Better latent spaces for better autoencoders

Dillon, Barry M.; Plehn, Tilman; Sauer, Christof; Sorrenson, Peter

doi:10.21468/scipostphys.11.3.061

Cited by 52 publications

(53 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Again, VAEs have been shown to work for anomaly searches with LHC jets [6,7], and we can replace the reconstruction loss by an anomaly score defined in latent space. At this point, an interesting and relevant question becomes the choice of latent space and specific anomaly score, for instance in a Dirichlet latent space [8,9] which encourages mode separation [4].…”

Section: What Is Anomalous?mentioning

confidence: 99%

What's Anomalous in LHC Jets?

Buss¹,

Dillon²,

Finke³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Searches for anomalies are the main motivation for the LHC and define key analysis steps, including triggers. We discuss how LHC anomalies can be defined through probability density estimates, evaluated in a physics space or in an appropriate neural network latent space. We illustrate this for classical k-means clustering, a Dirichlet variational autoencoder, and invertible neural networks. For two especially challenging scenarios of jets from a dark sector we evaluate the strengths and limitations of each method.

show abstract

Section: What Is Anomalous?mentioning

confidence: 99%

What's Anomalous in LHC Jets?

Buss¹,

Dillon²,

Finke³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…We hypothesize that this is because, unlike many experimental signatures, SUEP is less complex than its QCD background, and it has smaller values of ∆R than QCD. This complexity bias has been noted before [45,52,53]. The sigmoid function reduces sensitivity to large values of ∆R and p T by mapping them to values very close to 1, while remaining approximately linear for small input values, but offset to a minimum value of 0.5.…”

Section: Unsupervised Autoencodermentioning

confidence: 71%

“…To improve on this, we utilize supervised as well as unsupervised machine learning (ML). Unsupervised analysis concepts along the lines of autoencoder neural networks [39] have the potential to transform LHC analyses [40][41][42][43][44][45][46][47][48][49][50][51][52][53][54][55][56], including searches for dark showers [45]. We point out how unsupervised methods are especially appealing for difficult-to-simulate signals like SUEP since they only rely on the known QCD background for training, while yielding at least several times greater SUEP sensitivity than the cut-and-count analysis.…”

Section: Introductionmentioning

confidence: 99%

Unsupervised hadronic SUEP at the LHC

Barron

Curtin

Kasieczka

et al. 2021

J. High Energ. Phys.

Self Cite

View full text Add to dashboard Cite

Confining dark sectors with pseudo-conformal dynamics produce SUEPs, or Soft Unclustered Energy Patterns, at colliders: isotropic dark hadrons with soft and democratic energies. We target the experimental nightmare scenario, SUEPs in exotic Higgs decays, where all dark hadrons decay promptly to SM hadrons. First, we identify three promising observables: the charged particle multiplicity, the event ring isotropy, and the matrix of geometric distances between charged tracks. Their patterns can be exploited through a cut-and-count search, supervised machine learning, or an unsupervised autoencoder. We find that the HL-LHC will probe exotic Higgs branching ratios at the per-cent level, even without a detailed knowledge of the signal features. Our techniques can be applied to other SUEP searches, especially the unsupervised strategy, which is independent of overly specific model assumptions and the corresponding precision simulations.

show abstract

“…Once a model is trained to recognize the background, one can evaluate it on unlabeled data and use the reconstruction errors as a measure of similarity of a given sample to the training data set. Typically, a model trained on QCD will return low reconstruction loss for samples similar to those it was trained on, and high loss for samples that are more complex [24,25]. In case of a search for new physics, an autoencoder would be trained and evaluated, with the largest 10%, 1%, and 0.1% of its reconstruction losses isolated and analyzed.…”

Section: Autoencodersmentioning

confidence: 99%

Autoencoders for semivisible jet detection

Canelli

Cosa

Pottier

et al. 2022

J. High Energ. Phys.

View full text Add to dashboard Cite

The production of dark matter particles from confining dark sectors may lead to many novel experimental signatures. Depending on the details of the theory, dark quark production in proton-proton collisions could result in semivisible jets of particles: collimated sprays of dark hadrons of which only some are detectable by particle collider experiments. The experimental signature is characterised by the presence of reconstructed missing momentum collinear with the visible components of the jets. This complex topology is sensitive to detector inefficiencies and mis-reconstruction that generate artificial missing momentum. With this work, we propose a signal-agnostic strategy to reject ordinary jets and identify semivisible jets via anomaly detection techniques. A deep neural autoencoder network with jet substructure variables as input proves highly useful for analyzing anomalous jets. The study focuses on the semivisible jet signature; however, the technique can apply to any new physics model that predicts signatures with anomalous jets from non-SM particles.

show abstract

Better latent spaces for better autoencoders

Cited by 52 publications

References 51 publications

What's Anomalous in LHC Jets?

What's Anomalous in LHC Jets?

Unsupervised hadronic SUEP at the LHC

Autoencoders for semivisible jet detection

Contact Info

Product

Resources

About