JulieProst

from tensorflow.keras.datasets import cifar10
# Load data
(x_train, y_train), (x_test, y_test) = cifar10.load_data()
# Pre-processing
x_train = x_train.astype('float32') / 255.
x_test = x_test.astype('float32') / 255.

JulieProst

from tensorflow import keras
from tensorflow.keras.layers import (
    Conv2D,
    Dense,
    Dropout,
    Flatten,
    MaxPooling2D
)

INPUT_SHAPE = (32, 32, 3)

JulieProst

from kerastuner import HyperModel


class CNNHyperModel(HyperModel):
    def __init__(self, input_shape, num_classes):
        self.input_shape = input_shape
        self.num_classes = num_classes

    def build(self, hp):
        model = keras.Sequential()

JulieProst

from kerastuner.tuners import RandomSearch

NUM_CLASSES = 10  # cifar10 number of classes
INPUT_SHAPE = (32, 32, 3)  # cifar10 images input shape

hypermodel = CNNHyperModel(input_shape=INPUT_SHAPE, num_classes=NUM_CLASSES)

tuner = RandomSearch(
    hypermodel,
    objective='val_accuracy',

JulieProst

from kerastuner.tuners import Hyperband


tuner = Hyperband(
    hypermodel,
    max_epochs=HYPERBAND_MAX_EPOCHS,
    objective='val_accuracy',
    seed=SEED,
    executions_per_trial=EXECUTION_PER_TRIAL,
    directory='hyperband',

JulieProst

filters=hp.Choice(
    'num_filters',
    values=[32, 64],
    default=64,
),

JulieProst

Dense(
    units=hp.Int(
        'units',
        min_value=32,
        max_value=512,
        step=32,
        default=128
    ),
    activation=hp.Choice(
        'dense_activation',

JulieProst

model.compile(
    optimizer=keras.optimizers.Adam(1e-3),
    loss='sparse_categorical_crossentropy',
    metrics=['accuracy']
)

JulieProst

hp.Float(
    'learning_rate',
    min_value=1e-5,
    max_value=1e-2,
    sampling='LOG',
    default=1e-3
)

JulieProst

# Show a summary of the search
tuner.results_summary()

# Retrieve the best model.
best_model = tuner.get_best_models(num_models=1)[0]

# Evaluate the best model.
loss, accuracy = best_model.evaluate(x_test, y_test)