Data-driven approaches to error mitigation are the current state-of-the-art. We unify two such approaches. One approach involves circuits with variable noise, and the other involves Clifford circuits. Our approach uses both types of data (Clifford circuits with variable noise), resulting in a method more powerful than the previous ones.
Our method unifies zero-noise extrapolation (ZNE) with Clifford data regression (CDR). It can be viewed as guiding the extrapolation (in ZNE) with Clifford circuit examples. Rather than extrapolating “into the dark”, Clifford circuits guide one’s extrapolation. In numerical benchmarks, our method outperforms both ZNE and CDR.
For details see “Unified approach to data-driven quantum error mitigation”, A. Lowe, M. Hunter Gordon, P. Czarnik, A. Arrasmith, P.J. Coles, L. Cincio. arXiv:2011.01157.