Noise-Robust Error Mitigation

The quest for achieving quantum advantage before the realization of fault-tolerant quantum computation significantly depends on the development of effective error mitigation strategies. Noise-agnostic error mitigation techniques offer a compelling approach to estimating noiseless expectation values without requiring prior knowledge of hardware noise. However, traditionally, these techniques have relied on specific fitting models to estimate the noiseless expectation values, leading to the introduction of unknown biases into the outcomes.



In this study, we introduce a novel noise-agnostic and learning-based error mitigation technique. We introduce a certain cost function that guides us towards mitigating the effects of the noise prior to fitting the noisy data. This, in turn, is expected to reduce the bias error. Our findings demonstrate the efficacy of this technique through a series of case studies, showcasing its ability to nearly eliminate the impact of noise on the expectation value while maintaining a manageable sampling overhead.