def test_circuit_init(self):
"""Test initialization from a circuit."""
circuit, target = self.simple_circuit_no_measure()
op = Choi(circuit)
target = Choi(target)
self.assertEqual(op, target)
def test_multiply_except(self):
"""Test multiply method raises exceptions."""
chan = Choi(self.choiI)
self.assertRaises(QiskitError, chan.multiply, 's')
self.assertRaises(QiskitError, chan.multiply, chan)
def test_negate(self):
"""Test negate method"""
chan = Choi(self.choiI)
targ = Choi(-1 * self.choiI)
self.assertEqual(-chan, targ)
def test_add_except(self):
"""Test add method raises exceptions."""
chan1 = Choi(self.choiI)
chan2 = Choi(np.eye(8))
self.assertRaises(QiskitError, chan1.add, chan2)
self.assertRaises(QiskitError, chan1.add, 5)
def test_subtract_except(self):
"""Test subtract method raises exceptions."""
chan1 = Choi(self.choiI)
chan2 = Choi(np.eye(8))
self.assertRaises(QiskitError, chan1.subtract, chan2)
self.assertRaises(QiskitError, chan1.subtract, 5)
def test_compose_front(self):
"""Test front compose method."""
# UnitaryChannel evolution
chan1 = Choi(self.choiX)
chan2 = Choi(self.choiY)
chan = chan1.compose(chan2, front=True)
targ = Choi(self.choiZ)
self.assertEqual(chan, targ)
# 50% depolarizing channel
chan1 = Choi(self.depol_choi(0.5))
chan = chan1.compose(chan1, front=True)
targ = Choi(self.depol_choi(0.75))
self.assertEqual(chan, targ)
# Measure and rotation
Zp, Zm = np.diag([1, 0]), np.diag([0, 1])
Xp, Xm = np.array([[1, 1], [1, 1]]) / 2, np.array([[1, -1], [-1, 1]
]) / 2
chan1 = Choi(np.kron(Zp, Xp) + np.kron(Zm, Xm))
chan2 = Choi(self.choiX)
# X-gate second does nothing
chan = chan2.compose(chan1, front=True)
targ = Choi(np.kron(Zp, Xp) + np.kron(Zm, Xm))
self.assertEqual(chan, targ)
# X-gate first swaps Z states
chan = chan1.compose(chan2, front=True)
targ = Choi(np.kron(Zm, Xp) + np.kron(Zp, Xm))
self.assertEqual(chan, targ)
# Compose different dimensions
chan1 = Choi(np.eye(8) / 4, input_dims=2, output_dims=4)
chan2 = Choi(np.eye(8) / 2, input_dims=4, output_dims=2)
chan = chan1.compose(chan2, front=True)
self.assertEqual(chan.dim, (4, 4))
chan = chan2.compose(chan1, front=True)
self.assertEqual(chan.dim, (2, 2))
def test_power_except(self):
"""Test power method raises exceptions."""
chan = Choi(self.depol_choi(1))
# Non-integer power raises error
self.assertRaises(QiskitError, chan.power, 0.5)
def test_is_cptp(self):
"""Test is_cptp method."""
self.assertTrue(Choi(self.depol_choi(0.25)).is_cptp())
# Non-CPTP should return false
self.assertFalse(
Choi(1.25 * self.choiI - 0.25 * self.depol_choi(1)).is_cptp())
def test_compose_except(self):
"""Test compose different dimension exception"""
self.assertRaises(QiskitError,
Choi(np.eye(4)).compose, Choi(np.eye(8)))
self.assertRaises(QiskitError, Choi(np.eye(4)).compose, 2)
def thermal_relaxation_error(t1, t2, time, excited_state_population=0):
"""
Single-qubit thermal relaxation quantum error channel.
Args:
t1 (double): the T_1 relaxation time constant.
t2 (double): the T_2 relaxation time constant.
time (double): the gate time for relaxation error.
excited_state_population (double): the population of |1> state at
equilibrium (default: 0).
Returns:
QuantumError: a quantum error object for a noise model.
Raises:
NoiseError: If noise parameters are invalid.
Additional information:
For parameters to be valid T_2 <= 2 * T_1.
If T_2 <= T_1 the error can be expressed as a mixed reset and unitary
error channel.
If T_1 < T_2 <= 2 * T_1 the error must be expressed as a general
non-unitary Kraus error channel.
