没有被特指将会自动生成。
• dtype: 预期的输入数据类型
• sparse: 特定的布尔值，占位符是否为sparse
• tensor: 可选的存在的向量包装到Input层，如果设置了，该层将不会创建一个占位张量。

#返回

一个张量

---

下面贴上tensorflow官方文档上的示例：

inputs = tf.keras.Input(shape=(32,))  # Returns a placeholder tensor

# A layer instance is callable on a tensor, and returns a tensor.
x = layers.Dense(64, activation='relu')(inputs)
x = layers.Dense(64, activation='relu')(x)
predictions = layers.Dense(10, activation='softmax')(x)
# 在给定输入和输出的情况下实例化模型。
model = tf.keras.Model(inputs=inputs, outputs=predictions)

# The compile step specifies the training configuration.
model.compile(optimizer=tf.train.RMSPropOptimizer(0.001),
              loss='categorical_crossentropy',
              metrics=['accuracy'])

# Trains for 5 epochs
model.fit(data, labels, batch_size=32, epochs=5)

下面为基于keras函数式编程的Text-CNN模型实现：

"""
-------------------------------------------------
   File Name：     text-cnn
   Description :
   Author :       chaorenfei
   date：          2019/4/25
-------------------------------------------------
   Change Activity:
                   2019/4/25:
-------------------------------------------------
"""
"""
CNN模型首次使用在文本分类，是Yoon Kim发表的“Convolutional Neural Networks for Sentence Classification”论文中。
"""
import tensorflow as tf
from tensorflow import keras
import numpy as np
# from tensorflow.keras.layers import Input, Dense


imdb = keras.datasets.imdb

(train_data, train_labels), (test_data, test_labels) = imdb.load_data(num_words=10000)
############################
# explore data
print("Training entries: {}, labels: {}".format(len(train_data), len(train_labels)))
# >> Training entries: 25000, labels: 25000
print(train_data[0])
# >>
print(len(train_data[0]), len(train_data[1]))
############################
# 将整数转换回字词:了解如何将整数转换回文本可能很有用。在以下代码中，我们将创建一个辅助函数来查询包含整数到字符串映射的字典对象：
# A dictionary mapping words to an integer index
word_index = imdb.get_word_index()

# The first indices are reserved
word_index = {k: (v + 3) for k, v in word_index.items()}
word_index["<PAD>"] = 0
word_index["<START>"] = 1
word_index["<UNK>"] = 2  # unknown
word_index["<UNUSED>"] = 3

reverse_word_index = dict([(value, key) for (key, value) in word_index.items()])


def decode_review(text):
    return ' '.join([reverse_word_index.get(i, '?') for i in text])


# 现在，我们可以使用 decode_review 函数显示第一条影评的文本：
print(decode_review(train_data[0]))
####################################
# prepare data
# 我们可以填充数组，使它们都具有相同的长度，然后创建一个形状为 max_length * num_reviews 的整数张量。我们可以使用一个能够处理这种形状的嵌入层作为网络中的第一层。
# 使用 pad_sequences 函数将长度标准化
train_data = keras.preprocessing.sequence.pad_sequences(train_data,
                                                        value=word_index["<PAD>"],
                                                        padding='post',
                                                        maxlen=256)

test_data = keras.preprocessing.sequence.pad_sequences(test_data,
                                                       value=word_index["<PAD>"],
                                                       padding='post',
                                                       maxlen=256)
# now, the len of data is 256
print(train_data[0])
##########################################
#### stucture the model

