def test_tensor_kraus_maps():
damping = damping_kraus_map()
k1, k2, k3, k4 = tensor_kraus_maps(damping, damping)
assert k1.shape == (4, 4)
assert k2.shape == (4, 4)
assert k3.shape == (4, 4)
assert k4.shape == (4, 4)
np.testing.assert_allclose(k1[-1, -1], 1 - 0.1)
def test_pauli_kraus_map():
probabilities = [.1, .2, .3, .4]
k1, k2, k3, k4 = pauli_kraus_map(probabilities)
assert np.allclose(k1, np.sqrt(.1) * np.eye(2), atol=1 * 10**-8)
assert np.allclose(k2,
np.sqrt(.2) * np.array([[0, 1.], [1., 0]]),
atol=1 * 10**-8)
assert np.allclose(k3,
np.sqrt(.3) * np.array([[0, -1.j], [1.j, 0]]),
atol=1 * 10**-8)
assert np.allclose(k4,
np.sqrt(.4) * np.array([[1, 0], [0, -1]]),
atol=1 * 10**-8)
two_q_pauli_kmaps = pauli_kraus_map(
np.kron(probabilities, list(reversed(probabilities))))
q1_pauli_kmaps = [k1, k2, k3, k4]
q2_pauli_kmaps = pauli_kraus_map(list(reversed(probabilities)))
tensor_kmaps = tensor_kraus_maps(q1_pauli_kmaps, q2_pauli_kmaps)
assert np.allclose(two_q_pauli_kmaps, tensor_kmaps)
def _modified_decoherence_noise_model(
gates: Sequence[Gate],
T1: Union[Dict[int, float], float] = 30e-6,
T2: Union[Dict[int, float], float] = 30e-6,
gate_time_1q: float = 50e-9,
gate_time_2q: float = 150e-09,
ro_fidelity: Union[Dict[int, float], float] = 0.95,
) -> NoiseModel:
"""
The default noise parameters
- T1 = 30 us
- T2 = 30 us
- 1q gate time = 50 ns
- 2q gate time = 150 ns
are currently typical for near-term devices.
This function will define new gates and add Kraus noise to these gates. It will translate
the input program to use the noisy version of the gates.
:param gates: The gates to provide the noise model for.
:param T1: The T1 amplitude damping time either globally or in a
dictionary indexed by qubit id. By default, this is 30 us.
:param T2: The T2 dephasing time either globally or in a
dictionary indexed by qubit id. By default, this is also 30 us.
:param gate_time_1q: The duration of the one-qubit gates, namely RX(+pi/2) and RX(-pi/2).
By default, this is 50 ns.
:param gate_time_2q: The duration of the two-qubit gates, namely CZ.
By default, this is 150 ns.
:param ro_fidelity: The readout assignment fidelity
:math:`F = (p(0|0) + p(1|1))/2` either globally or in a dictionary indexed by qubit id.
:return: A NoiseModel with the appropriate Kraus operators defined.
"""
all_qubits = set(sum(([t.index for t in g.qubits] for g in gates), []))
if isinstance(T1, dict):
all_qubits.update(T1.keys())
if isinstance(T2, dict):
all_qubits.update(T2.keys())
if isinstance(ro_fidelity, dict):
all_qubits.update(ro_fidelity.keys())
if not isinstance(T1, dict):
T1 = {q: T1 for q in all_qubits}
if not isinstance(T2, dict):
T2 = {q: T2 for q in all_qubits}
if not isinstance(ro_fidelity, dict):
ro_fidelity = {q: ro_fidelity for q in all_qubits}
kraus_maps = []
for g in gates:
targets = tuple(t.index for t in g.qubits)
key = (g.name, tuple(g.params))
if g.name in NO_NOISE:
if not g.dd:
g.gate_time = gate_time_1q
continue
matrix, _ = get_modified_noisy_gate(g.name, g.params)
if len(targets) == 1:
if g.gate_time == None:
g.gate_time = gate_time_1q
noisy_I = damping_after_dephasing(T1.get(targets[0], INFINITY),
T2.get(targets[0], INFINITY),
g.gate_time)
else:
if len(targets) != 2:
raise ValueError(
"Noisy gates on more than 2Q not currently supported")
if g.gate_time == None:
g.gate_time = gate_time_2q
# note this ordering of the tensor factors is necessary due to how the QVM orders
# the wavefunction basis
noisy_I = tensor_kraus_maps(
damping_after_dephasing(T1.get(targets[1], INFINITY),
T2.get(targets[1], INFINITY),
g.gate_time),
damping_after_dephasing(T1.get(targets[0], INFINITY),
T2.get(targets[0], INFINITY),
g.gate_time))
kraus_maps.append(
KrausModel(g.name, tuple(g.params), targets,
combine_kraus_maps(noisy_I, [matrix]), 1.0))
aprobs = {}
for q, f_ro in ro_fidelity.items():
aprobs[q] = np.array([[f_ro, 1. - f_ro], [1. - f_ro, f_ro]])
return NoiseModel(kraus_maps, aprobs)
