Universal Quantum Error Mitigation via Random Inverse Depolarizing Approximation

Given the noise on present-day quantum computers, error mitigation methods are essential in order to harness their power. However, existing quantum error mitigation methods frequently require significant overhead, call for unrealistically accurate noise models, scale poorly in the face of high error, or are application-specific. To circumvent these difficulties, we introduce Random Inverse Depolarizing Approximation (RIDA), which models the global noise channel as a depolarizing channel and determines the depolarization probability of each circuit by running a randomly extracted half of the circuit followed by its inverse. RIDA then inverts the noise channel to estimate the error-free expectation.

RIDA requires minimal overhead (as little as a single reusable estimation circuit), simultaneously mitigates both gate and measurement error, scales well to high errors, requires no prior noise model information, and applies to any quantum circuit of interest. In numerical experiments on simulated quantum computers, we show RIDA achieves lower error than leading methods. This persists across the wide range of tested scenarios including 75% less error than the next best method in low-error, low-shot tests and 89% less error in high-error, high-shot tests. These advantages bode well for the immediate practical use of RIDA in near-term quantum computing.