Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “classification models”

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 847 records · Page 47
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Optimized allocation method of the VSC-MTDC system for frequency regulation reserves considering the cost

In recent years, the interconnection of asynchronous power grids through the VSC-MTDC system has been proposed and extensively studied in light of the potential benefits of economical bulk power exchanges and frequency regulation reserves sharing. This paper proposed an optimized allocation method for sharing frequency regulation reserves among the interconnected power systems and the corresponding frequency regulation control of the VSC-MTDC system under emergency frequency deviation events. Firstly, the frequency regulation reserve classification is proposed. In the classification, the available frequency response capacity reserves of each interconnection are divided into commercial reserves and regular reserves. While the commercial reserves are procured through long-term contracts, the regular reserves are purchased based on market prices of frequency regulation services. Secondly, based on the proposed frequency regulation reserve classification, a novel frequency regulation control is then introduced for the VSC-MTDC system. This control method could minimize the costs of the disturbed power grid for the needed frequency response supports from the other power grids. Simulation verifications are performed on a modified IEEE 39 bus system and a highly reduced power system model representing the North American grids. The simulation verification indicates that the developed frequency regulation control significantly reduced ancillary service costs of the disturbed power grid.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2025

Evaluating algorithmic bias on biomarker classification of breast cancer pathology reports

Objectives: This work evaluated algorithmic bias in biomarkers classification using electronic pathology reports from female breast cancer cases. Bias was assessed across 5 subgroups: cancer registry, race, Hispanic ethnicity, age at diagnosis, and socioeconomic status. Materials and Methods: We utilized 594 875 electronic pathology reports from 178 121 tumors diagnosed in Kentucky, Louisiana, New Jersey, New Mexico, Seattle, and Utah to train 2 deep-learning algorithms to classify breast cancer patients using their biomarkers test results. We used balanced error rate (BER), demographic parity (DP), equalized odds (EOD), and equal opportunity (EOP) to assess bias. Results: We found differences in predictive accuracy between registries, with the highest accuracy in the registry that contributed the most data (Seattle Registry, BER ratios for all registries >1.25). BER showed no significant algorithmic bias in extracting biomarkers (estrogen receptor, progesterone receptor, human epidermal growth factor receptor 2) for race, Hispanic ethnicity, age at diagnosis, or socioeconomic subgroups (BER ratio <1.25). DP, EOD, and EOP all showed insignificant results. Discussion: We observed significant differences in BER by registry, but no significant bias using the DP, EOD, and EOP metrics for socio-demographic or racial categories. This highlights the importance of employing a diverse set of metrics for a comprehensive evaluation of model fairness. Conclusion: A thorough evaluation of algorithmic biases that may affect equality in clinical care is a critical step before deploying algorithms in the real world. We found little evidence of algorithmic bias in our biomarker classification tool. Artificial intelligence tools to expedite information extraction from clinical records could accelerate clinical trial matching and improve care.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2024

Learning From User Behavior: A Survey-Assist Algorithm for Longitudinal Mobility Data Collection

GPS-based travel surveys are widely used in mobility studies to gather crucial qualitative data, like purpose, transportation mode and replaced mode. However, survey response still poses a burden to users, especially in long-term mobility studies, leading to response fatigue. We explore a survey-assist strategy to ease this burden by a novel, user-level modeling approach that leverages past responses from each user to predict responses for new trips, without relying on external data sources like GIS data. We investigate three main algorithms for predicting responses: (i) clustering trips and extrapolating responses for similar trips, (ii) using random forest classification, and (iii) clustering that uses a hybrid algorithm to determine spatial structure, which is then fed as input to a classic random forest classifier. The clustering approach can flexibly predict responses for even complex qualitative survey questions; it achieved F-scores of 65%. The random forest pipeline uses architecture that restricts it to predicting three predetermined survey questions: trip purpose, mode, and replaced mode. However, it achieved F-scores of 78%. While the survey-assist approach has been implemented by several proprietary systems, to our knowledge, this is the first exploration in the academic literature. It follows that this is also the first rigorous evaluation of multiple algorithms that can implement the approach. The evaluation uses a large scale, publicly available, longitudinal dataset consisting of ~ 92k trips from 235 users over a period of roughly one and a half years. With this approach, travel surveys can be pre-filled with the predicted responses for each trip, thus streamlining the survey process for users. Combined with an active learning system that requests user input on low-confidence predictions, models can be updated and improved over time to better support the long-term collection of longitudinal qualitative data.

clustering↗
DOE OSTI2025

How does ion temperature gradient turbulence depend on magnetic geometry? Insights from data and machine learning

