Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “multiclass”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 37 records · Page 2

Performance Evaluation of Vertical Federated Machine Learning Against Adversarial Threats on Wide-Area Control System: Preprint

Federated machine learning (FL) is gaining significant popularity to develop cybersecurity solutions in power grids because of its advanced capability to support decentralized data handing at local devices, its privacy preservation, and its low-bandwidth requirement. However, the evolving adversarial machine learning (AML) threats raise significant concerns for the cybersecurity of FL architectures. The FL-based split neural network (SplitNN) achieves high performance through the decentralized training of local neural network models while preserving data privacy across multiple entities. In this paper, we propose a methodology for evaluating the performance of a vertical FLbased anomaly detector against different types of AML attacks, including denial-of-service attacks, adversarial data injection attacks, and replay attacks on the trained local models deployed in the grid network. For a case study, we consider the modified IEEE 13-bus system, and we develop SplitNN-based binary and multiclass classification models to detect, locate, and identify different types of data integrity attacks on the volt-watt control with two pooling layers: maximum pooling and AvgPool. Our experimental results, computed through performance metrics, reveal that the severity of these AML attacks varies with the integrated pooling mechanism, the type of classification model, and the nature of the cyberattack. Further, the AML attacks negatively impacted the prediction time per sample for the pretrained SplitNN during the online testing.

adversarial threats↗

Measurements of inclusive and differential cross sections for top quark production in association with a Z boson in proton-proton collisions at $\sqrt{s} $ = 13 TeV

Measurements are presented of inclusive and differential cross sections for Z boson associated production of top quark pairs ($ \textrm{t}\overline{\textrm{t}}\textrm{Z} $) and single top quarks (tZq or tWZ). The data were recorded in proton-proton collisions at a center-of-mass energy of 13 TeV, corresponding to an integrated luminosity of 138 fb$^{−1}$. Events with three or more leptons, electrons or muons, are selected and a multiclass deep neural network is used to separate three event categories, the $ \textrm{t}\overline{\textrm{t}}\textrm{Z} $ and tWZ processes, the tZq process, and the backgrounds. A profile likelihood approach is used to unfold the differential cross sections, to account for systematic uncertainties, and to determine the correlations between the two signal categories in one global fit. The inclusive cross sections for a dilepton invariant mass between 70 and 110 GeV are measured to be 1.14 ± 0.07 pb for the sum of $ \textrm{t}\overline{\textrm{t}}\textrm{Z} $ and tWZ, and 0.81 ± 0.10 pb for tZq, in good agreement with theoretical predictions.[graphic not available: see fulltext]

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Measurement of the cross section of top quark-antiquark pair production in association with a W boson in proton-proton collisions at $ \sqrt{s} $ = 13 TeV

The production of a top quark-antiquark pair in association with a W boson ($ \textrm{t}\overline{\textrm{t}}\textrm{W} $) is measured in proton-proton collisions at a center-of-mass energy of 13 TeV. The analyzed data was recorded by the CMS experiment at the CERN LHC and corresponds to an integrated luminosity of 138 fb$^{−1}$. Events with two or three leptons (electrons and muons) and additional jets are selected. In events with two leptons, a multiclass neural network is used to distinguish between the signal and background processes. Events with three leptons are categorized based on the number of jets and of jets originating from b quark hadronization, and the lepton charges. The inclusive $ \textrm{t}\overline{\textrm{t}}\textrm{W} $ production cross section in the full phase space is measured to be 868 ± 40(stat) ± 51(syst) fb. The $ \textrm{t}\overline{\textrm{t}}\textrm{W} ^{+}$ and $ \textrm{t}\overline{\textrm{t}}\textrm{W} ^{−}$ cross sections are also measured as 553 ± 30(stat) ± 30(syst) and 343 ± 26(stat) ± 25(syst) fb, respectively, and the corresponding ratio of the two cross sections is found to be $ 1.61\pm 0.15{\left(\textrm{stat}\right)}_{-0.05}^{+0.07}\left(\textrm{syst}\right) $. The measured cross sections are larger than but consistent with the standard model predictions within two standard deviations, and represent the most precise measurement of these cross sections to date.[graphic not available: see fulltext]

