ML-aided searches for periodic features and dark jets at the LHC

Abstract:
In the first part of the talk I will discuss strategies to search for
new physics in LHC data using wavelet transforms of kinematics
distributions and machine learning. Such searches are intended to be
sensitive to periodic contributions, for example due to a dense
spectrum of KK modes. I will also present a recent implementation of
such a search in ATLAS.
In the second part I will discuss a search strategy for anomalous jets
with displaced vertices. Such jets may arise due to a "dark" QCD-like
sector. Since the dark sector parameters are largely unconstrained,
and the dynamics of a dark QCD-like theory are difficult to compute or
simulate reliably, model-independent, data-based searches for such
scenarios are desirable. This can be accomplished with weakly
supervised machine learning.