Skip to main content
OSTI.GOVU.S. Department of Energy Office of Scientific and Technical Information
U.S. Department of Energy
Office of Scientific and Technical Information
 

Challenges for unsupervised anomaly detection in particle physics

Journal Article · · Journal of High Energy Physics (Online)
 [1];  [2];  [2];  [2];  [2]
  1. Harvard University, Cambridge, MA (United States); Harvard Univ., Cambridge, MA (United States)
  2. Harvard University, Cambridge, MA (United States)
+ Show Author Affiliations

Anomaly detection relies on designing a score to determine whether a particular event is uncharacteristic of a given background distribution. One way to define a score is to use autoencoders, which rely on the ability to reconstruct certain types of data (background) but not others (signals). In this paper, we study some challenges associated with variational autoencoders, such as the dependence on hyperparameters and the metric used, in the context of anomalous signal (top and W) jets in a QCD background. We find that the hyperparameter choices strongly affect the network performance and that the optimal parameters for one signal are non-optimal for another. In exploring the networks, we uncover a connection between the latent space of a variational autoencoder trained using mean-squared-error and the optimal transport distances within the dataset. We then show that optimal transport distances to representative events in the background dataset can be used directly for anomaly detection, with performance comparable to the autoencoders. Whether using autoencoders or optimal transport distances for anomaly detection, we find that the choices that best represent the background are not necessarily best for signal identification. These challenges with unsupervised anomaly detection bolster the case for additional exploration of semi-supervised or alternative approaches.

Research Organization:
Harvard University, Cambridge, MA (United States)
Sponsoring Organization:
USDOE Office of Science (SC); National Science Foundation (NSF); National Aeronautics and Space Administration (NASA); Alfred P. Sloan Foundation
Grant/Contract Number:
SC0013607; SC0020223
OSTI ID:
1976605
Journal Information:
Journal of High Energy Physics (Online), Journal Name: Journal of High Energy Physics (Online) Journal Issue: 3; ISSN 1029-8479
Publisher:
Springer NatureCopyright Statement
Country of Publication:
United States
Language:
English

References (61)

Nonlinear principal component analysis using autoassociative neural networks journal February 1991
Optimal Transport book January 2009
Tag N’ Train: a technique to train improved classifiers on unlabeled data journal January 2021
DELPHES 3: a modular framework for fast simulation of a generic collider experiment journal February 2014
Jet-images: computer vision inspired techniques for jet tagging journal February 2015
(Machine) learning to do more with less journal February 2018
Does SUSY have friends? A new approach for LHC event analysis journal February 2021
Identifying boosted objects with N-subjettiness journal March 2011
Erratum to: Mass Unspecific Supervised Tagging (MUST) for boosted jets journal April 2021
Topological obstructions to autoencoding journal April 2021
Weakly supervised classification in high energy physics journal May 2017
Variational autoencoders for new physics mining at the Large Hadron Collider journal May 2019
Spheres to jets tuning event shapes with 5d simplified models journal May 2021
Autoencoders for unsupervised anomaly detection in high energy physics journal June 2021
The automated computation of tree-level and next-to-leading order differential cross sections, and their matching to parton shower simulations journal July 2014
The hidden geometry of particle collisions journal July 2020
The efficacy of event isotropy as an event shape observable journal July 2021
A robust measure of event isotropy at colliders journal August 2020
Anomaly detection with convolutional Graph Neural Networks journal August 2021
Combining outlier analysis algorithms to identify new physics at the LHC journal September 2021
Pulling out all the tops with computer vision and deep learning journal October 2018
Adversarially-trained autoencoders for robust unsupervised new physics searches journal October 2019
Learning the latent structure of collider events journal October 2020
A generic anti-QCD jet tagger journal November 2017
An introduction to PYTHIA 8.2 journal June 2015
Variational Inference: A Review for Statisticians journal July 2016
The anti- k t jet clustering algorithm journal April 2008
The LHC Olympics 2020 a community challenge for anomaly detection in high energy physics journal December 2021
Unsupervised outlier detection in heavy-ion collisions journal April 2021
Anomalous jet identification via sequence modeling journal August 2021
Uncovering latent jet substructure journal September 2019
Exploring the space of jets with CMS open data journal February 2020
Transferability of deep learning models in searches for new physics at colliders journal February 2020
Searching for new physics with deep autoencoders journal April 2020
Anomaly detection with density estimation journal April 2020
Novelty detection meets collider physics journal April 2020
Simulation assisted likelihood-free anomaly detection journal May 2020
Linearized optimal transport for collider events journal December 2020
Unsupervised clustering for collider physics journal May 2021
Simulation-assisted decorrelation for resonant anomaly detection journal August 2021
Learning to classify from impure samples with high-dimensional data journal July 2018
Extending the search for new resonances with machine learning journal January 2019
Learning new physics from a machine journal January 2019
Anomaly Detection for Resonant New Physics with Machine Learning journal December 2018
Metric Space of Collider Events journal July 2019
Anomaly Detection with Conditional Variational Autoencoders conference December 2019
FastJet user manual: (for version 3.0.2) journal March 2012
Guiding new physics searches with unsupervised learning journal March 2019
Lund jet images from generative and cycle-consistent adversarial networks journal November 2019
Finding new physics without learning about it: anomaly detection as a tool for searches at colliders journal January 2021
Learning multivariate new physics journal January 2021
Use of a generalized energy Mover’s distance in the search for rare phenomena at colliders journal February 2021
Comparing weak- and unsupervised methods for resonant anomaly detection journal July 2021
Adversarially Learned Anomaly Detection on CMS open data: re-discovering the top quark journal February 2021
An Introduction to Variational Autoencoders journal January 2019
Better latent spaces for better autoencoders journal January 2021
The Dark Machines Anomaly Score Challenge: Benchmark Data and Model Independent Event Classification for the Large Hadron Collider journal January 2022
Deep Set Auto Encoders for Anomaly Detection in Particle Physics journal January 2022
QCD or what? journal January 2019
How to GAN LHC events journal January 2019
Uncovering hidden new physics patterns in collider events using Bayesian probabilistic models conference February 2021

Similar Records

Deep Set Auto Encoders for Anomaly Detection in Particle Physics
Journal Article · Sun Jan 30 23:00:00 EST 2022 · SciPost Physics · OSTI ID:1842863
Improving Variational Autoencoders for New Physics Detection at the LHC With Normalizing Flows
Journal Article · Sun Feb 27 23:00:00 EST 2022 · Frontiers in Big Data · OSTI ID:1833285

Related Subjects

71 CLASSICAL AND QUANTUM MECHANICS
GENERAL PHYSICS
Hadron-Hadron scattering
beyond standard model
jet substructure
jets