Magnetic geometry has a significant effect on the level of turbulent transport in fusion plasmas. Here, we model and analyse this dependence using multiple machine learning methods and a dataset of >200 000 nonlinear gyrokinetic simulations of ion-temperature-gradient turbulence in diverse non-axisymmetric geometries. The dataset is generated using a large collection of both optimised and randomly generated stellarator equilibria. At fixed gradients and other input parameters, the turbulent heat flux varies between geometries by several orders of magnitude. Trends are apparent among the configurations with particularly high or particularly low heat flux. Regression and classification techniques from machine learning are then applied to extract patterns in the dataset. Due to a symmetry of the gyrokinetic equation, the heat flux and regressions thereof should be invariant to translations of the raw features in the parallel coordinate, similar to translation invariance in computer vision applications. Multiple regression models including convolutional neural networks (CNNs) and decision trees can achieve reasonable predictive power for the heat flux in held-out test configurations, with highest accuracy for the CNNs. Using Spearman correlation, sequential feature selection and Shapley values to measure feature importance, it is consistently found that the most important geometric lever on the heat flux is the flux surface compression in regions of bad curvature. The second most important geometric feature relates to the magnitude of geodesic curvature. These two features align remarkably with surrogates that have been proposed based on theory, while the methods here allow a natural extension to more features for increased accuracy. The dataset, released with this publication, may also be used to test other proposed surrogates, and we find that many previously published proxies do correlate well with both the heat flux and stability boundary.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2020

Learning to Count Grave Sites for Cemetery Observation Models With Satellite Imagery

Understanding how people occupy open spaces is important for research in support of population modeling, policy, national security, emergency response, and sustainability. For the past decade, there has been an increase in research toward capturing and reporting population dynamics and patterns of life at the building level and in some open public spaces such as cemeteries and parks. This is done through observation models developed from local sociocultural information acquired at various spatiotemporal scales to inform night, day, and episodic population occupancy estimates (people/1000 sq ft). Sociocultural information for cemeteries and parks is scarcely available and often collected manually. The process is not only marred by inconsistencies but is laborious and time consuming. In this study, we leverage convolutional neural networks (CNNs) and satellite imagery to derive grave site counts as proxy variables to support scalable and accurate sociocultural data required in a population observation model. Through a hybrid workflow (weak localization plus regression model), we characterize a large scale automation process to counting of grave sites. We evaluate and demonstrate the efficacy of proposed workflow using out-of-data set large satellite imagery and establish its broader impact on cemetery observation models.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2024

Towards Precision Photometric Type ia Supernova Cosmology With Machine Learning

The revolutionary discovery of dark energy and accelerating cosmic expansion was made with just 42 type Ia supernovae (SNe Ia) in 1999. Since then, large synoptic surveys, e.g., Dark Energy Survey (DES), have observed thousands more SNe Ia and the upcoming Rubin Legacy Survey of Space and Time (LSST) and Roman Space Telescope promise to deliver millions in the next decade. This unprecedented data volume can produce the required precision to unambiguously test concordance cosmology which could represent a monumental shift in our understanding of dark energy and its role in cosmic history. However, extracting a pure SN Ia sample with accurate redshifts for such a large dataset will be a challenge. Specifically, spectroscopic classification will not be possible for the vast majority of discovered objects, and only ~25% will have spectroscopic redshifts. This thesis presents a series of observational and methodological studies designed to address the questions associated with this new era of photometric SN Ia cosmology. First, we present a machine learning method for photometric classification of SNe, Supernova Classification with a COnvolutional Neural Network (SCONE). Photometric classification enables SNe with no spectroscopic information to be confidently categorized, and is a critical component of current and future analysis pipelines. SCONE achieves >99% accuracy distinguishing simulated SNe Ia from non-Ia SNe, and has been integrated into DES, LSST, and Roman analysis pipelines. We also demonstrate the efficacy of SCONE on early-time photometric classification, which will be vital for optimal allocation of spectroscopic resources. We show that SCONE can distinguish between 6 SN types with 75% accuracy on the night of initial discovery, comparable to results in the literature for full-phase SNe. Next, we study current methods for estimating SN Ia redshifts and propose a machine learning alternative that uses SN photometry alone to extract redshift information. Most SNe Ia inherit redshift information from their host galaxy, but the process of matching SNe to the correct host galaxy can be challenging. We systematically analyze the impact of incorrect redshifts from host galaxy mismatch on 5 years of DES SN data, and conclude that improved host matching or redshift estimation methods can reduce our systematic errors by ~10%. In response to this finding, we present a SN photometry-only method for estimating redshifts independent of host galaxy information, Photo-zSNthesis. We show that Photo-zSNthesis redshift estimates are accurate to within 2% across the full redshift range of LSST, a first in the literature. Finally, we focus on the robustness of machine learning (ML) algorithms for real-world and scientific applications. ML models generalize poorly beyond their training set and often experience severe performance degradation when deployed on new data. We demonstrate a general method for improving robustness that achieves new state-of-the-art results on astronomical object classification, wildlife identification, and tumor detection.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Comparison of Machine Learning and Deep Learning for View Identification from Cardiac Magnetic Resonance Images

