期待値推定のための演算子バックプロパゲーション（OBP）

import numpy as np
import matplotlib.pyplot as plt

from qiskit.primitives import StatevectorEstimator
from qiskit.transpiler.preset_passmanagers import generate_preset_pass_manager
from qiskit.quantum_info import SparsePauliOp
from qiskit.transpiler import CouplingMap
from qiskit.synthesis import LieTrotter

from qiskit_addon_utils.problem_generators import generate_xyz_hamiltonian
from qiskit_addon_utils.problem_generators import (
    generate_time_evolution_circuit,
)
from qiskit_addon_utils.slicing import slice_by_depth, combine_slices
from qiskit_addon_obp.utils.simplify import OperatorBudget
from qiskit_addon_obp import backpropagate
from qiskit_addon_obp.utils.truncating import setup_budget

from rustworkx.visualization import graphviz_draw

from qiskit_ibm_runtime import QiskitRuntimeService
from qiskit_ibm_runtime import EstimatorV2, EstimatorOptions

num_qubits = 10
layout = [(i - 1, i) for i in range(1, num_qubits)]

# Instantiate a CouplingMap object
coupling_map = CouplingMap(layout)
graphviz_draw(coupling_map.graph, method="circo")

# Get a qubit operator describing the Heisenberg XYZ model
hamiltonian = generate_xyz_hamiltonian(
    coupling_map,
    coupling_constants=(np.pi / 8, np.pi / 4, np.pi / 2),
    ext_magnetic_field=(np.pi / 3, np.pi / 6, np.pi / 9),
)
print(hamiltonian)

circuit = generate_time_evolution_circuit(
    hamiltonian,
    time=0.2,
    synthesis=LieTrotter(reps=2),
)
circuit.draw("mpl", style="iqp", fold=-1)

slices = slice_by_depth(circuit, max_slice_depth=1)
print(f"Separated the circuit into {len(slices)} slices.")

op_budget = OperatorBudget(max_qwc_groups=8)

observable = SparsePauliOp.from_sparse_list(
    [("Z", [i], 1 / num_qubits) for i in range(num_qubits)],
    num_qubits=num_qubits,
)
observable

# Backpropagate slices onto the observable
bp_obs, remaining_slices, metadata = backpropagate(
    observable, slices, operator_budget=op_budget
)
# Recombine the slices remaining after backpropagation
bp_circuit = combine_slices(remaining_slices)

print(f"Backpropagated {metadata.num_backpropagated_slices} slices.")
print(
    f"New observable has {len(bp_obs.paulis)} terms, which can be combined into "
    f"{len(bp_obs.group_commuting(qubit_wise=True))} groups."
)
print(
    f"Note that backpropagating one more slice would result in "
    f"{metadata.backpropagation_history[-1].num_paulis[0]} terms "
    f"across {metadata.backpropagation_history[-1].num_qwc_groups} groups."
)
print("The remaining circuit after backpropagation looks as follows:")
bp_circuit.draw("mpl", fold=-1, scale=0.6)

service = QiskitRuntimeService()
backend = service.least_busy(
    operational=True, simulator=False, min_num_qubits=133
)
print(backend)

pm_basis = generate_preset_pass_manager(
    optimization_level=3, basis_gates=backend.configuration().basis_gates
)
isa_circuit = pm_basis.run(circuit)
isa_bp_circuit = pm_basis.run(bp_circuit)

pubs = [(isa_circuit, observable), (isa_bp_circuit, bp_obs)]

rng = np.random.default_rng()
estimator = StatevectorEstimator(seed=rng)
job = estimator.run(pubs)

primitive_result = job.result()
circuit_expval = primitive_result[0].data.evs.item()
bp_circuit_expval = primitive_result[1].data.evs.item()

methods = [
    "No backpropagation",
    "Backpropagation",
]
values = [circuit_expval, bp_circuit_expval]

ax = plt.gca()
plt.bar(methods, values, color="#a56eff", width=0.4, edgecolor="#8a3ffc")
ax.set_ylim([0.6, 0.92])
ax.set_ylabel(r"$M_Z$", fontsize=12)

num_qubits = 50
layout = [(i - 1, i) for i in range(1, num_qubits)]

# Instantiate a CouplingMap object
coupling_map = CouplingMap(layout)

hamiltonian = generate_xyz_hamiltonian(
    coupling_map,
    coupling_constants=(np.pi / 8, np.pi / 4, np.pi / 2),
    ext_magnetic_field=(np.pi / 3, np.pi / 6, np.pi / 9),
)