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

A Method to Recognize Anatomical Site and Image Acquisition View in X-ray Images

A method was developed to recognize anatomical site and image acquisition view automatically in 2D X-ray images that are used in image-guided radiation therapy. The purpose is to enable site and view dependent automation and optimization in the image processing tasks including 2D-2D image registration, 2D image contrast enhancement, and independent treatment site confirmation. The X-ray images for 180 patients of six disease sites (the brain, head-neck, breast, lung, abdomen, and pelvis) were included in this study with 30 patients each site and two images of orthogonal views each patient. A hierarchical multiclass recognition model was developed to recognize general site first and then specific site. Each node of the hierarchical model recognized the images using a feature extraction step based on principal component analysis followed by a binary classification step based on support vector machine. Given two images in known orthogonal views, the site recognition model achieved a 99% average F1 score across the six sites. If the views were unknown in the images, the average F1 score was 97%. If only one image was taken either with or without view information, the average F1 score was 94%. The accuracy of the site-specific view recognition models was 100%.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

HIV drug resistance during antiretroviral therapy scale-up in Uganda, 2012–19: a population-based, longitudinal study

Background With scale-up of antiretroviral therapy (ART) in sub-Saharan Africa, increasing pretreatment HIV drug resistance has been reported; however, the broader effect of ART expansion on population-level resistance patterns remains insufficiently quantified. We aimed to estimate the longitudinal prevalence of drug resistance and resistance-conferring mutations. Methods This study used data collected as part of the Rakai Community Cohort Study (RCCS), an open population-based census and cohort study conducted in southern Uganda. At each survey round, residents aged 15–49 years are invited to participate and receive a structured questionnaire that obtains sociodemographic, behavioural, and health information, including self-reported past and current ART use. Voluntary HIV testing is conducted using a rapid test algorithm and a venous blood sample. People with HIV provide samples for viral load quantification and deep sequencing. We analysed RCCS survey, HIV viral load, and deep sequencing (which was used to predict resistance) data from five survey rounds. The key outcomes were the population prevalence of viraemic people with HIV with non-nucleoside reverse transcriptase inhibitor (NNRTI), nucleoside reverse transcriptase inhibitor (NRTI), protease inhibitor, or multiclass resistance among all participants (regardless of HIV serostatus) in the 2015 and 2017 surveys. Prevalence of class-specific resistance and resistance-conferring substitutions were estimated using robust log-Poisson regression. Findings Between Aug 10, 2011, and Nov 4, 2020, there were 43 361 participants in the RCCS and 7923 (18·27%) people with HIV. Over five survey rounds, 93 622 participant visits occurred, among which 17 460 (18·65%) were from people with HIV. Over the analysis period, the median age of study participants remained similar (28 years [22–35] in 2012 and 29 years [21–38] in 2019). Sufficient data were available to reliably genotype 4072 (90·03%) of 4523 participant visits from 3407 people with HIV for at least one drug. Overall population prevalence of resistance contributed by viraemic pretreatment people with HIV decreased between 2012 and 2017 from 0·56% (95% CI 0·42–0·75) to 0·25% (0·18–0·33) for NNRTI and from 0·24% (0·15–0·37) to 0·05% (0·02–0·10) for NRTI (prevalence ratio 0·44 [0·29–0·68] for NNRTI and 0·21 [0·09–0·47] for NRTI). Between 2012 and 2017, NNRTI resistance among viraemic pretreatment people with HIV increased from 4·86% (3·69–6·42) to 9·61% (7·27–12·7; prevalence ratio 1·98 [1·34–2·91]). The prevalence of NNRTI and NRTI resistance was substantially higher among viraemic treatment-experienced people with HIV (51·49% [46·24–57·34] for NNRTI and 36·46% [30·06–44·22] for NRTI in 2017) than among pretreatment people with HIV. NNRTI and NRTI resistance was predominantly attributable to rtK103N and rtM184V. inT97A was observed at a similar prevalence among viraemic treatment-experienced (9·96% [6·41–15·48]) and viraemic pretreatment (10·56% [8·01–13·93]) people with HIV; no major dolutegravir resistance mutations were observed. Interpretation Despite rising NNRTI resistance among pretreatment people with HIV, overall population prevalence of pretreatment HIV drug-resistant viraemia decreased due to increasing ART uptake and viral suppression. This finding underscores the crucial role of achieving and maintaining high ART coverage in reducing transmission of drug-resistant HIV. The high prevalence of mutations conferring resistance to components of first-line ART regimens among viraemic people with HIV is potentially concerning. Funding National Institutes of Health, Johns Hopkins University Center for AIDS Research, Bill & Melinda Gates Foundation, and the US Centers for Disease Control and Prevention.

