Autoencoders for unsupervised anomaly detection in high energy physics

Autoencoders are widely used in machine learning applications, in particular for anomaly detection. Hence, they have been introduced in high energy physics as a promising tool for model-independent new physics searches. We scrutinize the usage of autoencoders for unsupervised anomaly detection based on reconstruction loss to show their capabilities, but also their limitations. As a particle physics benchmark scenario, we study the tagging of top jet images in a background of QCD jet images. Although we reproduce the positive results from the literature, we show that the standard autoencoder setup cannot be considered as a model-independent anomaly tagger by inverting the task: due to the sparsity and the specific structure of the jet images, the autoencoder fails to tag QCD jets if it is trained on top jets even in a semi-supervised setup. Since the same autoencoder architecture can be a good tagger for a specific example of an anomaly and a bad tagger for a different example, we suggest improved performance measures for the task of model-independent anomaly detection. We also improve the capability of the autoencoder to learn non-trivial features of the jet images, such that it is able to achieve both top jet tagging and the inverse task of QCD jet tagging with the same setup. However, we want to stress that a truly model-independent and powerful autoencoder-based unsupervised jet tagger still needs to be developed.

Casting a graph net to catch dark showers

Strongly interacting dark sectors predict novel LHC signatures such as semi-visible jets resulting from dark showers that contain both stable and unstable dark mesons. Distinguishing such semi-visible jets from large QCD backgrounds is difficult and constitutes an exciting challenge for jet classification. In this article we explore the potential of supervised deep neural networks to identify semi-visible jets. We show that dynamic graph convolutional neural networks operating on so-called particle clouds outperform convolutional neural networks analysing jet images as well as other neural networks based on Lorentz vectors. We investigate how the performance depends on the properties of the dark shower and discuss training on mixed samples as a strategy to reduce model dependence. By modifying an existing mono-jet analysis we show that LHC sensitivity to dark sectors can be enhanced by more than an order of magnitude by using the dynamic graph network as a dark shower tagger.

Investigating the connection between γ-ray activity and the relativistic jet in 3C 273 during 2015−2019

Context. The powerful radiation over the entire electromagnetic spectrum and its radio jet activity of the blazar 3C 273 offer the opportunity of studying the physics of γ-ray emission from active galactic nuclei. Since the historically strong outburst in 2009, 3C 273 showed relatively weak emission in the γ-ray band over several years. However, recent Fermi-Large Area Telescope observations indicate higher activity during 2015−2019. Aims. We constrain the origin of the γ-ray outbursts toward 3C 273 and investigate their connection to the parsec-scale jet. Methods. We generated Fermi-LAT γ-ray light curves with multiple binning intervals and studied the spectral properties of the γ-ray emission. Using a 3 mm ALMA light curve, we studied the correlation between radio and γ-ray emission. The relevant activity in the parsec-scale jet of 3C 273 was investigated with 7 mm VLBA observations that were obtained close in time to notable γ-ray outbursts. Results. We find two prominent γ-ray outbursts in 2016 (MJD 57382) and 2017 (MJD 57883) accompanied by millimeter-wavelength flaring activity. The γ-ray photon index time series show a weak hump-like feature around the γ-ray outbursts. The monthly γ-ray flux–index plot indicates a transition from softer-when-brighter to harder-when-brighter states at 1.03 × 10−7 ph cm−2 s−1. A significant correlation between the γ-ray and millimeter-wavelength emission is found, and the radio lags the γ-rays by about 105−112 days. The 43 GHz jet images reveal the known stationary features (i.e., the core, S1, and S2) in a region upstream of the jet. We find an indication for a propagating disturbance and a polarized knot between the stationary components at about the times of the two γ-ray outbursts. Conclusions. Our results support a parsec-scale origin for the observed higher γ-ray activity, which suggests that this is associated with standing shocks in the jet.

Reports of my demise are greatly exaggerated: $N$-subjettiness taggers take on jet images

We compare the performance of a convolutional neural network (CNN) trained on jet images with dense neural networks (DNNs) trained on nn-subjettiness variables to study the distinguishing power of these two separate techniques applied to top quark decays. We find that they perform almost identically and are highly correlated once jet mass information is included, which suggests they are accessing the same underlying information which can be intuitively understood as being contained in 4-, 5-, 6-, and 8-body kinematic phase spaces depending on the sample. This suggests both of these methods are highly useful for heavy object tagging and provides a tentative answer to the question of what the image network is actually learning.

Investigation on Surface Wave Characteristic of Water Jet

Water jet surface stability is important to enhance fighting efficiency of fire water monitor. In this paper, a visualization experimental method is designed to capture the surface waves of water jet out from fire water monitor, and the wavelength and the amplitudes are captured and measured from the obtained water jet images. The Sobel horizontal gradient direction template and Burg method are selected to obtain wavelength characteristic of water jet. Based on surface morphology, the relationship between wave characteristic of water jet and Weber number is also discussed. The structure of the surface wave changes from cosmic turbulence to stochastic small-scale waves with increasing Weber number and becomes fully chaotic finally. The average wavelength of water jet out from fire water monitor decreases with increasing Weber number. The growth rate in the presence of wavelength with the lower Weg is less than that of the higher Weg. Furthermore, the amplitudes of the water jet increase continuously with increasing flow distance and Weg. In other words, the larger the Weber number is, the faster the velocity of surface waves on water jet is. The main objective of the present work is to give the basis for better understanding the microstructure of water jet, which in turn improves the performance of fire water monitor.

ATLAS Jet Reconstruction, Calibration, and Tagging of Lorentzboosted Objects

Jet reconstruction in the ATLAS detector takes multiple forms, as motivated by the intended usage of the jet. Different jet definitions are used in particular for the study of QCD jets and jets containing the hadronic decay of boosted massive particles. These different types of jets are calibrated through a series of mostly sequential steps, providing excellent uncertainties, including a first in situ calibration of the jet mass scale. Jet tagging is investigated, including both not-top-quark vs gluon discrimination as well as W/Z boson, H → bb, and top-quark identification. This includes a first look at the use of Boosted Decision Trees and Deep Neural Networks built from jet substructure variables, as well as Convolutional Neural Networks built from jet images. In all cases, these advanced techniques are seen to provide gains over the standard approaches, with the magnitude of the gain depending on the use case. Future methods for improving jet tagging are briefly discussed, including jet substructure-oriented particle flow primarily for W/Z tagging and new subjet reconstruction strategies for H → bb tagging.