Search for: All records

Self-supervised learning (SSL) is at the core of training modern large machine learning models, providing a scheme for learning powerful representations that can be used in a variety of downstream tasks. However, SSL strategies must be adapted to the type of training data and downstream tasks required. We propose resimulation-based self-supervised representation learning (RS3L), a novel simulation-based SSL strategy that employs a method of to drive data augmentation for contrastive learning in the physical sciences, particularly, in fields that rely on stochastic simulators. By intervening in the middle of the simulation process and rerunning simulation components downstream of the intervention, we generate multiple realizations of an event, thus producing a set of augmentations covering all physics-driven variations available in the simulator. Using experiments from high-energy physics, we explore how this strategy may enable the development of a foundation model; we show how RS3L pretraining enables powerful performance in downstream tasks such as discrimination of a variety of objects and uncertainty mitigation. In addition to our results, we make the RS3L dataset publicly available for further studies on how to improve SSL strategies. Published by the American Physical Society2025 

Abstract We report a gravitational-wave parameter estimation algorithm,AMPLFI, based on likelihood-free inference using normalizing flows. The focus ofAMPLFIis to perform real-time parameter estimation for candidates detected by machine-learning based compact binary coalescence search,Aframe. We present details of our algorithm and optimizations done related to data-loading and pre-processing on accelerated hardware. We train our model using binary black-hole (BBH) simulations on real LIGO-Virgo detector noise. Our model has $\sim 6$million trainable parameters with training times $\lesssim 24$h. Based on online deployment on a mock data stream of LIGO-Virgo data,Aframe+AMPLFIis able to pick up BBH candidates and infer parameters for real-time alerts from data acquisition with a net latency of $\sim 6$s. 

Discoveries of new phenomena often involve a dedicated search for a hypothetical physics signature. Recently, novel deep learning techniques have emerged for anomaly detection in the absence of a signal prior. However, by ignoring signal priors, the sensitivity of these approaches is significantly reduced. We present a new strategy dubbed Quasi Anomalous Knowledge (QUAK), whereby we introduce alternative signal priors that capture some of the salient features of new physics signatures, allowing for the recovery of sensitivity even when the alternative signal is incorrect. This approach can be applied to a broad range of physics models and neural network architectures. In this paper, we apply QUAK to anomaly detection of new physics events at the CERN Large Hadron Collider utilizing variational autoencoders with normalizing flow. 

Improving the quantity and quality of STEM teachers is one of the goals of the Noyce Scholarship Grant Program implemented through the National Science Foundation. Here, we examine how the change to remote learning impacted ASU'’s Noyce Scholarship program. Our research question was as follows: what kinds of successes and struggles did STEM Noyce (a) in-service and (b) pre-service scholars experience in the first year of remote teaching during the pandemic? We interviewed three different cohorts of Noyce scholars: preservice teachers, first year teachers and second year teachers. Results showed that some scholars saw benefits to remote learning including more free time for other learning activities. They also observed teachers dedicated to student success. Challenges included an inability to reach all students, a continued focus on standardized tests, and a lack of flexibility within the districts. 

A<sc>bstract</sc> The measurements of the Higgs boson (H) production cross sections performed by the CMS Collaboration in the four-lepton (4ℓ, ℓ= e,μ) final state at a center-of-mass energy$$\sqrt{s}$$= 13.6 TeV are presented. These measurements are based on data collected with the CMS detector at the CERN LHC in 2022, corresponding to an integrated luminosity of 34.7 fb⁻¹. Cross sections are measured in a fiducial region closely matching the experimental acceptance, both inclusively and differentially, as a function of the transverse momentum and the absolute value of the rapidity of the four-lepton system. The H → ZZ → 4ℓinclusive fiducial cross section is measured to be$${2.89}_{-0.49}^{+0.53}{\left({\text{stat}}\right)}_{-0.21}^{+0.29}\left({\text{syst}}\right)$$fb, in agreement with the standard model expectation of$${3.09}_{-0.24}^{+0.27}$$fb. 

A measurement of the Higgs boson mass and width via its decay to two $Z$bosons is presented. Proton-proton collision data collected by the CMS experiment, corresponding to an integrated luminosity of $138{\mathrm{fb}}^{-1}$at a center-of-mass energy of 13 TeV, is used. The invariant mass distribution of four leptons in the on-shell Higgs boson decay is used to measure its mass and constrain its width. This yields the most precise single measurement of the Higgs boson mass to date, $125.04\pm 0.12\mathrm{GeV}$, and an upper limit on the width ${\mathrm{\Gamma }}_H<330\mathrm{MeV}$at 95% confidence level. A combination of the on- and off-shell Higgs boson production decaying to four leptons is used to determine the Higgs boson width, assuming that no new virtual particles affect the production, a premise that is tested by adding new heavy particles in the gluon fusion loop model. This result is combined with a previous CMS analysis of the off-shell Higgs boson production with decay to two leptons and two neutrinos, giving a measured Higgs boson width of ${3.0}_{-1.5}^{+2.0}\mathrm{MeV}$, in agreement with the standard model prediction of 4.1 MeV. The strength of the off-shell Higgs boson production is also reported. The scenario of no off-shell Higgs boson production is excluded at a confidence level corresponding to 3.8 standard deviations. © 2025 CERN, for the CMS Collaboration2025CERN 

A<sc>bstract</sc> A search for heavy, long-lived, charged particles with large ionization energy loss within the silicon tracker of the CMS experiment is presented. A data set of proton-proton collisions at a center of mass energy at$$ \sqrt{s} $$ $\sqrt{s}$= 13 TeV, collected in 2017 and 2018 at the CERN LHC, corresponding to an integrated luminosity of 101 fb⁻¹, is used in this analysis. Two different approaches for the search are taken. A new method exploits the independence of the silicon pixel and strips measurements, while the second method improves on previous techniques using ionization to determine a mass selection. No significant excess of events above the background expectation is observed. The results are interpreted in the context of the pair production of supersymmetric particles, namely gluinos, top squarks, and tau sleptons, and of the Drell-Yan pair production of fourth generation (τ′) leptons with an electric charge equal to or twice the absolute value of the electron charge (e). An interpretation of a Z’ boson decaying to twoτ′ leptons with an electric charge equal to 2eis presented for the first time. The 95% confidence upper limits on the production cross section are extracted for each of these hypothetical particles.