Comparing weak- and unsupervised methods for resonant anomaly detection

Anomaly detection techniques are growing in importance at the Large Hadron Collider (LHC), motivated by the increasing need to search for new physics in a model-agnostic way. In this work, we provide a detailed comparative study between a well-studied unsupervised method called the autoencoder (AE) and a weakly-supervised approach based on the Classification Without Labels (CWoLa) technique. We examine the ability of the two methods to identify a new physics signal at different cross sections in a fully hadronic resonance search. By construction, the AE classification performance is independent of the amount of injected signal. In contrast, the CWoLa performance improves with increasing signal abundance. When integrating these approaches with a complete background estimate, we find that the two methods have complementary sensitivity. In particular, CWoLa is effective at finding diverse and moderately rare signals while the AE can provide sensitivity to very rare signals, but only with certain topologies. We therefore demonstrate that both techniques are complementary and can be used together for anomaly detection at the LHC.

Tag N’ Train: a technique to train improved classifiers on unlabeled data

There has been substantial progress in applying machine learning techniques to classification problems in collider and jet physics. But as these techniques grow in sophistication, they are becoming more sensitive to subtle features of jets that may not be well modeled in simulation. Therefore, relying on simulations for training will lead to sub-optimal performance in data, but the lack of true class labels makes it difficult to train on real data. To address this challenge we introduce a new approach, called Tag N’ Train (TNT), that can be applied to unlabeled data that has two distinct sub-objects. The technique uses a weak classifier for one of the objects to tag signal-rich and background-rich samples. These samples are then used to train a stronger classifier for the other object. We demonstrate the power of this method by applying it to a dijet resonance search. By starting with autoencoders trained directly on data as the weak classifiers, we use TNT to train substantially improved classifiers. We show that Tag N’ Train can be a powerful tool in model-agnostic searches and discuss other potential applications.

LHC signals for KK graviton from an extended warped extra dimension

We analyze signals at the Large Hadron Collider (LHC) from production and decay of Kaluza-Klein (KK) gravitons in the context of “extended” warped extra-dimensional models, where the standard model (SM) Higgs and fermion fields are restricted to be in-between the usual ultraviolet/Planck brane and a ∼ O(10) TeV (new, “intermediate”) brane, whereas the SM gauge fields (and gravity) propagate further down to the ∼ O(TeV) infrared brane. Such a framework suppresses flavor violation stemming from KK particle effects, while keeping the KK gauge bosons and gravitons accessible to the LHC. We find that the signals from KK graviton are significantly different than in the standard warped model. This is because the usually dominant decay modes of KK gravitons into top quark, Higgs and longitudinal W/Z particles are suppressed by the above spatial separation between these two sets of particles, thus other decay channels are allowed to shine themselves. In particular, we focus on two novel decay channels of the KK graviton. The first one is the decay into a pair of radions, each of which decays (dominantly) into a pair of SM gluons, resulting in a resonant 4-jet final state consisting of two pairs of dijet resonance. On the other hand, if the radion is heavier and/or KK gluon is lighter, then the KK graviton mostly decays into a KK gluon and a SM gluon. The resulting KK gluon has a significant decay branching fraction into radion and SM gluon, thereby generating (again) a 4-jet signature, but with a different underlying event topology, i.e., featuring now three different resonances. We demonstrate that the High-Luminosity LHC (HL-LHC) has sensitivity to KK graviton of (up to) ∼ 4 TeV in both channels, in the specific model with only gluon field (and gravity) propagating in the extended bulk, whereas it is unlikely to have sensitivity in the standard dijet resonance search channel from KK graviton decay into two gluons.

Exclusion Limits on a Scalar Decaying to Photons and Distinguishing Its Production Mechanisms

LHC run-II has a great potential to search for new resonances in the diphoton channel. Latest 13 TeV data already put stringent limits on the cross sections in the diphoton channel assuming the resonance is produced through the gluon-gluon fusion. Many beyond the Standard Model (SM) theories predict TeV-scale scalars, which copiously decay to diphotons. Apart from the gluon-gluon fusion production, these scalars can also be dominantly produced in other ways too at the LHC, namely, through the quark-quark fusion or the gauge boson fusions like the photon-photon, photon-Z, WW, or ZZ fusions. In this paper we use an effective field theory approach where a heavy scalar can be produced in various ways and recast the latest ATLAS diphoton resonance search to put model-independent limits on its mass and effective couplings to the SM particles. If a new scalar is discovered at the LHC, it would be very important to identify its production mechanism in order to probe the nature of the underlying theory. We show that combining various kinematic variables in a multivariate analysis can be very powerful to distinguish different production mechanisms from one another.

Assessment Procedures for Mechanical Securing Components of LNG Cargo Containment System With Potential Damages

This research has been performed to evaluate the structural integrity of the LNG cargo containment system. The structural integrity under vibrational loading and hull deflection was evaluated for damaged and undamaged systems. The test Mock-ups had been constructed for vibrational loading tests and hull deflection tests to compare the system behaviors under the damaged and the undamaged conditions. The damaged Mock-up was constructed with particular damages at the mechanical joints of the top and the bottom boards. The damages of the joints were applied at the designated points, which were selected based on the FEM analysis that compares the stress levels for the various damaged systems. The static tests were performed when the bottom board and the top board was installed, respectively. The visual inspection had been performed to compare the structural integrity of the damaged and the undamaged system during and after the tests. The vibrational loading tests had also been performed to compare the fatigue performance of both the damaged and the undamaged system. Endurance tests were performed for both vertical and lateral directions based on the results of resonance search tests. The specimens were vibrated at the resonant frequencies determined to most seriously affect the structural integrity. Abnormality of the acceleration time history had been monitored during the endurance tests and the acceleration level at the beginning of the test was compared with the level at the end of the test. After the endurance tests, the structural integrity of both systems had been thoroughly investigated by visual inspection.