keras 的 example 文件 imdb_bidirectional_lstm.py 解析

imdb是一个文本情感分析的数据集，通过评论来分析观众对电影是好评还是差评

其网络结构比较简单

________________________________________________________________________________
Layer (type)                        Output Shape                    Param #     
================================================================================
embedding_1 (Embedding)             (None, 100, 128)                2560000     
________________________________________________________________________________
bidirectional_1 (Bidirectional)     (None, 128)                     98816       
________________________________________________________________________________
dropout_1 (Dropout)                 (None, 128)                     0           
________________________________________________________________________________
dense_1 (Dense)                     (None, 1)                       129         
================================================================================
Total params: 2,658,945
Trainable params: 2,658,945
Non-trainable params: 0
________________________________________________________________________________

对imdb数据集稍微分析一下，

通过函数load_data获取到的x_train, y_train，是一堆编号，这个编号不太直接，可以通过下面代码解析出来：

word_index = imdb.get_word_index()

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])

for i in range(10):
    print(decode_review(x_train[i]))
    print(y_train[i])

就可以看到评论的具体内容，而y_train打印出来的是0和1，分别代表差评和好评

x_train和y_train的shape分别为

(25000, 100)
(25000,)

 

——————————————————————————————————

不另开帖子了，把其他几个网络的结构也贴出来备忘：

imdb_cnn_lstm.py的神经网络结构如下：

________________________________________________________________________________
Layer (type)                        Output Shape                    Param #     
================================================================================
embedding_1 (Embedding)             (None, 100, 128)                2560000     
________________________________________________________________________________
dropout_1 (Dropout)                 (None, 100, 128)                0           
________________________________________________________________________________
conv1d_1 (Conv1D)                   (None, 96, 64)                  41024       
________________________________________________________________________________
max_pooling1d_1 (MaxPooling1D)      (None, 24, 64)                  0           
________________________________________________________________________________
lstm_1 (LSTM)                       (None, 70)                      37800       
________________________________________________________________________________
dense_1 (Dense)                     (None, 1)                       71          
________________________________________________________________________________
activation_1 (Activation)           (None, 1)                       0           
================================================================================
Total params: 2,638,895
Trainable params: 2,638,895
Non-trainable params: 0
________________________________________________________________________________

imdb_cnn.py的神经网络结构如下：

____________________________________________________________________________________________________
Layer (type)                                 Output Shape                            Param #        
====================================================================================================
embedding_1 (Embedding)                      (None, 400, 50)                         250000         
____________________________________________________________________________________________________
dropout_1 (Dropout)                          (None, 400, 50)                         0              
____________________________________________________________________________________________________
conv1d_1 (Conv1D)                            (None, 398, 250)                        37750          
____________________________________________________________________________________________________
global_max_pooling1d_1 (GlobalMaxPooling1D)  (None, 250)                             0              
____________________________________________________________________________________________________
dense_1 (Dense)                              (None, 250)                             62750          
____________________________________________________________________________________________________
dropout_2 (Dropout)                          (None, 250)                             0              
____________________________________________________________________________________________________
activation_1 (Activation)                    (None, 250)                             0              
____________________________________________________________________________________________________
dense_2 (Dense)                              (None, 1)                               251            
____________________________________________________________________________________________________
activation_2 (Activation)                    (None, 1)                               0              
====================================================================================================
Total params: 350,751
Trainable params: 350,751
Non-trainable params: 0
____________________________________________________________________________________________________

 

imdb_lstm.py的神经网络结构为：

_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
embedding_1 (Embedding)      (None, None, 128)         2560000
_________________________________________________________________
lstm_1 (LSTM)                (None, 128)               131584
_________________________________________________________________
dense_1 (Dense)              (None, 1)                 129
=================================================================
Total params: 2,691,713
Trainable params: 2,691,713
Non-trainable params: 0
_________________________________________________________________

——————————————————————

总目录

keras的example文件解析