def test_clifford_state_str():
(q0, q1) = (cirq.LineQubit(0), cirq.LineQubit(1))
state = cirq.CliffordState(qubit_map={q0: 0, q1: 1})
assert str(state) == "|00⟩"
def test_clifford_state_state_vector():
(q0, q1) = (cirq.LineQubit(0), cirq.LineQubit(1))
state = cirq.CliffordState(qubit_map={q0: 0, q1: 1})
np.testing.assert_equal(state.state_vector(),
[1.0 + 0.0j, 0.0 + 0.0j, 0.0 + 0.0j, 0.0 + 0.0j])
def direct_fidelity_estimation(circuit: cirq.Circuit, qubits: List[cirq.Qid],
sampler: cirq.Sampler,
n_measured_operators: Optional[int],
samples_per_term: int):
"""
Implementation of direct fidelity estimation, as per 'Direct Fidelity
Estimation from Few Pauli Measurements' https://arxiv.org/abs/1104.4695 and
'Practical characterization of quantum devices without tomography'
https://arxiv.org/abs/1104.3835.
Args:
circuit: The circuit to run the simulation on.
qubits: The list of qubits.
sampler: Either a noisy simulator or an engine.
n_measured_operators: The total number of Pauli measurements, or None to
explore each Pauli state once.
samples_per_term: if set to 0, we use the 'sampler' parameter above as
a noise (must be of type cirq.DensityMatrixSimulator) and
simulate noise in the circuit. If greater than 0, we instead use the
'sampler' parameter directly to estimate the characteristic
function.
Returns:
The estimated fidelity and a log of the run.
"""
# n_measured_operators is upper-case N in https://arxiv.org/abs/1104.3835
# Number of qubits, lower-case n in https://arxiv.org/abs/1104.3835
n_qubits = len(qubits)
clifford_circuit = True
clifford_state: Optional[cirq.CliffordState] = None
try:
clifford_state = cirq.CliffordState(
qubit_map={qubits[i]: i
for i in range(len(qubits))})
for gate in circuit.all_operations():
clifford_state.apply_unitary(gate)
except ValueError:
clifford_circuit = False
# Computes for every \hat{P_i} of https://arxiv.org/abs/1104.3835
# estimate rho_i and Pr(i). We then collect tuples (rho_i, Pr(i), \hat{Pi})
# inside the variable 'pauli_traces'.
if clifford_circuit:
assert clifford_state is not None
pauli_traces = _estimate_pauli_traces_clifford(
n_qubits, cast(cirq.CliffordState, clifford_state),
n_measured_operators)
else:
pauli_traces = _estimate_pauli_traces_general(qubits, circuit,
n_measured_operators)
p = np.asarray([x.Pr_i for x in pauli_traces])
if n_measured_operators is None:
# Since we enumerate all the possible traces, the probs should add to 1.
assert np.isclose(np.sum(p), 1.0, atol=1e-6)
p /= np.sum(p)
fidelity = 0.0
if samples_per_term == 0:
# sigma in https://arxiv.org/abs/1104.3835
if not isinstance(sampler, cirq.DensityMatrixSimulator):
raise TypeError('sampler is not a cirq.DensityMatrixSimulator '
'but samples_per_term is zero.')
noisy_simulator = cast(cirq.DensityMatrixSimulator, sampler)
noisy_density_matrix = cast(
cirq.DensityMatrixTrialResult,
noisy_simulator.simulate(circuit)).final_density_matrix
if clifford_circuit and n_measured_operators is None:
# In case the circuit is Clifford and we compute an exhaustive list of
# Pauli traces, instead of sampling we can simply enumerate them because
# they all have the same probability.
measured_pauli_traces = pauli_traces
else:
# Otherwise, randomly sample as per probability.
measured_pauli_traces = np.random.choice(pauli_traces,
size=len(pauli_traces),
p=p)
trial_results: List[TrialResult] = []
for pauli_trace in measured_pauli_traces:
measure_pauli_string: cirq.PauliString = pauli_trace.P_i
rho_i = pauli_trace.rho_i
if samples_per_term > 0:
sigma_i = asyncio.get_event_loop().run_until_complete(
estimate_characteristic_function(circuit, measure_pauli_string,
qubits, sampler,
samples_per_term))
else:
sigma_i, _ = compute_characteristic_function(
circuit, measure_pauli_string, qubits, noisy_density_matrix)
trial_results.append(
TrialResult(pauli_trace=pauli_trace, sigma_i=sigma_i))
fidelity += sigma_i / rho_i
estimated_fidelity = fidelity / len(pauli_traces)
std_dev_estimate: Optional[float]
std_dev_bound: Optional[float]
if clifford_circuit:
std_dev_estimate, std_dev_bound = _estimate_std_devs_clifford(
estimated_fidelity, len(measured_pauli_traces))
else:
std_dev_estimate, std_dev_bound = None, None
dfe_intermediate_result = DFEIntermediateResult(
clifford_state=clifford_state,
pauli_traces=pauli_traces,
trial_results=trial_results,
std_dev_estimate=std_dev_estimate,
std_dev_bound=std_dev_bound)
return estimated_fidelity, dfe_intermediate_result
def test_clifford_state_initial_state():
q0 = cirq.LineQubit(0)
with pytest.raises(ValueError, match='Out of range'):
_ = cirq.CliffordState(qubit_map={q0: 0}, initial_state=2)
state = cirq.CliffordState(qubit_map={q0: 0}, initial_state=1)
np.testing.assert_allclose(state.state_vector(), [0, 1])
def test_stabilizerStateChForm_H():
(q0, q1) = (cirq.LineQubit(0), cirq.LineQubit(1))
state = cirq.CliffordState(qubit_map={q0: 0, q1: 1})
with pytest.raises(ValueError, match="|y> is equal to |z>"):
state.ch_form._H_decompose(0, 1, 1, 0)
def test_clifford_stabilizerStateChForm_repr():
(q0, q1) = (cirq.LineQubit(0), cirq.LineQubit(1))
state = cirq.CliffordState(qubit_map={q0: 0, q1: 1})
assert repr(state) == 'StabilizerStateChForm(num_qubits=2)'
def test_clifford_tableau_repr():
(q0, q1) = (cirq.LineQubit(0), cirq.LineQubit(1))
state = cirq.CliffordState(qubit_map={q0: 0, q1: 1})
f = cirq.DensePauliString
assert (repr(state.tableau) == "stabilizers: [{!r}, {!r}]".format(
f("ZI"), f("IZ")))
def test_clifford_state_wave_function():
(q0, q1) = (cirq.LineQubit(0), cirq.LineQubit(1))
state = cirq.CliffordState(qubit_map={q0: 0, q1: 1})
np.testing.assert_equal(state.wave_function(),
[1. + 0.j, 0. + 0.j, 0. + 0.j, 0. + 0.j])
def direct_fidelity_estimation(circuit: cirq.Circuit, qubits: List[cirq.Qid],
sampler: cirq.Sampler, n_trials: int,
n_clifford_trials: int, samples_per_term: int):
"""
Implementation of direct fidelity estimation, as per 'Direct Fidelity
Estimation from Few Pauli Measurements' https://arxiv.org/abs/1104.4695 and
'Practical characterization of quantum devices without tomography'
https://arxiv.org/abs/1104.3835.