59 BASIC BIOLOGICAL SCIENCES↗

Levenberg–Marquardt multi-classification using hinge loss function

Incorporating higher-order optimization functions, such as Levenberg-Marquardt (LM) have revealed better generalizable solutions for deep learning problems. However, these higher-order optimization functions suffer from very large processing time and training complexity especially as training datasets become large, such as in multi-view classification problems, where finding global optima is a very costly problem. To solve this issue, we develop a solution for LM-enabled classification with, to the best of knowledge first-time implementation of hinge loss, for multiview classification. Hinge loss allows the neural network to converge faster and perform better than other loss functions such as logistic or square loss rates. Here we prove our method by experimenting with various multiclass classification challenges of varying complexity and training data size. The empirical results show the training time and accuracy rates achieved, highlighting how our method outperforms in all cases, especially when training time is limited. Our paper presents important results in the relationship between optimization and loss functions and how these can impact deep learning problems.

97 MATHEMATICS AND COMPUTING↗

Search for flavour-changing neutral-current couplings between the top quark and the photon with the ATLAS detector at s = 13 TeV

This letter documents a search for flavour-changing neutral currents (FCNCs), which are strongly suppressed in the Standard Model, in events with a photon and a top quark with the ATLAS detector. The analysis uses data collected in pp collisions at TeV during Run 2 of the LHC, corresponding to an integrated luminosity of 139 fb –1 . Both FCNC top-quark production and decay are considered. The final state consists of a charged lepton, missing transverse momentum, a b-tagged jet, one high-momentum photon and possibly additional jets. A multiclass deep neural network is used to classify events either as signal in one of the two categories, FCNC production or decay, or as background. No significant excess of events over the background prediction is observed and 95% CL upper limits are placed on the strength of left- and right-handed FCNC interactions. The 95% CL bounds on the branching fractions for the FCNC top-quark decays, estimated (expected) from both top-quark production and decay, are B(t → uγ) < 0.85 (0.88$^{+0.37}_{-0.25}$) × 10 –5 and B(t → cγ) < 4.2 (3.40$^{+1.35}_{-0.95}$) × 10 –5 for a left-handed tqγ coupling, and B(t → uγ) < 1.2 (1.20$^{+0.50}_{-1.33}$) ×10 –5 and B(t → cγ) < 4.5 (3.70$^{+1.47}_{-1.03}$) × 10 –5 for a right-handed coupling.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Resilience-based Recovery Scheduling of Transportation Network in Mixed Traffic Environment: A Deep-Ensemble-Assisted Active Learning Approach

Devising effective post-hazard recovery strategies is critical in enhancing the resilience of transportation networks (TNs). However, existing work does not consider the multiclass users’ travel behavior in network functionality quantification and the metaheuristic solution procedures often suffer from extensive computational burden due to the exploration need in large solution space and the expensive functionality quantification. Here, this study develops a bilevel decision-making framework for the resilience-based recovery scheduling of the TN in a mixed traffic environment with connected and autonomous vehicles (CAVs) and human-driven vehicles (HDVs). The lower level quantifies the TN's functionality over time considering different travel behavior of CAV and HDV users arisen from their different levels of information perception. The upper level presents a novel deep-ensemble-assisted active learning approach to balance optimization performance and computational cost. This framework can help decision makers better quantify the TN's functionality to support effective recovery scheduling of TN with different mixed traffic scenarios ranging from HDV-only to future CAV-dominant traffic. The optimization approach bears the potential to be extended to solving general large-scale network recovery scheduling problems effectively and efficiently. The proposed methodology is demonstrated using a real-world traffic network in Southern California under earthquake considering deterministic and stochastic repair durations.