"""
if excited_state_population < 0:
raise NoiseError("Invalid excited state population "
"({} < 0).".format(excited_state_population))
if excited_state_population > 1:
raise NoiseError("Invalid excited state population "
"({} > 1).".format(excited_state_population))
if time < 0:
raise NoiseError("Invalid gate_time ({} < 0)".format(time))
if t1 <= 0:
raise NoiseError("Invalid T_1 relaxation time parameter: T_1 <= 0.")
if t2 <= 0:
raise NoiseError("Invalid T_2 relaxation time parameter: T_2 <= 0.")
if t2 - 2 * t1 > 0:
raise NoiseError(
"Invalid T_2 relaxation time parameter: T_2 greater than 2 * T_1.")
# T1 relaxation rate
if t1 == np.inf:
rate1 = 0
p_reset = 0
else:
rate1 = 1 / t1
p_reset = 1 - np.exp(-time * rate1)
# T2 dephasing rate
if t2 == np.inf:
rate2 = 0
exp_t2 = 1
else:
rate2 = 1 / t2
exp_t2 = np.exp(-time * rate2)
# Qubit state equilibrium probabilities
p0 = 1 - excited_state_population
p1 = excited_state_population
if t2 > t1:
# If T_2 > T_1 we must express this as a Kraus channel
# We start with the Choi-matrix representation:
chan = Choi(
np.array([[1 - p1 * p_reset, 0, 0, exp_t2],
[0, p1 * p_reset, 0, 0], [0, 0, p0 * p_reset, 0],
[exp_t2, 0, 0, 1 - p0 * p_reset]]))
return QuantumError(Kraus(chan))
else:
# If T_2 < T_1 we can express this channel as a probabilistic
# mixture of reset operations and unitary errors:
circuits = [[{
'name': 'id',
'qubits': [0]
}], [{
'name': 'z',
'qubits': [0]
}], [{
'name': 'reset',
'qubits': [0]
}], [{
'name': 'reset',
'qubits': [0]
}, {
'name': 'x',
'qubits': [0]
}]]
# Probability
p_reset0 = p_reset * p0
p_reset1 = p_reset * p1
p_z = (1 - p_reset) * (1 - np.exp(-time * (rate2 - rate1))) / 2
p_identity = 1 - p_z - p_reset0 - p_reset1
probabilities = [p_identity, p_z, p_reset0, p_reset1]
return QuantumError(zip(circuits, probabilities))
def test_equal(self):
"""Test __eq__ method"""
mat = self.rand_matrix(4, 4)
self.assertEqual(Choi(mat), Choi(mat))
def test_evolve(self):
"""Test evolve method."""
input_psi = [0, 1]
input_rho = [[0, 0], [0, 1]]
# Identity channel
chan = Choi(self.choiI)
target_rho = np.array([[0, 0], [0, 1]])
self.assertAllClose(chan._evolve(input_psi), target_rho)
self.assertAllClose(chan._evolve(np.array(input_psi)), target_rho)
self.assertAllClose(chan._evolve(input_rho), target_rho)
self.assertAllClose(chan._evolve(np.array(input_rho)), target_rho)
# Hadamard channel
chan = Choi(self.choiH)
target_rho = np.array([[1, -1], [-1, 1]]) / 2
self.assertAllClose(chan._evolve(input_psi), target_rho)
self.assertAllClose(chan._evolve(np.array(input_psi)), target_rho)
self.assertAllClose(chan._evolve(input_rho), target_rho)
self.assertAllClose(chan._evolve(np.array(input_rho)), target_rho)
# Completely depolarizing channel
chan = Choi(self.depol_choi(1))
target_rho = np.eye(2) / 2
self.assertAllClose(chan._evolve(input_psi), target_rho)
self.assertAllClose(chan._evolve(np.array(input_psi)), target_rho)
self.assertAllClose(chan._evolve(input_rho), target_rho)
self.assertAllClose(chan._evolve(np.array(input_rho)), target_rho)
def diamond_norm(choi, **kwargs):
r"""Return the diamond norm of the input quantum channel object.
This function computes the completely-bounded trace-norm (often
referred to as the diamond-norm) of the input quantum channel object
using the semidefinite-program from reference [1].
Args:
choi(Choi or QuantumChannel): a quantum channel object or
Choi-matrix array.
kwargs: optional arguments to pass to CVXPY solver.
