e2efs 0.0.3

E2E-FS Feature Selection Method

E2E-FS

E2E-FS: An End-to-End Feature Selection Method for Neural Networks

CONTACT

This project is hosted at https://github.com/braisCB/E2E-FS.

REFERENCE

If you plan to use this code, please cite the following paper:

Cancela, B., Bolón-Canedo, V., & Alonso-Betanzos, A. (2020). E2E-FS: An End-to-End Feature Selection Method for Neural Networks. IEEE Transactions on Pattern Analysis and Machine Intelligence. (Pending on publication)

EXAMPLE OF USE

# REQUIRED IMPORTS TO CREATE THE MODEL
from keras.datasets import mnist
from keras.callbacks import LearningRateScheduler
from keras.utils import to_categorical
from keras import optimizers, models, layers
# E2EFS IMPORT
from e2efs import models


# DEFINE YOUR CLASSIFIER
def three_layer_nn(input_shape, nclasses):
    return models.Sequential([
        layers.Flatten(input_shape=input_shape),
        layers.Dense(50),
        layers.BatchNormalization(),
        layers.Activation('relu'),
        layers.Dense(25),
        layers.BatchNormalization(),
        layers.Activation('relu'),
        layers.Dense(10, activation='softmax')
    ])


if __name__ == '__main__':

    ## LOAD DATA
    (x_train, y_train), (x_test, y_test) = mnist.load_data()
    x_train = np.expand_dims(x_train, axis=-1)
    x_test = np.expand_dims(x_test, axis=-1)
    y_train = to_categorical(y_train)
    y_test = to_categorical(y_test)

    ## LOAD MODEL AND COMPILE IT (NEVER FORGET TO COMPILE!)
    model = three_layer_nn(input_shape=x_train.shape[1:], nclasses=10)
    model.compile(optimizer=optimizers.SGD(), metrics=['acc'], loss='categorical_crossentropy')

    ## LOAD E2EFS AND RUN IT
    fs_class = models.E2EFSSoft(n_features_to_select=39).attach(model).fit(
        x_train, y_train, batch_size=128, validation_data=(x_test, y_test), verbose=2
    )

    ## FINE TUNING
    def scheduler(epoch):
        if epoch < 20:
            return .1
        elif epoch < 40:
            return .02
        elif epoch < 50:
            return .004
        else:
            return .0008

    fs_class.fine_tuning(x_train, y_train, epochs=60, batch_size=128, 
                         validation_data=(x_test, y_test),
                         callbacks=[LearningRateScheduler(scheduler)], verbose=2)
    print('FEATURE_RANKING :', fs_class.get_ranking())
    print('ACCURACY : ', fs_class.get_model().evaluate(x_test, y_test, batch_size=128)[-1])
    print('FEATURE_MASK NNZ :', np.count_nonzero(fs_class.get_mask()))