> Model Architecture > Inspect Model

Sequential Model
Python For Data Science

>>> model.output_shape #Model output shape

>>> model.summary() #Model summary representation

>>> model.get_config() #Model configuration

>>> from tensorflow.keras.models import Sequential

>>> model.get_weights() #List all weight tensors in the model

Keras Cheat Sheet

>>> model = Sequential()

>>> model2 = Sequential()

>>> model3 = Sequential()

Multilayer Perceptron (MLP)

> Compile Model
Learn Keras online at www.DataCamp.com
MLP: Binary Classification
Binary Classification
>>> model.compile(optimizer='adam',

>>> from tensorflow.keras.layers import Dense

loss='binary_crossentropy',

>>> model.add(Dense(12,
metrics=['accuracy'])
input_dim=8,

MLP: Multi-Class Classification

kernel_initializer='uniform',

Keras activation='relu'))

>>> model.add(Dense(8,kernel_initializer='uniform',activation='relu'))

>>> model.compile(optimizer='rmsprop',

loss='categorical_crossentropy',

>>> model.add(Dense(1,kernel_initializer='uniform',activation='sigmoid')) metrics=['accuracy'])

Keras is a powerful and easy-to-use deep learning library forTheano and Multi-Class Classification MLP: Regression
TensorFlow that provides a high-level neural networks API to develop and >>> from tensorflow.keras.layers import Dropout
>>> model.compile(optimizer='rmsprop',

>>> model.add(Dense(512,activation='relu',input_shape=(784,)))
loss='mse',

evaluate deep learning models. >>> model.add(Dropout(0.2))

metrics=['mae'])
>>> model.add(Dense(512,activation='relu'))

A Basic Example >>> model.add(Dropout(0.2))

Recurrent Neural Network

>>> import numpy as np
>>> model.add(Dense(10,activation='softmax')) >>> model3.compile(loss='binary_crossentropy',

>>> from tensorflow.keras.models import Sequential

optimizer='adam',

>>> from tensorflow.keras.layers import Dense

Regression metrics=['accuracy'])
>>> data = np.random.random((1000,100))
>>> model.add(Dense(64,activation='relu',input_dim=train_data.shape[1]))

>>> labels = np.random.randint(2,size=(1000,1))

>>> model.add(Dense(1))
>>> model = Sequential()

>>> model.add(Dense(32,

activation='relu',
Convolutional Neural Network (CNN) > Model Training
input_dim=100))

>>> model.add(Dense(1, activation='sigmoid'))

>>> from tensorflow.keras.layers import Activation,Conv2D,MaxPooling2D,Flatten
>>> model3.fit(x_train4,

>>> model.compile(optimizer='rmsprop',
>>> model2.add(Conv2D(32,(3,3),padding='same',input_shape=x_train.shape[1:]))
y_train4,

loss='binary_crossentropy',
>>> model2.add(Activation('relu'))
batch_size=32,

metrics=['accuracy'])
>>> model2.add(Conv2D(32,(3,3)))
epochs=15,

>>> model.fit(data,labels,epochs=10,batch_size=32)
>>> model2.add(Activation('relu'))
verbose=1,

>>> predictions = model.predict(data) >>> model2.add(MaxPooling2D(pool_size=(2,2)))

validation_data=(x_test4,y_test4))
>>> model2.add(Dropout(0.25))

>>> model2.add(Conv2D(64,(3,3), padding='same'))

> Data
>>> model2.add(Activation('relu'))

>>>
>>>
model2.add(Conv2D(64,(3, 3)))

model2.add(Activation('relu'))
> Evaluate Your Model's Performance
>>> model2.add(MaxPooling2D(pool_size=(2,2)))

Your data needs to be stored as NumPy arrays or as a list of NumPy arrays. Ideally, you split the data in training and >>> model2.add(Dropout(0.25))
>>> score = model3.evaluate(x_test,

test sets, for which you can also resort to the train_test_split module of sklearn.cross_validation. >>> model2.add(Flatten())
y_test,

>>> model2.add(Dense(512))
batch_size=32)
>>> model2.add(Activation('relu'))

Keras Data Sets >>> model2.add(Dropout(0.5))

>>> model2.add(Dense(num_classes))

>>>
>>>
from tensorflow.keras.datasets import boston_housing, mnist, cifar10, imdb

(x_train,y_train),(x_test,y_test) = mnist.load_data()

>>> model2.add(Activation('softmax'))
> Save/ Reload Models
>>>
>>>
(x_train2,y_train2),(x_test2,y_test2) = boston_housing.load_data()

(x_train3,y_train3),(x_test3,y_test3) = cifar10.load_data()

Recurrent Neural Network (RNN) >>> from keras.models import load_model

>>> (x_train4,y_train4),(x_test4,y_test4) = imdb.load_data(num_words=20000)

>>> model3.save('model_file.h5')

>>> num_classes = 10 >>> from tensorflow.keras.klayers import Embedding,LSTM

>>> my_model = load_model('my_model.h5')
>>> model3.add(Embedding(20000,128))

>>> model3.add(LSTM(128,dropout=0.2,recurrent_dropout=0.2))

Other >>> model3.add(Dense(1,activation='sigmoid'))

>>> from urllib.request import urlopen

> Model Fine-tuning
> Prediction
>>> data =
np.loadtxt(urlopen("http://archive.ics.uci.edu/ml/machine-learning-databases/pima-indians-di
abetes/pima-indians-diabetes.data"),delimiter=",")
Optimization Parameters
>>> X = data[:,0:8]

>>> y = data [:,8] >>> model3.predict(x_test4, batch_size=32)

>>> from tensorflow.keras.optimizers import RMSprop

>>> model3.predict_classes(x_test4,batch_size=32) >>> opt = RMSprop(lr=0.0001, decay=1e-6)

>>> model2.compile(loss='categorical_crossentropy',

optimizer=opt,

> Preprocessing Also see NumPy & Scikit-Learn metrics=['accuracy'])

Early Stopping
Sequence Padding Train and Test Sets
>>> from tensorflow.keras.callbacks import EarlyStopping

>>> from tensorflow.keras.preprocessing import sequence

>>> from sklearn.model_selection import train_test_split
>>> early_stopping_monitor = EarlyStopping(patience=2)

>>> x_train4 = sequence.pad_sequences(x_train4,maxlen=80)

>>> X_train5,X_test5,y_train5,y_test5 = train_test_split(X, y,
>>> model3.fit(x_train4,

>>> x_test4 = sequence.pad_sequences(x_test4,maxlen=80) test_size=0.33,

y_train4,

random_state=42) batch_size=32,

epochs=15,

One-Hot Encoding validation_data=(x_test4,y_test4),

Standardization/Normalization callbacks=[early_stopping_monitor])
>>> from tensorflow.keras.utils import to_categorical

>>> Y_train = to_categorical(y_train, num_classes)

>>> from sklearn.preprocessing import StandardScaler

>>> Y_test = to_categorical(y_test, num_classes)

>>> scaler = StandardScaler().fit(x_train2)

>>> Y_train3 = to_categorical(y_train3, num_classes)

>>> standardized_X = scaler.transform(x_train2)

>>> Y_test3 = to_categorical(y_test3, num_classes) >>> standardized_X_test = scaler.transform(x_test2) Learn Data Skills Online at www.DataCamp.com