Automatically find cuts using CKT

Create a circuit and observables

import numpy as np
from qiskit.circuit.random import random_circuit
from qiskit.quantum_info import SparsePauliOp

circuit = random_circuit(7, 6, max_operands=2, seed=1242)
observable = SparsePauliOp(["ZIIIIII", "IIIZIII", "IIIIIIZ"])


circuit.draw("mpl", scale=0.8)

Find cut locations, given a maximum of 4 qubits per subcircuit. This circuit can be separated in two by making a single wire cut and cutting one CRZGate

from circuit_knitting.cutting.automated_cut_finding import (
    find_cuts,
    OptimizationParameters,
    DeviceConstraints,
)

# Specify settings for the cut-finding optimizer
optimization_settings = OptimizationParameters(seed=111)

# Specify the size of the QPUs available
device_constraints = DeviceConstraints(qubits_per_subcircuit=4)

cut_circuit, metadata = find_cuts(circuit, optimization_settings, device_constraints)
print(
    f'Found solution using {len(metadata["cuts"])} cuts with a sampling '
    f'overhead of {metadata["sampling_overhead"]}.'
)
for cut in metadata["cuts"]:
    print(f"{cut[0]} at circuit instruction index {cut[1]}")
cut_circuit.draw("mpl", scale=0.8, fold=-1)

Found solution using 2 cuts with a sampling overhead of 127.06026169907257.
Wire Cut at circuit instruction index 19
Gate Cut at circuit instruction index 28

Add ancillas for wire cuts and expand the observables to account for ancilla qubits

from circuit_knitting.cutting import cut_wires, expand_observables

qc_w_ancilla = cut_wires(cut_circuit)
observables_expanded = expand_observables(observable.paulis, circuit, qc_w_ancilla)
qc_w_ancilla.draw("mpl", scale=0.8, fold=-1)

Partition the circuit and observables into subcircuits and subobservables. Calculate the sampling overhead incurred from cutting these gates and wires.

from circuit_knitting.cutting import partition_problem

partitioned_problem = partition_problem(
    circuit=qc_w_ancilla, observables=observables_expanded
)
subcircuits = partitioned_problem.subcircuits
subobservables = partitioned_problem.subobservables
print(
    f"Sampling overhead: {np.prod([basis.overhead for basis in partitioned_problem.bases])}"
)

Sampling overhead: 127.06026169907257

subobservables

{0: PauliList(['IIII', 'IZII', 'IIIZ']),
 1: PauliList(['ZIII', 'IIII', 'IIII'])}

subcircuits[0].draw("mpl", style="iqp", scale=0.8)

subcircuits[1].draw("mpl", style="iqp", scale=0.8)

Generate the experiments to run on the backend.

from circuit_knitting.cutting import generate_cutting_experiments

subexperiments, coefficients = generate_cutting_experiments(
    circuits=subcircuits, observables=subobservables, num_samples=1_000
)
print(
    f"{len(subexperiments[0]) + len(subexperiments[1])} total subexperiments to run on backend."
)

96 total subexperiments to run on backend.