Skip to content

Anomaly detection

Section status

Contributor: Unclaimed — this section has no named author yet. See Contribute.

Last reviewed: —

Status: Stub — a starting point only. The curated list is not yet written.

Overview

Anomaly detection asks whether collider data contains something the Standard Model does not predict, without specifying in advance what that something is. It responds to a structural tension in the search program: a dedicated search is far more sensitive to the signal it was designed for, but the space of possible signals is vastly larger than the number of searches anyone can perform.

The literature divides by how much supervision is available, and that distinction matters more than the choice of architecture. Unsupervised methods flag events improbable under a learned model of the data. Weakly supervised methods exploit the expectation that a signal is localized in some variable, so a classifier separating a signal region from sidebands approximates the optimal discriminant without ever seeing a labeled signal. The latter has proven substantially more powerful.

A point easily missed on a first pass: the hard part is usually not the anomaly score but the background estimate, since a selection made with a learned score distorts the background in ways that must be controlled before any significance can be quoted.

  • Machine Learning for Anomaly Detection in Particle Physics, Belis et al. (2023) (arXiv:2312.14190) — the standard review of the area; read this first
  • Anomaly Detection for Physics Analysis and Less than Supervised Learning, Nachman (2020) (arXiv:2010.14554) — shorter and more conceptual, on the supervision spectrum and why weak supervision helps
  • The LHC Olympics 2020: A Community Challenge for Anomaly Detection in High Energy Physics, Kasieczka et al. (2021) (arXiv:2101.08320) — the LHC Olympics summary, doubling as an empirical survey of what many methods achieve on common data

Curated paper list

This list is a seed, not a curated selection

A few landmark papers are listed to give the section a starting shape. A proper curated list — thematically grouped, with an annotation on every entry — is what this section still needs. See Contribute.

  • Anomaly Detection for Resonant New Physics with Machine Learning, Collins et al. (2018) (arXiv:1805.02664) — CWoLa hunting: introduced the weakly supervised, sideband-based framing most later work builds on
  • Searching for New Physics with Deep Autoencoders, Farina et al. (2018) (arXiv:1808.08992) — autoencoder reconstruction error as an anomaly score — the canonical unsupervised baseline
  • Anomaly Detection with Density Estimation, Nachman et al. (2020) (arXiv:2001.04990) — ANODE: densities estimated in signal region and sidebands separately, making the background estimate intrinsic to the method
  • Classifying Anomalies THrough Outer Density Estimation (CATHODE), Hallin et al. (2021) (arXiv:2109.00546) — CATHODE: sampling from the sideband-interpolated density to train a classifier, now a standard baseline

Benchmarks, datasets & software

  • The LHC Olympics 2020: A Community Challenge for Anomaly Detection in High Energy Physics, Kasieczka et al. (2021) (arXiv:2101.08320) — LHC Olympics 2020 R&D and black-box datasets, with a leaderboard
  • The Dark Machines Anomaly Score Challenge: Benchmark Data and Model Independent Event Classification for the Large Hadron Collider, Aarrestad et al. (2021) (arXiv:2105.14027) — Dark Machines anomaly score challenge — complementary in signals and event representation

Open questions

What is settled here, what is contested, and what remains unsolved? This is the part a bibliography structurally cannot provide, and often the most useful paragraph on the page.

Further reading

Relevant work that is not an entry point — too specialized, too recent, or simply not where a newcomer should start. Suggestions that do not fit the curated list above belong here rather than being turned away.

Nothing listed yet.

Cross-references