Improving Probabilistic Error Cancellation in the Presence of Nonstationary Noise

In this article, we investigate the stability of probabilistic error cancellation (PEC) outcomes in the presence of nonstationary noise, which is an obstacle to achieving accurate observable estimates. Leveraging Bayesian methods, we design a strategy to enhance PEC stability and accuracy. Our experiments using a five-qubit implementation of the Bernstein–Vazirani algorithm and conducted on the ibm_kolkata device reveal a 42% improvement in accuracy and a 60% enhancement in stability compared to nonadaptive PEC. These results underscore the importance of adaptive estimation processes to effectively address nonstationary noise, vital for advancing PEC utility.