Disentangling quark and gluon jets with normalizing flows

Abstract

We use the CMS Open Data to examine the performance of weakly-supervised learning for tagging quark and gluon jets at the LHC. We target Z+jet and dijet events as respective quark-and gluonenriched mixtures and derive samples both from data taken in 2011 at 7 TeV, and from Monte Carlo. CWoLa and TopicFlow models are trained on real data and compared to fully-supervised classifiers trained on simulation. In order to obtain estimates for the discrimination power in real data, we consider three different estimates of the quark/gluon mixture fractions in the data. Compared to when the models are evaluated on simulation, we find reversed rankings for the fully-and weaklysupervised approaches. Further, these rankings based on data are robust to the estimate of the mixture fraction in the test set. Finally, we use TopicFlow to smooth statistical fluctuations in the small testing set, and to provide uncertainty on the performance in real data.

Performance versus resilience in modern quark-gluon tagging

“…In spite of the serious theoretical challenges [26][27][28][29][30][31][32][33][34][35], efficient ML-approaches have been devised to separate "quark jets" from "gluon jets" [36][37][38][39][40][41]. They include study of hadronization and detector effects [42] and modern network architecture like transformers [43], Lorentzequivariant networks [44], and normalizing flows [45]. One way to overcome the fundamental problem of defining quark and gluon jets is to instead use well-defined hypotheses in terms of LHC signatures, for instance mostly quarks in LHC signals like weak boson fusion vs. gluons in QCD backgrounds [46][47][48][49][50].…”