Args:
circuit: The circuit to run the simulation on.
qubits: The list of qubits.
sampler: Either a noisy simulator or an engine.
n_trial: The total number of Pauli measurements.
n_clifford_trials: In case the circuit is Clifford, we specify the
number of trials to estimate the noise-free pauli traces.
samples_per_term: if set to 0, we use the 'sampler' parameter above as
a noise (must be of type cirq.DensityMatrixSimulator) and
simulate noise in the circuit. If greater than 0, we instead use the
'sampler' parameter directly to estimate the characteristic
function.
Returns:
The estimated fidelity.
"""
# n_trials is upper-case N in https://arxiv.org/abs/1104.3835
# Number of qubits, lower-case n in https://arxiv.org/abs/1104.3835
n_qubits = len(qubits)
d = 2**n_qubits
clifford_circuit = True
clifford_state: Optional[cirq.CliffordState] = None
try:
clifford_state = cirq.CliffordState(
qubit_map={qubits[i]: i
for i in range(len(qubits))})
for gate in circuit.all_operations():
clifford_state.apply_unitary(gate)
except ValueError:
clifford_circuit = False
# Computes for every \hat{P_i} of https://arxiv.org/abs/1104.3835
# estimate rho_i and Pr(i). We then collect tuples (rho_i, Pr(i), \hat{Pi})
# inside the variable 'pauli_traces'.
if clifford_circuit:
print('Circuit is Clifford')
assert clifford_state is not None
pauli_traces = _estimate_pauli_traces_clifford(
n_qubits, cast(cirq.CliffordState, clifford_state),
n_clifford_trials)
else:
print('Circuit is not Clifford')
pauli_traces = _estimate_pauli_traces_general(qubits, circuit)
p = np.asarray([x['Pr_i'] for x in pauli_traces])
if not clifford_circuit:
# For Clifford circuits, we do a Monte Carlo simulations, and thus there
# is no guarantee that it adds up to 1.0 (but it should to the limit).
assert np.isclose(np.sum(p), 1.0, atol=1e-6)
# The package np.random.choice() is quite sensitive to probabilities not
# summing up to 1.0. Even an absolute difference below 1e-6 (as checked just
# above) does bother it, so we re-normalize the probs.
p /= np.sum(p)
fidelity = 0.0
if samples_per_term == 0:
# sigma in https://arxiv.org/abs/1104.3835
if not isinstance(sampler, cirq.DensityMatrixSimulator):
raise TypeError('sampler is not a cirq.DensityMatrixSimulator '
'but samples_per_term is zero.')
noisy_simulator = cast(cirq.DensityMatrixSimulator, sampler)
noisy_density_matrix = cast(
cirq.DensityMatrixTrialResult,
noisy_simulator.simulate(circuit)).final_density_matrix
for _ in range(n_trials):
# Randomly sample as per probability.
i = np.random.choice(len(pauli_traces), p=p)
Pr_i = pauli_traces[i]['Pr_i']
measure_pauli_string: cirq.PauliString = pauli_traces[i]['P_i']
rho_i = pauli_traces[i]['rho_i']
if samples_per_term > 0:
sigma_i = asyncio.get_event_loop().run_until_complete(
estimate_characteristic_function(circuit, measure_pauli_string,
qubits, sampler,
samples_per_term))
else:
sigma_i, _ = compute_characteristic_function(
circuit, measure_pauli_string, qubits, noisy_density_matrix)
fidelity += Pr_i * sigma_i / rho_i
return fidelity / n_trials * d
def test_valid_apply_measurement():
q0 = cirq.LineQubit(0)
state = cirq.CliffordState(qubit_map={q0: 0}, initial_state=1)
measurements = {}
_ = state.apply_measurement(cirq.measure(q0), measurements, np.random.RandomState())
assert measurements == {'0': [1]}
def test_clifford_tableau_repr():
(q0, q1) = (cirq.LineQubit(0), cirq.LineQubit(1))
state = cirq.CliffordState(qubit_map={q0: 0, q1: 1})
assert (repr(state.tableau) == "stabilizers: [Z0, Z1]")
