def main():
# GENERAL SETTINGS
num_system_qubits = 3
num_ancillae = 1
num_qubits = num_system_qubits + num_ancillae
n_attempts = 20
logging.info('We are going for 20 attempts.')
# SET TARGET GATE AND INTERACTIONS
target_gate = qutip.fredkin()
# interactions = 'all'
fredkin_model = make_XX_model(num_qubits)
# HYPERPARAMETERS
training_dataset_size = 200
test_dataset_size = 100
n_epochs = 300
batch_size = 2
sgd_method = 'adadelta'
learning_rate = 1
decay_rate = 0.1
initial_values = 'random'
logging.info('Random initial values')
# TAKE CARE NOT TO OVERWRITE PREVIOUSLY SAVED FILES
prefix = os.path.join(OUTPUT_DIR, OUTPUT_FILES_NAME)
ext = '.pickle'
files = glob.glob(prefix + '*')
pre_idx = 0
if files:
pre_idx = max([int(f[len(prefix):-len(ext)]) for f in files])
# STARTING MESSAGES
for idx in range(n_attempts):
logging.info('Starting training no.{}'.format(str(idx + 1)))
model = qn.model.QubitNetworkGateModel(
sympy_expr=fredkin_model,
initial_values=initial_values,
num_system_qubits=num_system_qubits)
optimizer = qn.Optimizer.Optimizer(
net=model,
learning_rate=learning_rate,
decay_rate=decay_rate,
n_epochs=n_epochs,
batch_size=batch_size,
target_gate=target_gate,
training_dataset_size=training_dataset_size,
test_dataset_size=test_dataset_size,
sgd_method=sgd_method,
headless=True)
optimizer.run()
newpath = os.path.join(
OUTPUT_DIR, OUTPUT_FILES_NAME + str(1 + pre_idx + idx) + '.pickle')
optimizer.save_results(newpath)
logging.info('Fidelity obtained: {}'.format(model.average_fidelity()))
def main():
# GENERAL SETTINGS
num_system_qubits = 3
num_ancillae = 0
num_qubits = num_system_qubits + num_ancillae
# the attempts will be equally (when possible) distributed among these
# initial values. Each initial value will be used the same amount of
# times.
attempted_initvalues = [0, 1, 2, 3, 4]
num_attempts_per_initvalue = 10
num_attempts = num_attempts_per_initvalue * len(attempted_initvalues)
logging.info('We are going for a total of {}'
'attempts.'.format(num_attempts))
logging.info('The following initial values will be used:'
'{}.'.format(attempted_initvalues))
# SET TARGET GATE AND INTERACTIONS
target_gate = qutip.fredkin()
# interactions = 'all'
# fredkin_model = make_reduced_fredkin_model(num_qubits)
# HYPERPARAMETERS
training_dataset_size = 200
test_dataset_size = 100
n_epochs = 100
batch_size = 2
sgd_method = 'momentum'
learning_rate = 1
decay_rate = 0.1
# TAKE CARE NOT TO OVERWRITE PREVIOUSLY SAVED FILES
prefix = os.path.join(OUTPUT_DIR, OUTPUT_FILES_NAME)
ext = '.pickle'
files = glob.glob(prefix + '*')
pre_idx = 0
if files:
pre_idx = max([int(f[len(prefix):-len(ext)]) for f in files])
# STARTING MESSAGES
for idx in range(num_attempts):
initial_values = attempted_initvalues[idx //
num_attempts_per_initvalue]
fredkin_model = make_reduced_fredkin_model(num_qubits)
logging.info('Starting training no.{}'.format(str(idx + 1)))
logging.info('Initial values: {}.'.format(initial_values))
model = qn.model.QubitNetworkGateModel(
sympy_expr=fredkin_model,
initial_values=initial_values,
num_system_qubits=num_system_qubits)
optimizer = qn.Optimizer.Optimizer(
net=model,
learning_rate=learning_rate,
decay_rate=decay_rate,
n_epochs=n_epochs,
batch_size=batch_size,
target_gate=target_gate,
training_dataset_size=training_dataset_size,
test_dataset_size=test_dataset_size,
sgd_method=sgd_method,
headless=True)
optimizer.run()
newpath = os.path.join(
OUTPUT_DIR, OUTPUT_FILES_NAME + str(1 + pre_idx + idx) + '.pickle')
optimizer.save_results(newpath)
logging.info('Fidelity obtained: {}'.format(model.average_fidelity()))
import qutip as qt
import numpy as np
"""
fredkin hadamard
ancilla---------O---------H------------Measure
f1---------X----------------------
f2---------X----------------------
"""
up = qt.qstate('u') #qstate down is (1,0), qstate up is (0,1)
down = qt.qstate('d')
ancilla = 1 / np.sqrt(2) * (up + down)
hadamard = qt.tensor(qt.snot(), qt.qeye(2), qt.qeye(2))
fredkin = qt.fredkin() #fredkin swapes f1 and f2 if ancilla = 1
#------------------------------------------------------------------------
#Define Input
f1 = 1 / np.sqrt(2) * (up + down)
f2 = up
#------------------------------------------------------------------------
psi0 = qt.tensor(ancilla, f1, f2)
output = hadamard * fredkin * psi0
measure = qt.tensor(down, f1,
f2).dag() * output #if f1 = f2 the output is (down,f1,f2)
#therefore measure is 1 if f1 = f2
#measure is 0.5 if f1, f2 are orthogonal
print(measure)
def make_reduced_fredkin_model(num_qubits):
import qubit_network.analytical_conditions as ac
fredkin_diagonal = ac.commuting_generator(qutip.fredkin(),
interactions='diagonal')
return fredkin_diagonal
return qt.phasegate(angle)
# 2 qubit gates
CNOT = qt.cnot()
CZ = qt.csign()
Berkeley = qt.berkeley()
SWAP = qt.swap()
iSWAP = qt.iswap()
SQSWAP = qt.sqrtswap()
SQiSWAP = qt.sqrtiswap()
def aSWAP(angle):
return qt.swapalpha(angle)
# 3 qubit gates
Fredkin = qt.fredkin()
Toffoli = qt.toffoli()
# Get unitary axis and angle
def bloch(U):
if isinstance(U, qt.Qobj):
U = U.full()
angle = np.real(2 * np.arccos(np.trace(U) / 2))
sin = np.sin(angle / 2)
eps = 1e-10
if sin < eps:
axis = [0, 0, 1]
else:
nz = np.imag(U[1, 1] - U[0, 0]) / (2 * sin)