# Generate a time evolution circuit for the Hamiltonian
circuit = generate_time_evolution_circuit(
    hamiltonian,
    time=0.2,
    synthesis=LieTrotter(reps=4),
)

# Define the observable to measure
observable = SparsePauliOp.from_sparse_list(
    [("Z", [i], 1 / num_qubits) for i in range(num_qubits)],
    num_qubits,
)

slices = slice_by_depth(circuit, max_slice_depth=1)

# Define the maximum number of qwc groups allowed in the
# backpropagated observable,
# and the truncation error budget
op_budget = OperatorBudget(max_qwc_groups=15)
truncation_error_budget = setup_budget(
    max_error_total=0.03, max_error_per_slice=0.005
)

# First backpropagation without truncation
bp_obs, remaining_slices, metadata = backpropagate(
    observable, slices, operator_budget=op_budget
)
bp_circuit = combine_slices(remaining_slices)

# Now backpropagate with truncation, using the same operator budget and
# the defined truncation error budget
bp_obs_trunc, remaining_slices_trunc, metadata = backpropagate(
    observable,
    slices,
    operator_budget=op_budget,
    truncation_error_budget=truncation_error_budget,
)
bp_circuit_trunc = combine_slices(
    remaining_slices_trunc, include_barriers=False
)

# Now we transpile the original circuit and the two backpropagated circuits,
# and apply the layout to the corresponding observables
pm = generate_preset_pass_manager(optimization_level=3, backend=backend)

isa_circuit = pm.run(circuit)
isa_bp_circuit = pm.run(bp_circuit)
isa_bp_circuit_trunc = pm.run(bp_circuit_trunc)

isa_observable = observable.apply_layout(isa_circuit.layout)
isa_bp_observable = bp_obs.apply_layout(isa_bp_circuit.layout)
isa_bp_observable_trunc = bp_obs_trunc.apply_layout(
    isa_bp_circuit_trunc.layout
)

# Compare the 2-qubit depth of each transpiled circuit to see how much
# depth backpropagation saved
print(
    f"2-qubit depth without backpropagation: "
    f"{isa_circuit.depth(lambda x: x.operation.num_qubits == 2)}"
)
print(
    f"2-qubit depth with backpropagation: "
    f"{isa_bp_circuit.depth(lambda x: x.operation.num_qubits == 2)}"
)
print(
    f"2-qubit depth with backpropagation and truncation: "
    f"{isa_bp_circuit_trunc.depth(lambda x: x.operation.num_qubits == 2)}"
)

pubs = [
    (isa_circuit, isa_observable),
    (isa_bp_circuit, isa_bp_observable),
    (isa_bp_circuit_trunc, isa_bp_observable_trunc),
]

# Now we instantiate the Estimator primitive for the hardware with
# ZNE and measurement error
# mitigation and compute the three circuits and observables
options = EstimatorOptions()
options.default_precision = 0.01
options.resilience_level = 2
options.resilience.zne.noise_factors = [1, 1.2, 1.4]
options.resilience.zne.extrapolator = ["linear"]
estimator = EstimatorV2(mode=backend, options=options)

estimator.options.environment.job_tags = ["TUT_OBP"]
job = estimator.run(pubs)

# Retrieve the results and the standard deviations
result_no_bp = job.result()[0].data.evs.item()
result_bp = job.result()[1].data.evs.item()
result_bp_trunc = job.result()[2].data.evs.item()

std_no_bp = job.result()[0].data.stds.item()
std_bp = job.result()[1].data.stds.item()
std_bp_trunc = job.result()[2].data.stds.item()

print(f"Expectation value without backpropagation: {result_no_bp}")
print(f"Backpropagated expectation value: {result_bp}")
print(f"Backpropagated expectation value with truncation: {result_bp_trunc}")

# Plot the results
methods = [
    "No backpropagation",
    "Backpropagation",
    "Backpropagation w/ truncation",
]
values = [result_no_bp, result_bp, result_bp_trunc]
error_bars = [std_no_bp, std_bp, std_bp_trunc]

ax = plt.gca()
plt.bar(methods, values, color="#a56eff", width=0.4, edgecolor="#8a3ffc")
plt.errorbar(methods, values, yerr=error_bars, fmt="o", color="r", capsize=5)
plt.axhline(0.89)
ax.set_ylim([0.8, 0.98])
plt.text(0.25, 0.895, "Exact result")
ax.set_ylabel(r"$M_Z$", fontsize=12)