Leveraging junk information to enhance the quantum error mitigation

Abstract

Noise in quantum information processing poses a significant obstacle to achieving precise results. Quantum error mitigation techniques are crucial for improving the accuracy of experimental expectation values in this process. In the experiments, it is commonly observed that some measured events violate certain principles, such as symmetry constraints. These events can be considered junk information and should be discarded in a post-selection process. In this work, we introduce a quantum error mitigation method named Self-Trained Quantum Noise Filter (SQNF), which leverages the junk information to differentiate errors from the experimental population distributions, thereby aiming to approximate the error-free distribution. Our numerical results demonstrate that the proposed method can significantly reduce the infidelity of population distributions compared to the traditional post-selection method. Notably, the infidelity reduction is achieved without additional experimental resource consumption. Our method is scalable and applicable to multi-qubit computing systems.