Returns:
float: The completely-bounded trace norm
:math:`\|\mathcal{E}\|_{\diamond}`.
Raises:
QiskitError: if CVXPY package cannot be found.
Additional Information:
The input to this function is typically *not* a CPTP quantum
channel, but rather the *difference* between two quantum channels
:math:`\|\Delta\mathcal{E}\|_\diamond` where
:math:`\Delta\mathcal{E} = \mathcal{E}_1 - \mathcal{E}_2`.
Reference:
J. Watrous. "Simpler semidefinite programs for completely bounded
norms", arXiv:1207.5726 [quant-ph] (2012).
.. note::
This function requires the optional CVXPY package to be installed.
Any additional kwargs will be passed to the ``cvxpy.solve``
function. See the CVXPY documentation for information on available
SDP solvers.
"""
_cvxpy_check('`diamond_norm`') # Check CVXPY is installed
if not isinstance(choi, Choi):
choi = Choi(choi)
def cvx_bmat(mat_r, mat_i):
"""Block matrix for embedding complex matrix in reals"""
return cvxpy.bmat([[mat_r, -mat_i], [mat_i, mat_r]])
# Dimension of input and output spaces
dim_in = choi._input_dim
dim_out = choi._output_dim
size = dim_in * dim_out
# SDP Variables to convert to real valued problem
r0_r = cvxpy.Variable((dim_in, dim_in))
r0_i = cvxpy.Variable((dim_in, dim_in))
r0 = cvx_bmat(r0_r, r0_i)
r1_r = cvxpy.Variable((dim_in, dim_in))
r1_i = cvxpy.Variable((dim_in, dim_in))
r1 = cvx_bmat(r1_r, r1_i)
x_r = cvxpy.Variable((size, size))
x_i = cvxpy.Variable((size, size))
iden = sparse.eye(dim_out)
# Watrous uses row-vec convention for his Choi matrix while we use
# col-vec. It turns out row-vec convention is requried for CVXPY too
# since the cvxpy.kron function must have a constant as its first argument.
c_r = cvxpy.bmat([[cvxpy.kron(iden, r0_r), x_r],
[x_r.T, cvxpy.kron(iden, r1_r)]])
c_i = cvxpy.bmat([[cvxpy.kron(iden, r0_i), x_i],
[-x_i.T, cvxpy.kron(iden, r1_i)]])
c = cvx_bmat(c_r, c_i)
# Convert col-vec convention Choi-matrix to row-vec convention and
# then take Transpose: Choi_C -> Choi_R.T
choi_rt = np.transpose(
np.reshape(choi.data, (dim_in, dim_out, dim_in, dim_out)),
(3, 2, 1, 0)).reshape(choi.data.shape)
choi_rt_r = choi_rt.real
choi_rt_i = choi_rt.imag
# Constraints
cons = [
r0 >> 0, r0_r == r0_r.T, r0_i == -r0_i.T,
cvxpy.trace(r0_r) == 1, r1 >> 0, r1_r == r1_r.T, r1_i == -r1_i.T,
cvxpy.trace(r1_r) == 1, c >> 0
]
# Objective function
obj = cvxpy.Maximize(
cvxpy.trace(choi_rt_r @ x_r) + cvxpy.trace(choi_rt_i @ x_i))
prob = cvxpy.Problem(obj, cons)
sol = prob.solve(**kwargs)
return sol
def test_sub_qargs(self):
"""Test subtract method with qargs."""
mat = self.rand_matrix(8 ** 2, 8 ** 2)
mat0 = self.rand_matrix(4, 4)
mat1 = self.rand_matrix(4, 4)
op = Choi(mat)
op0 = Choi(mat0)
op1 = Choi(mat1)
op01 = op1.tensor(op0)
eye = Choi(self.choiI)
with self.subTest(msg="qargs=[0]"):
value = op - op0([0])
target = op - eye.tensor(eye).tensor(op0)
self.assertEqual(value, target)
with self.subTest(msg="qargs=[1]"):
value = op - op0([1])
target = op - eye.tensor(op0).tensor(eye)
self.assertEqual(value, target)
with self.subTest(msg="qargs=[2]"):
value = op - op0([2])
target = op - op0.tensor(eye).tensor(eye)
self.assertEqual(value, target)
with self.subTest(msg="qargs=[0, 1]"):
value = op - op01([0, 1])
target = op - eye.tensor(op1).tensor(op0)
self.assertEqual(value, target)
with self.subTest(msg="qargs=[1, 0]"):
value = op - op01([1, 0])
target = op - eye.tensor(op0).tensor(op1)
self.assertEqual(value, target)
with self.subTest(msg="qargs=[0, 2]"):
value = op - op01([0, 2])
target = op - op1.tensor(eye).tensor(op0)
self.assertEqual(value, target)
with self.subTest(msg="qargs=[2, 0]"):
value = op - op01([2, 0])
target = op - op0.tensor(eye).tensor(op1)
self.assertEqual(value, target)
def process_fidelity(channel, target=None, require_cp=True, require_tp=True):
r"""Return the process fidelity of a noisy quantum channel.