42 ENGINEERING↗

A unified large language model–based framework for heterogeneous PV image diagnosis

With advances in imaging technologies, modern photovoltaic (PV) systems generate large volumes of heterogeneous image data, including visible, electroluminescence (EL), and infrared (IR) images. Existing PV image analysis models, particularly deep learning approaches, are typically task-specific and lack cross-modality generalization. To address this limitation, this paper proposes an open-source large language model (LLM)–based unified framework for heterogeneous PV image diagnostics. Through task-aware diagnostic prompting, the framework enables analysis of visible, EL, and IR images within a single pipeline, supporting both zero-shot and few-shot inference and binary and multiclass classification. It is compatible with state-of-the-art multimodal LLMs, including ChatGPT, Gemini, Claude, Qwen, and CLIP. The framework is evaluated on PV module condition classification (clean, soiling, snow, hail, and bird droppings) using visible images, cell crack detection using EL images, and hotspot detection using IR images. GPT-5.1 in few-shot mode achieves the best performance, with classification accuracy exceeding 97.3%. Open-source models such as Qwen and CLIP also deliver competitive results on visible images (around 90% accuracy), though their performance is more limited on EL and IR modalities. On the full ELPV dataset, the framework achieves 83.5% zero-shot accuracy, within 2.8% of the supervised CNN baseline, confirming scalability to larger benchmarks. Practical aspects such as reproducibility, response latency, and confidence estimation are systematically analyzed. The framework operates across PV image modalities without modality- or task-specific training, making it well suited as a rapid pre-screening tool to support downstream detailed diagnostics. A benchmark dataset of diverse labeled PV images is also released.

Li, Baojie↗

Toxicological relationships between proteins obtained from protein target predictions of large toxicity databases

The combination of models for protein target prediction with large databases containing toxicological information for individual molecules allows the derivation of 'toxiclogical' profiles, i.e., to what extent are molecules of known toxicity predicted to interact with a set of protein targets. To predict protein targets of drug-like and toxic molecules, we built a computational multiclass model using the Winnow algorithm based on a dataset of protein targets derived from the MDL Drug Data Report. A 15-fold Monte Carlo cross-validation using 50% of each class for training, and the remaining 50% for testing, provided an assessment of the accuracy of that model. We retained the 3 top-ranking predictions and found that in 82% of all cases the correct target was predicted within these three predictions. The first prediction was the correct one in almost 70% of cases. A model built on the whole protein target dataset was then used to predict the protein targets for 150 000 molecules from the MDL Toxicity Database. We analysed the frequency of the predictions across the panel of protein targets for experimentally determined toxicity classes of all molecules. This allowed us to identify clusters of proteins related by their toxicological profiles, as well as toxicities that are related. Literature-based evidence is provided for some specific clusters to show the relevance of the relationships identified.

60 APPLIED LIFE SCIENCES↗

Evaluating Limits of Machine Learning-Assisted Raman Spectroscopy in Classification of Biological Samples

Machine learning (ML)-assisted Raman spectroscopy has become a powerful analytical tool for the classification and identification of analytes; however, technical challenges impacting its detection accuracy have not been thoroughly investigated. This study explores experimental factors affecting classification performance. Among the evaluated ML models, ML algorithms show minimal impact on classification accuracy. Instead, experimental factors, including spectral similarity between tested samples and data quality, dominate detection performance. Increases in spectral noise and spectral similarity significantly reduce classification accuracy. In well-controlled samples with low experimental noise, ML-assisted Raman spectroscopy can discriminate lipid mixtures with a composition difference of 1.85 mol %. To assess the effect of biological heterogeneity, we analyzed single-cell Raman spectra from Saccharomyces cerevisiae strains carrying single, double, or triple gene mutations. Intrinsic cell-to-cell variability introduced substantial spectral differences, severely reducing the accuracy of multiclass classification of these genetically similar strains at the single-cell level. Averaging Raman spectra across multiple cells improved classification accuracy by reducing this spectral variability. We also assess the effectiveness of transfer learning across different Raman spectrometers, specifically by applying an ML model trained on one instrument to another Raman spectrometer. Transfer learning can be improved with proper instrument calibration, highlighting the importance of instrument standardization. Overall, our results demonstrate that data quality and spectral similarity are the primary bottlenecks in ML-assisted Raman spectroscopy. Careful attention to sample preparation, data acquisition, measurement conditions, and instrument calibration is critical to achieving robust and reliable classification performance.

