Keras: Deep Learning library

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

Keras: Deep Learning library for Theano and TensorFlowYou have just found Keras.

Keras is a high-level neural networks API, written in Python and capable of running on top of either TensorFlow orTheano. It was developed with a focus on enabling fast experimentation. Being able to go from idea to result with the least possible delay is key to doing good research.

Use Keras if you need a deep learning library that:

• Allows for easy and fast prototyping (through user friendliness, modularity, and extensibility).
• Supports both convolutional networks and recurrent networks, as well as combinations of the two.
• Runs seamlessly on CPU and GPU.
  

Read the documentation at Keras.io.

Keras is compatible with: Python 2.7-3.5.

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏2 回帖

关键词：Learning earning Library Learn BRARY developed learning possible library running

1. Getting started with the Keras Sequential model
   
2. 
   
3. The Sequential model is a linear stack of layers.
   
4. 
   
5. You can create a Sequential model by passing a list of layer instances to the constructor:
   
6. 
   
7. from keras.models import Sequential
   
8. from keras.layers import Dense, Activation
   
9. 
   
10. model = Sequential([
    
11.     Dense(32, input_shape=(784,)),
    
12.     Activation('relu'),
    
13.     Dense(10),
    
14.     Activation('softmax'),
    
15. ])

复制代码

1. Specifying the input shape
   
2. 
   
3. The model needs to know what input shape it should expect. For this reason, the first layer in a Sequential model (and only the first, because following layers can do automatic shape inference) needs to receive information about its input shape. There are several possible ways to do this:
   
4. 
   
5. Pass an input_shape argument to the first layer. This is a shape tuple (a tuple of integers or None entries, where None indicates that any positive integer may be expected). In input_shape, the batch dimension is not included.
   
6. Some 2D layers, such as Dense, support the specification of their input shape via the argument input_dim, and some 3D temporal layers support the arguments input_dim and input_length.
   
7. If you ever need to specify a fixed batch size for your inputs (this is useful for stateful recurrent networks), you can pass a batch_size argument to a layer. If you pass both batch_size=32 and input_shape=(6, 8) to a layer, it will then expect every batch of inputs to have the batch shape (32, 6, 8).
   
8. As such, the following snippets are strictly equivalent:
   
9. 
   
10. model = Sequential()
    
11. model.add(Dense(32, input_shape=(784,)))
    
12. model = Sequential()
    
13. model.add(Dense(32, input_dim=784))

复制代码

1. Compilation
   
2. 
   
3. # For a multi-class classification problem
   
4. model.compile(optimizer='rmsprop',
   
5.               loss='categorical_crossentropy',
   
6.               metrics=['accuracy'])
   
7. 
   
8. # For a binary classification problem
   
9. model.compile(optimizer='rmsprop',
   
10.               loss='binary_crossentropy',
    
11.               metrics=['accuracy'])
    
12. 
    
13. # For a mean squared error regression problem
    
14. model.compile(optimizer='rmsprop',
    
15.               loss='mse')
    
16. 
    
17. # For custom metrics
    
18. import keras.backend as K
    
19. 
    
20. def mean_pred(y_true, y_pred):
    
21.     return K.mean(y_pred)
    
22. 
    
23. model.compile(optimizer='rmsprop',
    
24.               loss='binary_crossentropy',
    
25.               metrics=['accuracy', mean_pred])

复制代码

1. # For a single-input model with 2 classes (binary classification):
   
2. 
   
3. model = Sequential()
   
4. model.add(Dense(32, activation='relu', input_dim=100))
   
5. model.add(Dense(1, activation='sigmoid'))
   
6. model.compile(optimizer='rmsprop',
   
7.               loss='binary_crossentropy',
   
8.               metrics=['accuracy'])
   
9. 
   
10. # Generate dummy data
    
11. import numpy as np
    
12. data = np.random.random((1000, 100))
    
13. labels = np.random.randint(2, size=(1000, 1))
    
14. 
    
15. # Train the model, iterating on the data in batches of 32 samples
    
16. model.fit(data, labels, epochs=10, batch_size=32)

复制代码

1. # For a single-input model with 10 classes (categorical classification):
   
2. 
   
3. model = Sequential()
   
4. model.add(Dense(32, activation='relu', input_dim=100))
   
5. model.add(Dense(10, activation='softmax'))
   
6. model.compile(optimizer='rmsprop',
   
7.               loss='categorical_crossentropy',
   
8.               metrics=['accuracy'])
   
9. 
   
10. # Generate dummy data
    
11. import numpy as np
    
12. data = np.random.random((1000, 100))
    
13. labels = np.random.randint(10, size=(1000, 1))
    
14. 
    
15. # Convert labels to categorical one-hot encoding
    
16. one_hot_labels = keras.utils.to_categorical(labels, num_classes=10)
    
17. 
    
18. # Train the model, iterating on the data in batches of 32 samples
    
19. model.fit(data, one_hot_labels, epochs=10, batch_size=32)

复制代码

1. Multilayer Perceptron (MLP) for multi-class softmax classification:
   
2. 
   
3. from keras.models import Sequential
   
4. from keras.layers import Dense, Dropout, Activation
   
5. from keras.optimizers import SGD
   
6. 
   
