Mitigating errors in noisy quantum computers

The effectiveness of Quantum Error Mitigation (QEM) protocols depends crucially upon the capability of reconstructing noise processes in experimental setups. In the available QEM protocols, this reconstruction is performed by analyzing individual gates acting on one or two qubits and does not account for noise processes that correlate many qubits, such as cross-talks. To overcome this limitation, we provide a set of QEM protocols that rely on noise reconstruction techniques designed to efficiently reconstruct noise processes affecting a large number of qubits, potentially the whole register. Exploring different approaches, we enhance existing methods (such as quasi-probabilistic error cancellation) and propose novel techniques (which we call “cycle extrapolation” and “circuit repetition”) to drastically upgrade the performance of noisy quantum computers. Making only mild assumptions on the hardware, we rigorously quantify the effectiveness and overhead of all our protocols. Overall, our work offers the tools to significantly improve the outputs of noisy quantum computers, as well as to build confidence in these outputs by reproducing them with different techniques.