Utilities

find_indices(circuit, bits_or_instruction, /)

Find the qubit-indices of the provided bit indices or circuit instructions.

A sequence of bits maps to a list of indices, in the order the bits were given:

>>> from qiskit import QuantumCircuit
>>> from qiskit_addon_slc.utils import find_indices

>>> circuit = QuantumCircuit(3)
>>> find_indices(circuit, [circuit.qubits[2], circuit.qubits[0]])
[2, 0]

A single bit maps to a single index, and an instruction to the indices of the qubits it acts on:

>>> find_indices(circuit, circuit.qubits[1])
1

>>> _ = circuit.cx(2, 0)
>>> find_indices(circuit, circuit.data[0])
[2, 0]

Parameters

• circuit (QuantumCircuit) – the quantum circuit whose qubit indices to find.
• bits_or_instruction (Bit |CircuitInstruction |Sequence[Bit]) – the bits whose indices to find. If this is a CircuitInstruction, the qubits which this instruction acts upon are used.

Returns

The indices of the queried bits in the circuit’s registers. If a single bit object was provided, a single int is returned for its index. Otherwise the return type will be a list[int] whose length equals the number of provided bits.

Raises

TypeError – when an unexpected type of bits_or_instruction gets provided.

Return type

int | list[int]

generate_noise_model_paulis(instructions, coupling_map=None, circuit=None)

Generate the 1- and 2-weight Pauli terms for each of the unique 2q layer boxes provided.

Parameters

• instructions (list[CircuitInstruction]) – the output of find_unique_box_instructions(). Any of the provided instructions are assumed to either consists of only measurement gates or correspond to a layer of 2-qubit gate instructions. For the former, the generated noise model will contain only single-qubit X errors, for the latter all 1- and 2-weight Pauli errors on the reduced coupling map will be included.
• coupling_map (CouplingMap | None) – the coupling map of the backend on which the instructions have been laid out. If this is None, a 1d line of qubits is assumed.
• circuit (QuantumCircuit | None) – the transpiled circuit which has been laid out on the provided coupling map. This may only be None when the coupling_map is also None.

Returns

A dictionary mapping the ref attributes of the InjectNoise annotation of each unique Box to the 1- and 2-weight Pauli terms whose errors are learned for this box.

Return type

dict[str, QubitSparsePauliList]

get_extremal_eigenvalue(spo, *, tol=1e-10, max_cycle=500, max_space=12, lindep=1e-11, **kwargs)

Compute the spectral norm of a Hermitian Pauli operator (as a signed extremal eigenvalue).

Given a Hermitian operator written as a weighted sum of Paulis, C = sum_k a_k P_k, this returns its most-negative eigenvalue. Because such an operator has a spectrum that is symmetric about zero, the magnitude of that eigenvalue is the operator’s spectral norm ||C|| (its largest eigenvalue in absolute value) – take abs() of the result if that is what you need. In this addon it is used to evaluate the commutator-norm error bound ||[E, O]|| for an error Pauli E and observable O.

This exploits how a sum of Paulis lives in a space of dimension 2^(p + c), where p is the number of independent anticommuting pairs the Paulis generate and c counts the remaining commuting directions. The commuting directions are conserved, so this reduced operator is block-diagonal, with 2^c blocks of size 2^p. When these blocks are small enough each is diagonalized exactly, giving a result exact to machine precision. Otherwise, the reduced operator is instead solved with the compiled Rust Davidson solver (a diagonally-preconditioned Davidson iteration, which checks a converged eigenvalue with a short Lanczos iteration for a lower one before reporting it). This path is accurate but not exact, and the arguments below affect only it. It builds the whole 2^(p + c)-dimensional operator in sparse form and holds a few vectors of that length, so its memory use grows with the qubit count of spo; callers are expected to bound it, as the bound computations do with eigval_max_qubits.

Note the sign convention: the returned eigenvalue is the most-negative one, so the spectral norm is its absolute value.