The process fidelity :math:`F_{\text{pro}}(\mathcal{E}, \mathcal{F})`
between two quantum channels :math:`\mathcal{E}, \mathcal{F}` is given by
.. math::
F_{\text{pro}}(\mathcal{E}, \mathcal{F})
= F(\rho_{\mathcal{E}}, \rho_{\mathcal{F}})
where :math:`F` is the :func:`~qiskit.quantum_info.state_fidelity`,
:math:`\rho_{\mathcal{E}} = \Lambda_{\mathcal{E}} / d` is the
normalized :class:`~qiskit.quantum_info.Choi` matrix for the channel
:math:`\mathcal{E}`, and :math:`d` is the input dimension of
:math:`\mathcal{E}`.
When the target channel is unitary this is equivalent to
.. math::
F_{\text{pro}}(\mathcal{E}, U)
= \frac{Tr[S_U^\dagger S_{\mathcal{E}}]}{d^2}
where :math:`S_{\mathcal{E}}, S_{U}` are the
:class:`~qiskit.quantum_info.SuperOp` matrices for the *input* quantum
channel :math:`\mathcal{E}` and *target* unitary :math:`U` respectively,
and :math:`d` is the input dimension of the channel.
Args:
channel (Operator or QuantumChannel): input quantum channel.
target (Operator or QuantumChannel or None): target quantum channel.
If `None` target is the identity operator [Default: None].
require_cp (bool): check if input and target channels are
completely-positive and if non-CP log warning
containing negative eigenvalues of Choi-matrix
[Default: True].
require_tp (bool): check if input and target channels are
trace-preserving and if non-TP log warning
containing negative eigenvalues of partial
Choi-matrix :math:`Tr_{\mbox{out}}[\mathcal{E}] - I`
[Default: True].
Returns:
float: The process fidelity :math:`F_{\text{pro}}`.
Raises:
QiskitError: if the channel and target do not have the same dimensions.
"""
# Format inputs
channel = _input_formatter(channel, SuperOp, "process_fidelity", "channel")
target = _input_formatter(target, Operator, "process_fidelity", "target")
if target:
# Validate dimensions
if channel.dim != target.dim:
raise QiskitError(
"Input quantum channel and target unitary must have the same "
"dimensions ({} != {}).".format(channel.dim, target.dim))
# Validate complete-positivity and trace-preserving
for label, chan in [("Input", channel), ("Target", target)]:
if chan is not None and require_cp:
cp_cond = _cp_condition(chan)
neg = cp_cond < -1 * chan.atol
if np.any(neg):
logger.warning(
"%s channel is not CP. Choi-matrix has negative eigenvalues: %s",
label,
cp_cond[neg],
)
if chan is not None and require_tp:
tp_cond = _tp_condition(chan)
non_zero = np.logical_not(
np.isclose(tp_cond, 0, atol=chan.atol, rtol=chan.rtol))
if np.any(non_zero):
logger.warning(
"%s channel is not TP. Tr_2[Choi] - I has non-zero eigenvalues: %s",
label,
tp_cond[non_zero],
)
if isinstance(target, Operator):
# Compute fidelity with unitary target by applying the inverse
# to channel and computing fidelity with the identity
channel = channel.compose(target.adjoint())
target = None
input_dim, _ = channel.dim
if target is None:
# Compute process fidelity with identity channel
if isinstance(channel, Operator):
# |Tr[U]/dim| ** 2
fid = np.abs(np.trace(channel.data) / input_dim)**2
else:
# Tr[S] / (dim ** 2)
fid = np.trace(SuperOp(channel).data) / (input_dim**2)
return float(np.real(fid))
# For comparing two non-unitary channels we compute the state fidelity of
# the normalized Choi-matrices. This is equivalent to the previous definition
# when the target is a unitary channel.
state1 = DensityMatrix(Choi(channel).data / input_dim)
state2 = DensityMatrix(Choi(target).data / input_dim)
return state_fidelity(state1, state2, validate=False)
def process_fidelity(channel, target=None, require_cp=True, require_tp=False):
r"""Return the process fidelity of a noisy quantum channel.
