Error Mitigation

NISQ devices are noisy due to the imperfect gate operations, unwanted interactions with environment, and measurement errors. Error mitigation is a promising way of reducing the effect of noise. In this tutorial, we will explore three error mitigation techniques: Reqdout error mitigation, Zero-noise Extrapolation (ZNE), and Clifford Data Regression (CDR).

📄️ Readout error mitigation

Due to the imperfection of real devices, errors occur in state preparation and measurement. Readout error mitigation reduces the effect of those errors by applying inverse of such noisy operations. The inverse of the noisy operation, here we call filter matrix, plays an important role in readout error mitigation. Noisy counts $C{\text}$ can be considered as a product of ideal counts $C{\text{ideal}}$ which we could get in noiseless world and the error matrix $E$.

📄️ Zero-noise extrapolation

ZNE is an error mitigation method that extrapolates the noiseless value from the multiple noise level values. The method consists of three steps:

📄️ Clifford data regression

CDR is an error mitigation method which predicts the noiseless value using training data, which can be generated by exact estimator such as a simulator and a noisy estimator such as a real device. The method consists of three steps: