BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Uncovering hidden new physics patterns at high-energy colliders wi
 th probabilistic models - Darius Faroughy (UZH)
DTSTART:20210225T160000Z
DTEND:20210225T170000Z
UID:TALK157417@talks.cam.ac.uk
CONTACT:Joseph Davighi
DESCRIPTION:*The seminar will take place via Zoom "here":https://maths-cam
 -ac-uk.zoom.us/j/95215212460?pwd=TWN5cjJ3azErSUYremw5UWRKL0NKUT09.*\n\nAbs
 tract: Individual events at high-energy colliders like the LHC can be repr
 esented by a sequence of measurements\, or ‘point patterns’ in a space
  of high-level observables. We build a simple generative probabilistic mod
 el for event patterns that can be used for unsupervised classification tas
 ks in Beyond the SM studies. In order to arrive to this model we assume th
 at event measurements are exchangeable\, discrete\, and generated from mul
 tiple latent distributions\, called themes. The resulting probabilistic mo
 del is a mixed-membership model known as Latent Dirichlet Allocation (LDA)
 \, a model extensively used in natural language processing\, biology and m
 any unsupervised machine learning applications. By training on point patte
 rns in the Lund jet plane\, we demonstrate that a two-theme LDA model can 
 discover heavy resonances hidden in dijet data in a fully unsupervised man
 ner.\n\nThe slides are available "here":https://www.hep.phy.cam.ac.uk/~con
 ference/Faroughy_2021-02-25.pdf
LOCATION:Virtual Seminar 
END:VEVENT
END:VCALENDAR