keras自然语言处理（三）——标准网络模型

110 阅读 0 评论 73 点赞

我是靠谱客的博主悲凉大象，这篇文章主要介绍keras自然语言处理（三）——标准网络模型，现在分享给大家，希望可以做个参考。

1.3 标准网路模型

在开始使用函数式API时，最好先了解一些标准神经网路模型的定义，本节我们将介绍如何定义一个简单的多层感知器、卷积神经网路和递归神经网路。这些示例将为后面更详细的示例提供基础。

1.3.1 多层感知器

在这里，我们将定义一个用于二分类的多层感知器模型，该模型有10个输入，3个隐藏层，分别有16,32和16个神经元，以及1个输出层，每个隐藏层中使用relu激活函数，输出层使用sigmoid激活函数，用于二分类。

复制代码

1
2
3
4
5
6
7
8
9
10
11
from keras.utils import plot_model
from keras.models import Model
from keras.layers import Input,Dense
input_tensor = Input(shape=(10,))
x1 = Dense(16,activation='relu')(input_tensor)
x2 = Dense(32,activation='relu')(x1)
x3 = Dense(16,activation='relu')(x2)
output = Dense(1,activation='sigmoid')(x3)
model = Model(inputs = input_tensor,outputs = output)
plot_model(model,to_file='model_graph.png')

注意创建模型图需要安装pydot和pygraphviz。

1.3.2 卷积神经网络

在此我们定义用于图像分类的卷积神经网络，该网络接受黑白64×64图像作为输入，然后具有两个卷积和池化层的序列作为特征提取器，接着全连接层，并用sigmoid来激活

复制代码

1
2
3
4
5
6
7
8
9
10
11
12
13
14
from keras.utils import plot_model
from keras.models import Model
from keras.layers import Input,Dense,Conv2D,MaxPooling2D
input_tensor = Input(shape=(64,64,1))
conv1 = Conv2D(32,kernel_size=4,activation='relu')(input_tensor)
pool1 = MaxPooling2D(pool_size=(2,2))(conv1)
conv2 = Conv2D(16,kernel_size=4,activation='relu')(pool1)
pool2 = MaxPooling2D(pool_size=(2,2))(conv2)
x = Dense(16,activation='relu')(pool2)
output = Dense(1,activation='sigmoid')(x)
model = Model(inputs=input_tensor,outputs = output)
model.summary()
plot_model(model,to_file='conv model.png')

model.summary()：

复制代码

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         (None, 64, 64, 1)         0         
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 61, 61, 32)        544       
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 30, 30, 32)        0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 27, 27, 16)        8208      
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 13, 13, 16)        0         
_________________________________________________________________
dense_1 (Dense)              (None, 13, 13, 16)        272       
_________________________________________________________________
dense_2 (Dense)              (None, 13, 13, 1)         17        
=================================================================
Total params: 9,041
Trainable params: 9,041
Non-trainable params: 0
_________________________________________________________________

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         (None, 64, 64, 1)         0         
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 61, 61, 32)        544       
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 30, 30, 32)        0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 27, 27, 16)        8208      
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 13, 13, 16)        0         
_________________________________________________________________
dense_1 (Dense)              (None, 13, 13, 16)        272       
_________________________________________________________________
dense_2 (Dense)              (None, 13, 13, 1)         17        
=================================================================
Total params: 9,041
Trainable params: 9,041
Non-trainable params: 0
_________________________________________________________________

使用函数式API定义的一个简单的CNN模型图
在这里插入图片描述

1.3.3 递归神经网络

在此我们定义一个用于序列分类的的长短期记忆递归神经网络，该模型需要100个时间步长作为输入，该模型具有单个LSTM隐藏层以从序列中提取特征，接着是全连接层来介绍LSTM的输出，接着用二元预测的输出层

复制代码

1
2
3
4
5
6
7
8
9
10
11
from keras.utils import plot_model
from keras.models import Model
from keras.layers import Input,Dense,LSTM
input_tensor = Input(shape=(100,1))
lstm = LSTM(16)(input_tensor)
x = Dense(16,activation='relu')(lstm)
output = Dense(1,activation='sigmoid')(x)
model = Model(input_tensor,output)
model.summary()
plot_model(model,to_file='recurrent_neural_network.png')

model.summary()：

复制代码

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         (None, 100, 1)            0         
_________________________________________________________________
lstm_1 (LSTM)                (None, 16)                1152      
_________________________________________________________________
dense_1 (Dense)              (None, 16)                272       
_________________________________________________________________
dense_2 (Dense)              (None, 1)                 17        
=================================================================
Total params: 1,441
Trainable params: 1,441
Non-trainable params: 0
_________________________________________________________________