| Details: |
Classification of jets with deep learning has gained significant
attention in recent times. The primary motivation has been to extract
the maximum information from the complex collision events and improve
our understanding of jets. In this talk, I'll first give a brief
introduction to Jets and the usefulness of machine learning techniques
associated with the LHC Physics. I'll then introduce an infrared and
collider safe observable S2(R), R being some angular scale, constructed
using the (boosted) jet constituents. The resulting spectrum of S2(R)
identifies both hard and soft radiations associated with the particle
decay. Using the jet spectrum and deep learning technique, we build a
classifier which improves the classification of different kinds of jets.
Note, the performance of deep neural networks is often achieved at the
cost of interpretability. I'll briefly mention how an interpretable
network trained on S2(R) provides a qualitative understanding of the
network predictions. |