>>> from qiskit.quantum_info import SparsePauliOp
>>> from qiskit_addon_slc.utils import get_extremal_eigenvalue

>>> operator = SparsePauliOp(["ZI", "IZ"], [1.0, 0.5])
>>> converged, eigenvalue = get_extremal_eigenvalue(operator)
>>> converged
True
>>> float(eigenvalue)
-1.5
>>> abs(eigenvalue)  # the spectral norm
1.5

Parameters

• spo (SparsePauliOp) – the Hermitian operator whose most-negative eigenvalue (and hence spectral norm) to compute.
• tol (float) – the convergence threshold of the iterative solver. It bounds the change of the eigenvalue between iterations and, unless rounding errors at the scale of the operator norm dominate, the residual norm of the eigenvector.
• max_cycle (int) – the maximum number of iterations of the iterative solver.
• max_space (int) – the maximum size of the subspace built up by the iterative solver. Once reached, the subspace is restarted from its max_space // 2 lowest Ritz vectors.
• lindep (float) – the fraction of its squared norm below which a new subspace vector, after being orthogonalized against the existing subspace, is considered linearly dependent on it (and, thus, discarded).
• kwargs – deprecated and ignored. Additional keyword arguments used to be passed on to the pyscf.lib.davidson1 solver that this function no longer uses. They are still accepted for backwards compatibility, but have no effect and raise a DeprecationWarning.

Returns

A (converged, eigenvalue) pair. converged reports whether the computation succeeded (may be False if Davidson fails to converge). eigenvalue is the most-negative eigenvalue of spo.

Raises

ValueError – if tol is not positive and finite, max_cycle is not positive, max_space is less than 2, or lindep is not in [0, 1). These are checked for every spo, even one that is solved exactly.

Return type

tuple[bool, float]

iter_circuit(circuit, *, reverse=False, log_process=True)

Iterates over the instructions in a circuit.

Parameters

• circuit (QuantumCircuit) – the circuit to iterate over.
• reverse (bool) – whether to iterate the circuit in reverse order.
• log_process (bool) – whether to log process.

Yields

Tuples of length four, consisting of the encountered circuit instruction, the canonical qubit indices (i.e. the integer indices of the acted-upon qubits in the context of the input circuit), the InjectNoise attributes: modifier_ref and ref. The latter two items may be None indicating a circuit instruction that was not part of an unrolled BoxOp.

Return type

Generator[tuple[CircuitInstruction, list[int], str | None, str | None]]

map_modifier_ref_to_ref(circuit)

Iterate a circuit and map InjectNoise annotation references.

The bounds computation keys its results by modifier_ref, while a noise model is keyed by ref, so this mapping is what relates the two. This requires every modifier_ref to be unique, as provided by boxing a circuit with inject_noise_strategy="individual_modification" (see compute_bounds()). Boxes without an InjectNoise annotation are skipped.

>>> from qiskit import QuantumCircuit
>>> from qiskit.circuit import BoxOp
>>> from samplomatic import InjectNoise
>>> from qiskit_addon_slc.utils import map_modifier_ref_to_ref

>>> circuit = QuantumCircuit(2)
>>> for idx in range(2):
...     body = QuantumCircuit(2)
...     _ = body.cx(0, 1)
...     annotation = InjectNoise(ref=f"noise_{idx}", modifier_ref=f"box_{idx}")
...     _ = circuit.append(BoxOp(body, annotations=[annotation]), [0, 1])

>>> map_modifier_ref_to_ref(circuit)
{'box_0': 'noise_0', 'box_1': 'noise_1'}

Parameters

circuit (QuantumCircuit) – the circuit to iterate over.

Returns

A dictionary mapping each InjectNoise.modifier_ref to its InjectNoise.ref.

Return type

dict[str, str]

remove_measure(circuit)

Remove any Measure operations from the provided circuit.

>>> from qiskit import QuantumCircuit
>>> from qiskit_addon_slc.utils import remove_measure

>>> circuit = QuantumCircuit(2, 2)
>>> _ = circuit.h(0)
>>> _ = circuit.measure(0, 0)
>>> _ = circuit.cx(0, 1)
>>> dict(sorted(circuit.count_ops().items()))
{'cx': 1, 'h': 1, 'measure': 1}

>>> dict(sorted(remove_measure(circuit).count_ops().items()))
{'cx': 1, 'h': 1}

The original circuit is left untouched:

>>> dict(sorted(circuit.count_ops().items()))
{'cx': 1, 'h': 1, 'measure': 1}

Parameters

circuit (QuantumCircuit) – the circuit whose measurements to remove.

Returns

The circuit without any Measure operations.

Return type

QuantumCircuit