Learning-based error mitigation for near-term quantum computers

Presenter: Piotr Czarnik from Jagiellonian University in Krakow

Abstract:

We introduce a novel approach to error mitigation for near-term quantum computers [1]. The approach is based on learning noise effects on expectation values of observables from classically simulable near-Clifford circuits similar to a circuit of interest [1]. We generalize the approach unifying it with other state-of-the-art approaches [2,3]. Furthermore, we improve its performance imposing symmetry constraints of the simulated system [4]. We provide its proof of principle obtaining orders of magnitude improvements of results for simulations of quantum-many body systems with IBM quantum computers [1, 4, 5].

References

[1] P. Czarnik, A. Arrasmith, P. Coles, L. Cincio, Quantum 5, 592 (2021)

[2] A. Lowe et al., Phys. Rev. Research 3, 033098 (2021)

[3] D. Bultrini et al. arXiv:2107.13470

[4] P. Czarnik et al., in prep.

[5] Y. Zhang et al., arXiv:2106.07619Bios:

Bio: Piotr Czarnik got PhD from theoretical physics at Jagiellonian University in Krakow. He started to work on near-term quantum computation as a Director's Post-Doctoral Fellow at Los Alamos National Laboratory. He continues that line of research as an assistant professor at Jagiellonian University