calorimeter = [
Input(shape=sizes[:2] + [1]),
Input(shape=sizes[2:4] + [1]),
Input(shape=sizes[4:] + [1])
]
input_energy = Input(shape=(1, ))
features = []
energies = []
for l in range(3):
# build features per layer of calorimeter
features.append(
build_discriminator(image=calorimeter[l],
mbd=True,
sparsity=True,
sparsity_mbd=True))
energies.append(calculate_energy(calorimeter[l]))
features = concatenate(features)
# This is a (None, 3) tensor with the individual energy per layer
energies = concatenate(energies)
# calculate the total energy across all rows
total_energy = Lambda(lambda x: K.reshape(K.sum(x, axis=-1), (-1, 1)),
name='total_energy')(energies)
# construct MBD on the raw energies
nb_features = 10
logger.info('Building discriminator')
calorimeter = [Input(shape=sizes[:2] + [1]),
Input(shape=sizes[2:4] + [1]),
Input(shape=sizes[4:] + [1])]
input_energy = Input(shape=(1, ))
features = []
energies = []
for l in range(3):
# build features per layer of calorimeter
features.append(build_discriminator(
image=calorimeter[l],
mbd=True,
sparsity=True,
sparsity_mbd=True
))
energies.append(calculate_energy(calorimeter[l]))
features = concatenate(features)
# This is a (None, 3) tensor with the individual energy per layer
energies = concatenate(energies)
# calculate the total energy across all rows
total_energy = Lambda(
lambda x: K.reshape(K.sum(x, axis=-1), (-1, 1)),
name='total_energy'
)(energies)