Fungi↗

Monitoring Fracture Saturation With Internal Seismic Sources and Twin Neural Networks

Seismic coda-wave analysis is a well-developed method for detecting subtle physical changes in complex media by measuring arrival times in the late-arriving energy from multiply scattered or reflected waves. However, a challenge arises when multiply scattered waves are not sufficiently separated in time from the direct arrivals to provide a clear coda wave train. Additional complications for monitoring changes in fracture systems arise when the signals originate from unsynchronized internal sources, such as natural or induced seismicity, from acoustic emission, or from transportable intra-fracture sources (chattering dust), that generate uncontrolled signals that vary in arrival time, amplitude and frequency content. Here, we use a twin neural network (TNN also known as a Siamese neural network) for dimensionality reduction to analyze signals from chattering dust to classify the fluid saturation state of a synthetic fracture system. The TNN with shared weights generates a low-dimensional representation of the data input by minimizing contrastive loss, serving as the input to a multiclass classifier that accurately classifies whether multiple fractures in a fracture system are fully saturated or partially saturated, or whether a change in saturation has occurred in different fractures in the system. Furthermore, these results show that information buried in unresolved codas from uncontrolled sources can be extracted using machine learning to monitor the evolution of fracture systems caused by physical and chemical processes even when the scattered and direct wave fields overlap.

58 GEOSCIENCES↗

Classifier based on support vector machine for JET plasma configurations

The last flux surface can be used to identify the plasma configuration of discharges. For automated recognition of JET configurations, a learning system based on support vector machines has been developed. Each configuration is described by 12 geometrical parameters. A multiclass system has been developed by means of the one-versus-the-rest approach. Results with eight simultaneous classes (plasma configurations) show a success rate close to 100%.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Organ-delimited gene regulatory networks provide high accuracy in candidate transcription factor selection across diverse processes

Organ-specific gene expression datasets that include hundreds to thousands of experiments allow the reconstruction of organ-level gene regulatory networks (GRNs). However, creating such datasets is greatly hampered by the requirements of extensive and tedious manual curation. Here, we trained a supervised classification model that can accurately classify the organ-of-origin for a plant transcriptome. This K-Nearest Neighbor-based multiclass classifier was used to create organ-specific gene expression datasets for the leaf, root, shoot, flower, and seed in Arabidopsis thaliana . A GRN inference approach was used to determine the: i. influential transcription factors (TFs) in each organ and, ii. most influential TFs for specific biological processes in that organ. These genome-wide, organ-delimited GRNs (OD-GRNs), recalled many known regulators of organ development and processes operating in those organs. Importantly, many previously unknown TF regulators were uncovered as potential regulators of these processes. As a proof-of-concept, we focused on experimentally validating the predicted TF regulators of lipid biosynthesis in seeds, an important food and biofuel trait. Of the top 20 predicted TFs, eight are known regulators of seed oil content, e.g., WRI1, LEC1, FUS3. Importantly, we validated our prediction of MybS2, TGA4, SPL12, AGL18, and DiV2 as regulators of seed lipid biosynthesis. We elucidated the molecular mechanism of MybS2 and show that it induces purple acid phosphatase family genes and lipid synthesis genes to enhance seed lipid content. This general approach has the potential to be extended to any species with sufficiently large gene expression datasets to find unique regulators of any trait-of-interest.