The process fidelity :math:`F_{\text{pro}}(\mathcal{E}, \methcal{F})`
between two quantum channels :math:`\mathcal{E}, \mathcal{F}` is given by
.. math:
F_{\text{pro}}(\mathcal{E}, \mathcal{F})
= F(\rho_{\mathcal{E}}, \rho_{\mathcal{F}})
where :math:`F` is the :func:`~qiskit.quantum_info.state_fidelity`,
:math:`\rho_{\mathcal{E}} = \Lambda_{\mathcal{E}} / d` is the
normalized :class:`~qiskit.quantum_info.Choi` matrix for the channel
:math:`\mathcal{E}`, and :math:`d` is the input dimension of
:math:`\mathcal{E}`.
When the target channel is unitary this is equivalent to
.. math::
F_{\text{pro}}(\mathcal{E}, U)
= \frac{Tr[S_U^\dagger S_{\mathcal{E}}]}{d^2}
where :math:`S_{\mathcal{E}}, S_{U}` are the
:class:`~qiskit.quantum_info.SuperOp` matrices for the *input* quantum
channel :math:`\mathcal{E}` and *target* unitary :math:`U` respectively,
and :math:`d` is the input dimension of the channel.
Args:
channel (Operator or QuantumChannel): input quantum channel.
target (Operator or QuantumChannel or None): target quantum channel.
If `None` target is the identity operator [Default: None].
require_cp (bool): require channel to be completely-positive
[Default: True].
require_tp (bool): require channel to be trace-preserving
[Default: False].
Returns:
float: The process fidelity :math:`F_{\text{pro}}`.
Raises:
QiskitError: if the channel and target do not have the same dimensions.
QiskitError: if the channel and target are not completely-positive
(with ``require_cp=True``) or not trace-preserving
(with ``require_tp=True``).
"""
# Format inputs
channel = _input_formatter(channel, SuperOp, 'process_fidelity', 'channel')
target = _input_formatter(target, Operator, 'process_fidelity', 'target')
if target:
# Validate dimensions
if channel.dim != target.dim:
raise QiskitError(
'Input quantum channel and target unitary must have the same '
'dimensions ({} != {}).'.format(channel.dim, target.dim))
# Validate complete-positivity and trace-preserving
for label, chan in [('Input', channel), ('Target', target)]:
if isinstance(chan, Operator) and (require_cp or require_tp):
is_unitary = chan.is_unitary()
# Validate as unitary
if require_cp and not is_unitary:
raise QiskitError(
'{} channel is not completely-positive'.format(label))
if require_tp and not is_unitary:
raise QiskitError(
'{} channel is not trace-preserving'.format(label))
elif chan is not None:
# Validate as QuantumChannel
if require_cp and not chan.is_cp():
raise QiskitError(
'{} channel is not completely-positive'.format(label))
if require_tp and not chan.is_tp():
raise QiskitError(
'{} channel is not trace-preserving'.format(label))
if isinstance(target, Operator):
# Compute fidelity with unitary target by applying the inverse
# to channel and computing fidelity with the identity
channel = channel @ target.adjoint()
target = None
input_dim, _ = channel.dim
if target is None:
# Compute process fidelity with identity channel
if isinstance(channel, Operator):
# |Tr[U]/dim| ** 2
fid = np.abs(np.trace(channel.data) / input_dim)**2
else:
# Tr[S] / (dim ** 2)
fid = np.trace(SuperOp(channel).data) / (input_dim**2)
return float(np.real(fid))
# For comparing two non-unitary channels we compute the state fidelity of
# the normalized Choi-matrices. This is equivalent to the previous definition
# when the target is a unitary channel.
state1 = DensityMatrix(Choi(channel).data / input_dim)
state2 = DensityMatrix(Choi(target).data / input_dim)
return state_fidelity(state1, state2, validate=False)
