https://arxiv.org/abs/2002.09935

Usability of the quantum machine learning (QML), based on the variational quantum algorithm for the HEP data analysis is demonstrated in this article. The example physics target is a SUSY signal, chargino-pair production via a Higgs boson where the chargino decays into a neutralino and a W boson, that is decaying into two charged leptons (plus neutrinos). The background process selected for the demonstration is the W boson pair production with each W boson decaying to a charged lepton and a neutrino, thus the final states are identical. Kinematic variables, number of variables is varying from 3 to 7, are used for the machine learning input, together with their parameters. The comparison has been perfomred between classical machine learnings (the BDT and the DNN) and the QMLs for gate-based quantum computer. Implementation of QML consists of two packages: Quantum Circuit Learning (QCL), used for simulation of the QML performance, and Variational Quantum Classification (VQC) that is used for testing the QML algorithm on real quantum computer and simulator as well. The comparison results are evaulated by the AUC (area under the curve) values by changing the number of training size/events and it is proven that QML has the ability to discriminate the signal from the background effectively as the classical MLs, especially at the region where the number of training events is relatively small. The authors also mention that the “the actual VQC performance varies when it runs on the simulator or real quantum computer, most likely due to erros in quantum hardware”