09 BIOMASS FUELS↗

CaloChallenge 2022: a community challenge for fast calorimeter simulation

Here, we present the results of the ‘Fast Calorimeter Simulation Challenge 2022’—the CaloChallenge. We study state-of-the-art generative models on four calorimeter shower datasets of increasing dimensionality, ranging from a few hundred voxels to a few tens of thousand voxels. The 31 individual submissions span a wide range of current popular generative architectures, including variational autoencoders (VAEs), generative adversarial networks (GANs), normalizing flows, diffusion models, and models based on conditional flow matching. We compare all submissions in terms of quality of generated calorimeter showers, as well as shower generation time and model size. To assess the quality we use a broad range of different metrics including differences in one-dimensional histograms of observables, KPD/FPD scores, AUCs of binary classifiers, and the log-posterior of a multiclass classifier. The results of the CaloChallenge provide the most complete and comprehensive survey of cutting-edge approaches to calorimeter fast simulation to date. In addition, our work provides a uniquely detailed perspective on the important problem of how to evaluate generative models. As such, the results presented here should be applicable for other domains that use generative AI and require fast and faithful generation of samples in a large phase space.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Single-trial classification of evoked responses to auditory tones using OPM- and SQUID-MEG

Abstract Objective. Optically pumped magnetometers (OPMs) are emerging as a near-room-temperature alternative to superconducting quantum interference devices (SQUIDs) for magnetoencephalography (MEG). In contrast to SQUIDs, OPMs can be placed in a close proximity to subject’s scalp potentially increasing the signal-to-noise ratio and spatial resolution of MEG. However, experimental demonstrations of these suggested benefits are still scarce. Here, to compare a 24-channel OPM-MEG system to a commercial whole-head SQUID system in a data-driven way, we quantified their performance in classifying single-trial evoked responses. Approach. We measured evoked responses to three auditory tones in six participants using both OPM- and SQUID-MEG systems. We performed pairwise temporal classification of the single-trial responses with linear discriminant analysis as well as multiclass classification with both EEGNet convolutional neural network and xDAWN decoding. Main results. OPMs provided higher classification accuracies than SQUIDs having a similar coverage of the left hemisphere of the participant. However, the SQUID sensors covering the whole helmet had classification scores larger than those of OPMs for two of the tone pairs, demonstrating the benefits of a whole-head measurement. Significance. The results demonstrate that the current OPM-MEG system provides high-quality data about the brain with room for improvement for high bandwidth non-invasive brain–computer interfacing.

Iivanainen, Joonas (ORCID:0000000160344604)↗

Calibrated particle identification for Belle II

We present several efforts aimed at improving charged particle identification at the Belle II experiment. We define an ablation test to quantify and evaluate the impact of each sub-detector on the global particle identification performance. We demonstrate that the performance of the identification scheme can be improved via a simple calibration of the sub-detector likelihoods through a set of per hypothesis, per sub-detector weights. Finally, we present preliminary results on an improved definition of the likelihoods contributed by the electromagnetic calorimeter. A set of multiclass boosted decision trees is trained to exploit the shape of energy depositions of different particle species. In simulated $B\bar{B}$ events, the pion-to-electron and muon fake rates are reduced by 55% and 31% respectively at low-medium momentum.

Hohmann, Marcel [Univ. of Melbourne, Parkville, VI↗

A deep neural network to search for new long-lived particles decaying to jets

A tagging algorithm to identify jets that are significantly displaced from the proton-proton (pp) collision region in the CMS detector at the LHC is presented. Displaced jets can arise from the decays of long-lived particles (LLPs), which are predicted by several theoretical extensions of the standard model. The tagger is a multiclass classifier based on a deep neural network, which is parameterised according to the proper decay length $\mathrm{c}\tau_0$ of the LLP. A novel scheme is defined to reliably label jets from LLP decays for supervised learning. Samples of pp collision data, recorded by the CMS detector at a centre-of-mass energy of 13 TeV, and simulated events are used to train the neural network. Domain adaptation by backward propagation is performed to improve the simulation modelling of the jet class probability distributions observed in pp collision data. The potential performance of the tagger is demonstrated with a search for long-lived gluinos, a manifestation of split supersymmetric models. The tagger provides a rejection factor of 10 000 for jets from standard model processes, while maintaining an LLP jet tagging efficiency of 30-80% for gluinos with 1 mm $\leq$ $c\tau_0$ $\leq$ 10 m. The expected coverage of the parameter space for split supersymmetry is presented.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