Skip to content

Differentiable error-mitigation product

Versioned mitigator taxonomy with differentiability classes, local ZNE and readout probes, and hard gaps for invent-green ideal-gradient restoration or live QPU mitigation.

Module: scpn_quantum_control.error_mitigation_product

Rules

Rule Behaviour
Product schema error_mitigation_product.v1
Ambient ZNE mitigation.zne.zne_extrapolate / zne_uncertainty
Ambient readout mitigation.readout_matrix
Studio composition executive_mitigate claim boundary
Ideal gradient restore Refuse invent-green
Live QPU mitigation Refused by the no-submit safety policy
mitiq optional_extra only — not a hard dependency

Differentiability classes

Class Meaning
analytic_fd Linear map / FD-friendly (e.g. readout invert)
fd_only Extrapolation arithmetic on scalars (ZNE)
non_diff Sampling / discrete post-selection (PEC, symmetry)
optional_extra Optional dependency path (mitiq)

Quick start

from scpn_quantum_control.error_mitigation_product import (
    assert_error_mitigation_product_integrity,
    build_error_mitigation_product_registry,
    decide_mitigation_path,
    materialise_demo_zne_probe,
    materialise_readout_probe,
)

reg = assert_error_mitigation_product_integrity(
    build_error_mitigation_product_registry()
)
assert reg["ideal_gradient_restore_policy"] is False

assert decide_mitigation_path(
    "zne_richardson", invent_green_ideal_gradient_restore=True
).allowed is False

probe = materialise_demo_zne_probe()
assert probe.invent_green_live_qpu is False

readout = materialise_readout_probe()
assert readout.n_qubits == 1

Residuals (honest)

  • Deeper open-system objective integration
  • Metamorphic registration for mitigated estimators
  • Ambient: scpn_quantum_control.mitigation.*
  • No-submit hardware safety · Studio executive mitigation

Authored by Anulum Fortis & Arcane Sapience (protoscience@anulum.li)