Quantum error correction (QEC) and quantum error mitigation (QEM) are
the two most popular techniques to deal with errors occurring during a
quantum computation.
QEC comes with a polynomial asymptotic overhead in terms of qubit and
gate count, but the overhead is massive in practice. For example, the
factorization of RSA-2,048 may require surface codes using 1,457
physical qubits per logical qubit. QEM is already used in today’s
machines because it consumes only few or no extra qubits. However, it is
not a viable solution for large quantum circuits because the sampling
cost, i.e., the number of circuit executions required, generally grows
exponentially with the size of the circuit.
In this work, we propose a scheme that fills the gap between the regime
of QEC and QEM. This Clifford noise reduction (CliNR) scheme reduces the
logical error rate of Clifford circuits in practically relevant regimes,
and it does so at the price of comparatively small overheads. It can be
made universal together with physical or logical single-qubit rotations.
Based on joined work with Edwin Tham
https://arxiv.org/abs/2407.06583