mitigation.readout

Readout Error Mitigation (REM) for MonarQ

Two methods are available:

method='matrix'

Inverse confusion matrix (mitiq approach). Builds the full 2ⁿ × 2ⁿ matrix as the tensor product of local calibration matrices, then applies its pseudo-inverse. Simple and exact, but limited to ~12 qubits due to exponential memory.

method='m3'

Matrix-free Measurement Mitigation (mthree). Works on the reduced sub-matrix of observed bitstrings only — scalable to 24 qubits and beyond. Recommended when the number of measured qubits exceeds ~10.

In both cases, the calibration fidelities P(0|0) and P(1|1) are read directly from the Anyon benchmark via the Calcul Québec API, without submitting additional calibration circuits to the hardware.

Classes

ReadoutMitigation(backend[, method, ...])

Readout Error Mitigation (REM) for MonarQ.