Feed-Forward NN

Here we need the training and testing data as numpy arrays, not as pandas dataframes, for instance. Thus, if we have our data as pandas dataframes can do something like:

X_train = X_train.values
y_train = y_train.values
X_test = X_test.values
y_test = y_test.values

Below we create and compile the model, and then train it.

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense,Dropout

model = Sequential()
model.add(Dense(79,'relu'))
model.add(Dropout(0.5))
model.add(Dense(39,'relu'))
model.add(Dropout(0.5))
model.add(Dense(19,'relu'))
model.add(Dense(1,'sigmoid'))

model.compile(loss='binary_crossentropy',optimizer='adam')

model.fit(x=X_train,y=y_train,epochs=30,verbose=0,batch_size=256,
	      validation_data=(X_test,y_test))

Saving / loading the model

model.save('my_03_model.h5')

from tensorflow.keras.models import load_model
model = load_model('my_03_model.h5')

Predictions

• If your model does binary classification (e.g. if it uses a sigmoid last-layer activation):

predictions = (model.predict(X_test) > 0.5).astype(int)

• If your model does multi-class classification: (e.g. if it uses a softmax last-layer activation):

predictions = np.argmax(model.predict(X_test), axis=-1)

See discussion here.