1. 載入資料

from keras.datasets import imdb
(train_data,train_labels),(test_data,test_labels) = imdb.load_data(num_words=10000)  # num_words表示保留訓練資料中前10000個出個最常出現的單詞，捨棄低頻單詞

<__array_function__ internals>:5: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
E:\my_software\anaconda3\lib\site-packages\tensorflow\python\keras\datasets\imdb.py:159: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
  x_train, y_train = np.array(xs[:idx]), np.array(labels[:idx])
E:\my_software\anaconda3\lib\site-packages\tensorflow\python\keras\datasets\imdb.py:160: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
  x_test, y_test = np.array(xs[idx:]), np.array(labels[idx:])
 

train_data[0]

[1,
 14,
 22,
 16,
 43,
 530,
 973,
 1622,
 1385,
 65,
 458,
 4468,
 66,
 3941,
 4,
 173,
 36,
 256,
 5,
 25,
 100,
 43,
 838,
 112,
 50,
 670,
 2,
 9,
 35,
 480,
 284,
 5,
 150,
 4,
 172,
 112,
 167,
 2,
 336,
 385,
 39,
 4,
 172,
 4536,
 1111,
 17,
 546,
 38,
 13,
 447,
 4,
 192,
 50,
 16,
 6,
 147,
 2025,
 19,
 14,
 22,
 4,
 1920,
 4613,
 469,
 4,
 22,
 71,
 87,
 12,
 16,
 43,
 530,
 38,
 76,
 15,
 13,
 1247,
 4,
 22,
 17,
 515,
 17,
 12,
 16,
 626,
 18,
 2,
 5,
 62,
 386,
 12,
 8,
 316,
 8,
 106,
 5,
 4,
 2223,
 5244,
 16,
 480,
 66,
 3785,
 33,
 4,
 130,
 12,
 16,
 38,
 619,
 5,
 25,
 124,
 51,
 36,
 135,
 48,
 25,
 1415,
 33,
 6,
 22,
 12,
 215,
 28,
 77,
 52,
 5,
 14,
 407,
 16,
 82,
 2,
 8,
 4,
 107,
 117,
 5952,
 15,
 256,
 4,
 2,
 7,
 3766,
 5,
 723,
 36,
 71,
 43,
 530,
 476,
 26,
 400,
 317,
 46,
 7,
 4,
 2,
 1029,
 13,
 104,
 88,
 4,
 381,
 15,
 297,
 98,
 32,
 2071,
 56,
 26,
 141,
 6,
 194,
 7486,
 18,
 4,
 226,
 22,
 21,
 134,
 476,
 26,
 480,
 5,
 144,
 30,
 5535,
 18,
 51,
 36,
 28,
 224,
 92,
 25,
 104,
 4,
 226,
 65,
 16,
 38,
 1334,
 88,
 12,
 16,
 283,
 5,
 16,
 4472,
 113,
 103,
 32,
 15,
 16,
 5345,
 19,
 178,
 32]
 

train_labels[0]

1

word_index = imdb.get_word_index() # 將單詞對映為整數索引的字典
# 將整數索引對映為單詞
reverse_word_index = dict(
    [(value,key) for (key,value) in word_index.items()]
)
# 將評論解碼
# i-3  paddding 、start of sequence、unknown、保留索引
decoded_review = ' '.join(
    [reverse_word_index.get(i-3,'?') for i in train_data[0]]
)
 

2. 將整數序列編碼為二進位制矩陣

import numpy as np
def vectorize_sequences(sequences,dimension=10000):
    # 建立形狀為(len(sequences),dimension)食物零矩陣
    results = np.zeros((len(sequences),dimension))
    for i, sequences in enumerate(sequences):
        results[i,sequences] = 1    # 將results[i]的指定索引設為1
    return results
x_train = vectorize_sequences(train_data)  # 將訓練資料向量化
x_test = vectorize_sequences(test_data)    # 將測試資料向量化

x_train.shape
x_train[0]

array([0., 1., 1., ..., 0., 0., 0.])

x_test.shape
x_test[0]

array([0., 1., 1., ..., 0., 0., 0.])

# 將標籤向量化
y_train = np.asarray(train_labels).astype('float32')
y_test = np.asarray(test_labels).astype('float32')

y_train

array([1., 0., 0., ..., 0., 1., 0.], dtype=float32)

y_test

array([0., 1., 1., ..., 0., 0., 0.], dtype=float32)

3. 構建網路

# 模型定義
from keras import models
from keras import layers
model = models.Sequential()
model.add(layers.Dense(16,activation='relu', input_shape=(10000,)))
model.add(layers.Dense(16,activation='relu'))
model.add(layers.Dense(1,activation='sigmoid'))

4. 編譯模型

model.compile(optimizer='rmsprop',
             loss='binary_crossentropy',
             metrics=['accuracy'])

5.配置優化器

from keras import optimizers
model.compile(optimizer=optimizers.RMSprop(lr=0.001),
             loss='binary_crossentropy',
             metrics=['accuracy'])

