Probes of Anomalous Events at LHC with Self-Organizing Maps

We propose an Unsupervised Learning Algorithm, Self-Organizing Maps (SOM), built on a neural network architecture, for the probe of a rare top decay, mediated by Flavor Changing Neutral Current (FCNC), to charm and the Higgs boson, with the Higgs boson further decaying to a pair of b-quarks or a pair of gauge bosons ($W^{\pm}/Z$) in a boosted regime. Ideally, the particles originating from the decay of the boosted top lead to the reconstruction of a large-R jet, comprising three-prong substructures, with b- and c-tagged subjets. The SOM algorithm has been demonstrated as a model-agnostic anomaly-finder for probing the rare decay at the LHC, by mapping distinct signal and background regions to separate non-overlapping clusters on the Kohnen map. This helps to identify signal regions with higher signal significances. We also discuss the robustness of this algorithm, especially for other BSM probes with model-agnostic and model-dependent searches.