Machine learning-based b-jet tagging in $pp$ collisions at $\sqrt{s}=13$ TeV

Abstract

Studying heavy-flavor jets in $pp$ collision is important since they can test pQCD calculations and be used as a reference for heavy-ion collisions. Jets in this analysis are reconstructed from charged particles using the anti-$k_{\mathrm{T}}$ algorithm with a resolution parameter $R=$ 0.4 and with pseudorapidity $|\eta |<$ 0.5. Beauty jets are tagged using a machine learning model that uses a convolutional neural network trained on information extracted from the jet, tracks, and secondary vertices. The results show that this model is superior compared to other traditional tagging methods.

9 pages, 6 captioned figures