# How do I use ZNE?#

Zero-noise extrapolation (ZNE) is an error mitigation technique in which an expectation value is computed at different noise levels and, as a second step, the ideal expectation value is inferred by extrapolating the measured results to the zero-noise limit (see the section What is the theory behind ZNE?).

Both steps — noise scaling and extrapolation — can be applied with Mitiq. The corresponding sub-modules are mitiq.zne.scaling and mitiq.zne.inference.

As with all techniques, ZNE is compatible with any frontend supported by Mitiq:

import mitiq

mitiq.SUPPORTED_PROGRAM_TYPES.keys()

dict_keys(['cirq', 'pyquil', 'qiskit', 'braket', 'pennylane'])


In the next cell you can select the frontend used in this tutorial. For example:

frontend = "cirq"


## Problem setup#

We first define the circuit of interest. For simplicity, in this example we use a randomized-benchmarking circuit whose ideal execution is equivalent to the identity operation.

from mitiq import benchmarks

circuit = benchmarks.generate_rb_circuits(
n_qubits=1, num_cliffords=2, return_type = frontend,
)

print(circuit)

0: ───X^-0.5───Y^-0.5───Y^0.5───X^0───X^0.5───Y^0───


As an example, below we define a simple executor function which inputs a circuit, executes it on a noisy simulator, and returns the probability of the ground state. See the Executors section for more information on how to define more advanced executors.

import numpy as np
from cirq import DensityMatrixSimulator, depolarize
from mitiq.interface import convert_to_mitiq

def execute(circuit, noise_level=0.01):
"""Returns Tr[ρ |0⟩⟨0|] where ρ is the state prepared by the circuit
executed with depolarizing noise.
"""
# Replace with code based on your frontend and backend.
mitiq_circuit, _ = convert_to_mitiq(circuit)
noisy_circuit = mitiq_circuit.with_noise(depolarize(p=noise_level))
rho = DensityMatrixSimulator().simulate(noisy_circuit).final_density_matrix
return rho[0, 0].real


The executor can be used to evaluate noisy (unmitigated) expectation values.

# Compute the expectation value of the |0><0| observable.
noisy_value = execute(circuit)
ideal_value = execute(circuit, noise_level=0.0)
print(f"Error without mitigation: {abs(ideal_value - noisy_value) :.5f}")

Error without mitigation: 0.03869


## Apply ZNE#

Zero-noise extrapolation can be easily implemented with the function mitiq.zne.zne.execute_with_zne().

from mitiq import zne

mitigated_result = zne.execute_with_zne(circuit, execute)

print(f"Error with mitigation (ZNE): {abs(ideal_value - mitigated_result):.{3}}")

Error with mitigation (ZNE): 0.000232


Here we observe that the application of ZNE reduces the estimation error when compared to the unmitigated result. In the example above, both the noise scaling and inference steps were taken behinds the scenes thanks to the default options of execute_with_zne(). Below we provide more details about these two aspects of ZNE in Mitiq.

## Select a noise scaling method#

In Mitiq, one can select a noise scaling method via noise scaling functions. A noise scaling function takes as input a circuit and a real scale factor and returns a new circuit. The returned circuit is equivalent to the input one (if executed on a noiseless backend), but is more sensitive to noise when executed on a real noisy backend. In practice, by applying a noise scaling function before the execution of a circuit, one can indirectly scale up the effect of noise. See the section What additional options are available in ZNE? for more details.

For example, the function fold_gates_at_random() applies the unitary folding map $$G -> G G^\dagger G$$ to a random subset of gates of the input circuit. The folded circuit is more sensitive to gate errors since it has a number of gates approximately equal to scale_factor * n, where n is the number of gates in the input circuit.

from mitiq.zne.scaling import fold_gates_at_random

folded = fold_gates_at_random(circuit, scale_factor=2.)
print("Folded circuit:", folded, sep="\n")

Folded circuit:
0: ───X^-0.5───X^0.5───X^-0.5───Y^-0.5───Y^0.5───Y^-0.5───Y^0.5───X^0───X^0.5───Y^0───Y^0───Y^0───


## Select a noise extrapolation method#

Define a Factory object to select the noise extrapolation method and the noise scale factors. The section What additional options are available when using ZNE? also contains more details on the available noise extrapolation methods in Mitiq.

For example, a linear extrapolation method with scale factors 1 and 2, can be initialized as follows:

from mitiq.zne.inference import LinearFactory

linear_fac = LinearFactory(scale_factors=[1.0, 2.0])


We can use the defined Factory and chosen noise scaling method to apply ZNE.

from mitiq import zne

mitigated_result = zne.execute_with_zne(
circuit, execute, factory=linear_fac, scale_noise=fold_gates_at_random,
)

print(f"Error with mitigation (ZNE): {abs(ideal_value - mitigated_result):.{3}}")

Error with mitigation (ZNE): 0.00299


The section What additional options are available when using ZNE? contains more information on both noise scaling and noise extrapolation methods in Mitiq.