7. # Generate dummy data
   
8. import numpy as np
   
9. x_train = np.random.random((1000, 20))
   
10. y_train = keras.utils.to_categorical(np.random.randint(10, size=(1000, 1)), num_classes=10)
    
11. x_test = np.random.random((100, 20))
    
12. y_test = keras.utils.to_categorical(np.random.randint(10, size=(100, 1)), num_classes=10)
    
13. 
    
14. model = Sequential()
    
15. # Dense(64) is a fully-connected layer with 64 hidden units.
    
16. # in the first layer, you must specify the expected input data shape:
    
17. # here, 20-dimensional vectors.
    
18. model.add(Dense(64, activation='relu', input_dim=20))
    
19. model.add(Dropout(0.5))
    
20. model.add(Dense(64, activation='relu'))
    
21. model.add(Dropout(0.5))
    
22. model.add(Dense(10, activation='softmax'))
    
23. 
    
24. sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
    
25. model.compile(loss='categorical_crossentropy',
    
26.               optimizer=sgd,
    
27.               metrics=['accuracy'])
    
28. 
    
29. model.fit(x_train, y_train,
    
30.           epochs=20,
    
31.           batch_size=128)
    
32. score = model.evaluate(x_test, y_test, batch_size=128)

复制代码

1. MLP for binary classification:
   
2. 
   
3. import numpy as np
   
4. from keras.models import Sequential
   
5. from keras.layers import Dense, Dropout
   
6. 
   
7. # Generate dummy data
   
8. x_train = np.random.random((1000, 20))
   
9. y_train = np.random.randint(2, size=(1000, 1))
   
10. x_test = np.random.random((100, 20))
    
11. y_test = np.random.randint(2, size=(100, 1))
    
12. 
    
13. model = Sequential()
    
14. model.add(Dense(64, input_dim=20, activation='relu'))
    
15. model.add(Dropout(0.5))
    
16. model.add(Dense(64, activation='relu'))
    
17. model.add(Dropout(0.5))
    
18. model.add(Dense(1, activation='sigmoid'))
    
19. 
    
20. model.compile(loss='binary_crossentropy',
    
21.               optimizer='rmsprop',
    
22.               metrics=['accuracy'])
    
23. 
    
24. model.fit(x_train, y_train,
    
25.           epochs=20,
    
26.           batch_size=128)
    
27. score = model.evaluate(x_test, y_test, batch_size=128)

复制代码

1. VGG-like convnet:
   
2. 
   
3. import numpy as np
   
4. import keras
   
5. from keras.models import Sequential
   
6. from keras.layers import Dense, Dropout, Flatten
   
7. from keras.layers import Conv2D, MaxPooling2D
   
8. from keras.optimizers import SGD
   
9. 
   
10. # Generate dummy data
    
11. x_train = np.random.random((100, 100, 100, 3))
    
12. y_train = keras.utils.to_categorical(np.random.randint(10, size=(100, 1)), num_classes=10)
    
13. x_test = np.random.random((20, 100, 100, 3))
    
14. y_test = keras.utils.to_categorical(np.random.randint(10, size=(20, 1)), num_classes=10)
    
15. 
    
16. model = Sequential()
    
17. # input: 100x100 images with 3 channels -> (100, 100, 3) tensors.
    
18. # this applies 32 convolution filters of size 3x3 each.
    
19. model.add(Conv2D(32, (3, 3), activation='relu', input_shape=(100, 100, 3)))
    
20. model.add(Conv2D(32, (3, 3), activation='relu'))
    
21. model.add(MaxPooling2D(pool_size=(2, 2)))
    
22. model.add(Dropout(0.25))
    
23. 
    
24. model.add(Conv2D(64, (3, 3), activation='relu'))
    
25. model.add(Conv2D(64, (3, 3), activation='relu'))
    
26. model.add(MaxPooling2D(pool_size=(2, 2)))
    
27. model.add(Dropout(0.25))
    
28. 
    
29. model.add(Flatten())
    
30. model.add(Dense(256, activation='relu'))
    
31. model.add(Dropout(0.5))
    
32. model.add(Dense(10, activation='softmax'))
    
33. 
    
34. sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
    
35. model.compile(loss='categorical_crossentropy', optimizer=sgd)
    
36. 
    
37. model.fit(x_train, y_train, batch_size=32, epochs=10)
    
38. score = model.evaluate(x_test, y_test, batch_size=32)

复制代码

1. Sequence classification with LSTM:
   
2. 
   
3. from keras.models import Sequential
   
4. from keras.layers import Dense, Dropout
   
5. from keras.layers import Embedding
   
6. from keras.layers import LSTM
   
7. 
   
8. model = Sequential()
   
9. model.add(Embedding(max_features, output_dim=256))
   
10. model.add(LSTM(128))
    
11. model.add(Dropout(0.5))
    
12. model.add(Dense(1, activation='sigmoid'))
    
13. 
    
14. model.compile(loss='binary_crossentropy',
    
15.               optimizer='rmsprop',
    
16.               metrics=['accuracy'])
    
17. 
    
18. model.fit(x_train, y_train, batch_size=16, epochs=10)
    
19. score = model.evaluate(x_test, y_test, batch_size=16)

复制代码