Abstract: Current noisy intermediate-scale quantum (NISQ) computers are subject to various noises, resulting in the errors between the quantum circuit operation results and the ideal results. In order to make the circuit output closer to the desired result, the operation results of the quantum circuit need to be calibrated. Based on the reversibility of quantum circuit, the state errors in the forward and reverse circuit operation data are collected as the main noise features, and then an output calibration method based on support vector machine (SVM) is proposed. According to the method, the noise characteristics are sorted firstly by the support vector machine-recursive feature elimination (SVM-RFE) method, then the overfitted features are removed to obtain better calibration results for quantum circuit output. Experimental results show that compared with the optimization-based mapping methods, the proposed SVM-based method yields quantum circuit output results that are closer to the ideal results. In comparison to the decision tree ensemble classification model (Qraft), when the gate count of the CNOT quantum circuit is 60, the improvement rate reaches 43.94%.

Key words: quantum computing, NISQ calculation, noise characteristic, support vector machine