Python深度学习笔记09--使用Keras建立循环神经网络

6.2 理解循环神经网络
6.2.1 Keras中的循环层：
 1 from keras.layers import SimpleRNN
 2 
 3 from keras.models import Sequential
 4 from keras.layers import Embedding, SimpleRNN
 5 
 6 # model = Sequential()
 7 # model.add(Embedding(10000, 32))
 8 # model.add(SimpleRNN(32))
 9 # print(model.summary())
10 '''
11 embedding层的参数计算：10000 * 32
12 imple_rnn层的参数计算：x*h + h*h + y*h + x + y
13 x是输入向量的维度，本例中是32(第7行定义的32)
14 h是循环层的向量维度，本例中是32(第8行定义的32)
15 y是0，本例中没有定义输出
16 因此计算方式为：32*32 + 32*32 + 0*32 + 32 + 0 = 2080
17 '''
18 # model = Sequential()
19 # model.add(Embedding(10000, 32))
20 # model.add(SimpleRNN(32, return_sequences=True))
21 # print(model.summary())
22 
23 
24 # model = Sequential()
25 # model.add(Embedding(10000, 32))
26 # model.add(SimpleRNN(32, return_sequences=True))
27 # model.add(SimpleRNN(32, return_sequences=True))
28 # model.add(SimpleRNN(32, return_sequences=True)) # 返回完整状态序列
29 # model.add(SimpleRNN(32))  # 最后一层只返回最终输出
30 # print(model.summary())
31 
32 
33 from keras.datasets import imdb
34 from keras.preprocessing import sequence
35 
36 max_features = 10000  # number of words to consider as features
37 maxlen = 500  # cut texts after this number of words (among top max_features most common words)
38 batch_size = 32
39 
40 print('Loading data...')
41 (input_train, y_train), (input_test, y_test) = imdb.load_data(num_words=max_features)
42 print(len(input_train), 'train sequences')
43 print(len(input_test), 'test sequences')
44 
45 print('Pad sequences (samples x time)')
46 input_train = sequence.pad_sequences(input_train, maxlen=maxlen)
47 input_test = sequence.pad_sequences(input_test, maxlen=maxlen)
48 print('input_train shape:', input_train.shape)
49 print('input_test shape:', input_test.shape)
50 
51 
52 from keras.layers import Dense
53 
54 model = Sequential()
55 model.add(Embedding(max_features, 32))
56 model.add(SimpleRNN(32))
57 model.add(Dense(1, activation='sigmoid'))
58 # #print(model.summary())
59 
60 
61 model.compile(optimizer='rmsprop', loss='binary_crossentropy', metrics=['acc'])
62 history = model.fit(input_train, y_train,
63                     epochs=10,
64                     batch_size=128,
65                     validation_split=0.2)
66 
67 
68 import matplotlib.pyplot as plt
69 
70 acc = history.history['acc']
71 val_acc = history.history['val_acc']
72 loss = history.history['loss']
73 val_loss = history.history['val_loss']
74 
75 epochs = range(len(acc))
76 
77 plt.plot(epochs, acc, 'bo', label='Training acc')
78 plt.plot(epochs, val_acc, 'b', label='Validation acc')
79 plt.title('Training and validation accuracy')
80 plt.legend()
81 
82 plt.figure()
83 
84 plt.plot(epochs, loss, 'bo', label='Training loss')
85 plt.plot(epochs, val_loss, 'b', label='Validation loss')
86 plt.title('Training and validation loss')
87 plt.legend()
88 
89 plt.show()
90 
91 # loss: 0.0107 - acc: 0.9970 - val_loss: 0.6327 - val_acc: 0.8286
posted on 2021-01-25 15:05 Sempron2800+ 阅读(236) 评论(0) 编辑收藏举报