6. 自定義損失和指標

from keras import losses
from keras import metrics

model.compile(optimizer=optimizers.RMSprop(lr=0.001),
             loss=losses.binary_crossentropy,
             metrics = [metrics.binary_accuracy])

7. 留出驗證集

x_val = x_train[:10000]
partial_x_train = x_train[10000:]

y_val = y_train[:10000]
partial_y_train = y_train[10000:]

8. 訓練模型

model.compile(optimizer='rmsprop',
             loss='binary_crossentropy',
             metrics=['acc'])

history = model.fit(partial_x_train,
                   partial_y_train,
                   epochs=20,
                   batch_size=512,
                   validation_data=(x_val,y_val))

Epoch 1/20
30/30 [==============================] - 6s 137ms/step - loss: 0.6149 - acc: 0.6893 - val_loss: 0.4173 - val_acc: 0.8633
Epoch 2/20
30/30 [==============================] - 1s 33ms/step - loss: 0.3482 - acc: 0.8993 - val_loss: 0.3179 - val_acc: 0.8861
Epoch 3/20
30/30 [==============================] - 1s 28ms/step - loss: 0.2478 - acc: 0.9261 - val_loss: 0.2951 - val_acc: 0.8857
Epoch 4/20
30/30 [==============================] - 1s 28ms/step - loss: 0.1934 - acc: 0.9394 - val_loss: 0.2740 - val_acc: 0.8904
Epoch 5/20
30/30 [==============================] - 1s 27ms/step - loss: 0.1523 - acc: 0.9566 - val_loss: 0.2881 - val_acc: 0.8845
Epoch 6/20
30/30 [==============================] - 1s 27ms/step - loss: 0.1294 - acc: 0.9624 - val_loss: 0.2855 - val_acc: 0.8859
Epoch 7/20
30/30 [==============================] - 1s 26ms/step - loss: 0.1056 - acc: 0.9713 - val_loss: 0.2986 - val_acc: 0.8846
Epoch 8/20
30/30 [==============================] - 1s 29ms/step - loss: 0.0858 - acc: 0.9791 - val_loss: 0.3196 - val_acc: 0.8824
Epoch 9/20
30/30 [==============================] - 1s 28ms/step - loss: 0.0714 - acc: 0.9821 - val_loss: 0.3501 - val_acc: 0.8742
Epoch 10/20
30/30 [==============================] - 1s 27ms/step - loss: 0.0595 - acc: 0.9862 - val_loss: 0.3557 - val_acc: 0.8801
Epoch 11/20
30/30 [==============================] - 1s 28ms/step - loss: 0.0464 - acc: 0.9898 - val_loss: 0.3814 - val_acc: 0.8785
Epoch 12/20
30/30 [==============================] - 1s 28ms/step - loss: 0.0356 - acc: 0.9933 - val_loss: 0.4075 - val_acc: 0.8765
Epoch 13/20
30/30 [==============================] - 1s 27ms/step - loss: 0.0306 - acc: 0.9944 - val_loss: 0.4390 - val_acc: 0.8730
Epoch 14/20
30/30 [==============================] - 1s 27ms/step - loss: 0.0241 - acc: 0.9960 - val_loss: 0.4865 - val_acc: 0.8744
Epoch 15/20
30/30 [==============================] - 1s 27ms/step - loss: 0.0186 - acc: 0.9975 - val_loss: 0.5017 - val_acc: 0.8692
Epoch 16/20
30/30 [==============================] - 1s 26ms/step - loss: 0.0132 - acc: 0.9989 - val_loss: 0.5573 - val_acc: 0.8700
Epoch 17/20
30/30 [==============================] - 1s 26ms/step - loss: 0.0118 - acc: 0.9989 - val_loss: 0.5675 - val_acc: 0.8711
Epoch 18/20
30/30 [==============================] - 1s 26ms/step - loss: 0.0082 - acc: 0.9993 - val_loss: 0.6111 - val_acc: 0.8701
Epoch 19/20
30/30 [==============================] - 1s 26ms/step - loss: 0.0057 - acc: 0.9998 - val_loss: 0.6308 - val_acc: 0.8685
Epoch 20/20
30/30 [==============================] - 1s 25ms/step - loss: 0.0040 - acc: 0.9998 - val_loss: 0.6980 - val_acc: 0.8664

history_dict = history.history
history_dict.keys()

dict_keys(['loss', 'acc', 'val_loss', 'val_acc'])

9. 繪製訓練損失和驗證損失

import matplotlib.pyplot as plt
history_dict = history.history
loss_values = history_dict['loss']
val_loss_values = history_dict['val_loss']

epochs = range(1,len(loss_values)+1)

plt.plot(epochs,loss_values,'bo',label='Training loss')
plt.plot(epochs,val_loss_values,'b',label='Validation loss')
plt.title('Training and validation loss')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.show()

10. 繪製訓練精度和驗證精度

plt.clf()  #清空影象
acc = history_dict['acc']
val_acc = 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()