基于SVM的量子线路输出校准方法研究  

Research on calibration method of quantum circuit output based on SVM

作　　者：李响 姜一博 曹可欣 朱明强 程学云[1] 朱鹏程[1] 管致锦[1] LI Xiang;JIANG Yibo;CAO Kexin;ZHU Mingqiang;CHENG Xueyun;ZHU Pengcheng;GUAN Zhijin(School of Information Science and Technology,Nantong University,Nantong 226019,China)

机构地区：[1]南通大学信息科学技术学院,江苏南通226019

出　　处：《量子电子学报》2024年第2期357-366,共10页Chinese Journal of Quantum Electronics

基　　金：国家自然科学基金面上项目(62072259);江苏省自然科学基金面上项目(BK20221411);南通大学博士启动基金(23B03);江苏省研究生科研与实践创新计划项目(SJCX21_1448)。

摘　　要：当前噪声中尺度量子(NISQ)计算机由于受到各种噪声的影响,量子线路运行结果和理想结果之间存在误差,因此需要对量子线路的运行结果进行校准。基于量子线路可逆性的特点,收集正反向线路运行数据中的状态误差作为主要噪声特征,提出了基于支持向量机(SVM)集成策略的输出校准方法。通过支持向量机-递归特征消除(SVM-RFE)方法对噪声特征进行排序,去除过拟合的特征,从而得到更优的量子线路输出校准结果。实验结果表明,与基于优化映射的方法相比,所提基于SVM的方法使量子线路输出结果更接近于理想结果,与基于决策树集成分类模型(Qraft)相比,当CNOT量子线路的门数为60时,改善率达到43.94%。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%.

关 键 词：量子计算 NISQ计算 噪声特征 支持向量机 

分 类 号：TP302.2[自动化与计算机技术—计算机系统结构]