convs = []
inputs = keras.layers.Input(shape=(256,))
embed1 = keras.layers.Embedding(10000, 32)(inputs)
# embed = keras.layers.Reshape(-1,256, 32, 1)(embed1)
print(embed1[0])
def reshapes(embed1):
    embed = tf.reshape(embed1, [-1, 256, 32, 1]);
    return embed
# embed = tf.reshape(embed1, [-1, 256, 32, 1])
embed = keras.layers.Lambda(reshapes)(embed1);
print(embed[0])
l_conv1 = keras.layers.Conv2D(filters=3, kernel_size=(2, 32), activation='relu')(embed)  #现长度 = 1+（原长度-卷积核大小+2*填充层大小） /步长 卷积核的形状（fsz，embedding_size）
l_pool1 = keras.layers.MaxPooling2D(pool_size=(255, 1))(l_conv1)  # 这里面最大的不同 池化层核的大小与卷积完的数据长度一样
l_pool11 = keras.layers.Flatten()(l_pool1)    #一般为卷积网络最近全连接的前一层，用于将数据压缩成一维
convs.append(l_pool11)
l_conv2 = keras.layers.Conv2D(filters=3, kernel_size=(3, 32), activation='relu')(embed)
l_pool2 = keras.layers.MaxPooling2D(pool_size=(254, 1))(l_conv2)
l_pool22 = keras.layers.Flatten()(l_pool2)
convs.append(l_pool22)
l_conv3 = keras.layers.Conv2D(filters=3, kernel_size=(4, 32), activation='relu')(embed)
l_pool3 = keras.layers.MaxPooling2D(pool_size=(253, 1))(l_conv3)
l_pool33 = keras.layers.Flatten()(l_pool2)
convs.append(l_pool33)
merge = keras.layers.concatenate(convs, axis=1)

out = keras.layers.Dropout(0.5)(merge)

output = keras.layers.Dense(32, activation='relu')(out)

pred = keras.layers.Dense(units=1, activation='sigmoid')(output)

model = keras.models.Model(inputs=inputs, outputs=pred)
# adam = optimizers.Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-08, decay=0.0)
model.summary()
model.compile(loss="binary_crossentropy", optimizer="adam", metrics=['accuracy'])

###############################################################

# validation data
x_val = train_data[:10000]
partial_x_train = train_data[10000:]

y_val = train_labels[:10000]
partial_y_train = train_labels[10000:]
# fit
history = model.fit(partial_x_train, partial_y_train, batch_size=512,
                    epochs=40,
                    validation_data=(x_val, y_val),
                    verbose=1)
# evalute model
results = model.evaluate(test_data, test_labels)

print(results)
# predict data
predictions = model.predict(test_data)
##################################################################
# 创建准确率和损失随时间变化的图
# model.fit() 返回一个 History 对象，该对象包含一个字典，其中包括训练期间发生的所有情况：
history_dict = history.history
print(history_dict.keys())
# >>dict_keys(['loss', 'val_loss', 'val_acc', 'acc'])

# 可以使用这些指标绘制训练损失与验证损失图表以进行对比，并绘制训练准确率与验证准确率图表：
import matplotlib.pyplot as plt

acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']

epochs = range(1, len(acc) + 1)
##########-------------画图方式1-----------------

# # "bo" is for "blue dot"
# plt.plot(epochs, loss, 'bo', label='Training loss')
# # b is for "solid blue line"
# plt.plot(epochs, val_loss, 'b', label='Validation loss')
# plt.title('Training and validation loss')
# plt.xlabel('Epochs')
# plt.ylabel('Loss')
# plt.legend()
#
# plt.show()
# # -----------------------------------------
# plt.clf()  # clear figure
# acc_values = history_dict['acc']
# val_acc_values = history_dict['val_acc']
#
# plt.plot(epochs, acc, 'bo', label='Training acc')
# plt.plot(epochs, val_acc, 'b', label='Validation acc')
# plt.title('Training and validation accuracy')
# plt.xlabel('Epochs')
# plt.ylabel('Accuracy')
# plt.legend()

# plt.show()

#######--------画图方式2-------------------
# fig = plt.figure()
# ax = plt.subplot(1,2,1)
# plt.plot(epochs, loss, 'bo', label='Training loss')
# plt.plot(epochs, val_loss, 'b', label='Validation loss')
# plt.title('Training and validation loss')
# plt.xlabel('Epochs')
# plt.ylabel('Loss')
# plt.legend()
#
# ax2 = plt.subplot(1,2,2)
# acc_values = history_dict['acc']
# val_acc_values = history_dict['val_acc']
#
# plt.plot(epochs, acc, 'bo', label='Training acc')
# plt.plot(epochs, val_acc, 'b', label='Validation acc')
# plt.title('Training and validation accuracy')
# plt.xlabel('Epochs')
# plt.ylabel('Accuracy')
# plt.legend()
#
# plt.show()