Background: Artificial intelligence is increasingly utilized to aid in the interpretation of cardiac magnetic resonance (CMR) studies. One of the first steps is the identification of the imaging plane depicted, which can be achieved by both deep learning (DL) and classical machine learning (ML) techniques without user input. We aimed to compare the accuracy of ML and DL for CMR view classification and to identify potential pitfalls during training and testing of the algorithms. Methods: To train our DL and ML algorithms, we first established datasets by retrospectively selecting 200 CMR cases. The models were trained using two different cohorts (passively and actively curated) and applied data augmentation to enhance training. Once trained, the models were validated on an external dataset, consisting of 20 cases acquired at another center. We then compared accuracy metrics and applied class activation mapping (CAM) to visualize DL model performance. Results: The DL and ML models trained with the passively-curated CMR cohort were 99.1% and 99.3% accurate on the validation set, respectively. However, when tested on the CMR cases with complex anatomy, both models performed poorly. After training and testing our models again on all 200 cases (active cohort), validation on the external dataset resulted in 95% and 90% accuracy, respectively. The CAM analysis depicted heat maps that demonstrated the importance of carefully curating the datasets to be used for training. Conclusions: Both DL and ML models can accurately classify CMR images, but DL outperformed ML when classifying images with complex heart anatomy.

artificial intelligence↗
DOE OSTI2021

Comparing weak- and unsupervised methods for resonant anomaly detection

Abstract 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.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2025

Proposed Classifications of Remote Operations for Nuclear Reactors Based on Physical and Cybersecurity Considerations

The incorporation of remote operations into reactor operations is a topic of high interest among advanced and small modular reactor (A/SMR) vendors, with some considering it essential to the success of their business models. However, remote operations are a concept novel to the nuclear industry. While various technical aspects of remote operations have been explored, a significant gap remains in understanding the security implications of integrating remote operations into reactor designs, particularly concerning the security requirements for remote-operations facilities and infrastructure. This report aims to address this gap by first defining classes of remote operation based on the extent of remote access to reactor control systems and grounded in the existing regulatory framework with compatible terminology. Secondly, the report outlines the physical and cybersecurity requirements applicable to remote-operations facilities and infrastructure at each defined class. These requirements are based on existing licensing frameworks provided by 10 Code of Federal Regulations (CFR) Part 50 and 10 CFR Part 52, as well as the upcoming A/SMR licensing framework in the proposed Part 53. The assessment focuses specifically on security regulations, such as 10 CFR Part 73, which includes provisions for both cybersecurity (§ 73.54) and physical security (§ 73.55). This report proposes five classes of remote reactor operations. Class 1 involves remote monitoring only, with no control over reactor systems. Class 2 allows for the remote issuance of allowlisted commands to the reactor facility. Class 3 extends control to non-safety-significant, non-safety-related, or not important to safety systems and equipment. Class 4 permits remote control of safety-significant systems. Finally, Class 5 allows remote control of safety-related systems. It is important to note that these classes were defined purely with functionality in mind, without considering the practicality or feasibility of implementation for each class under current or upcoming regulatory guidance. The intention behind this approach is to enable an assessment of which security requirements apply to each class, allowing readers to evaluate the implementation possibilities for their specific use cases. Following the definition of remote-operation classes, the report assesses the specific physical and cybersecurity requirements applicable to the remote-operations facility and infrastructure within each defined class. This includes defining the types and locations of operators that are possible at each class of operation and, based on operator type and location, as well as functionality within each class, outlining the physical and cybersecurity requirements. By detailing the security requirements by class, the report provides readers with the information needed to determine the type of security program they may need to implement for their desired concept of operation. The next contribution of this report was to assess the practicality of implementing each proposed class of remote operations based upon the security requirement assessment. In short, three of the five proposed remote-operation classes were found to possibly have a practical path forward to implementation under the U.S. regulatory framework. Class 1 remote operations are currently in use in the U.S. while Class 2 and 3 remote operations may be logistically possible to implement under the U.S. regulatory framework. The final two Classes, 4 and 5, would likely be logistically difficult, if not infeasible to implement within the current U.S. physical- and cybersecurity regulatory framework. Given the results of the feasibility assessment, an example architecture is proposed for both Class 2, remote allowlisted commands, and Class 3, remote control of non-safety systems as well as security implication assessments of each architecture. These example implementations are not meant to be prescriptive in terms of how Class 2 or Class 3 remote operations should be deployed; instead, they are intended to be informative to stakeholders on how Class 2 or Class 3 could potentially be applied in order to inform their system design. An example architecture for Class 1 remote monitoring was not provided as Class 1 in already in use in U.S. nuclear operations. Example architectures for Class 4 and Class 5 were not provided due to their assessment of being likely infeasible to implement. The final contribution is an assessment of the physical- and cybersecurity implications of introducing autonomous operations into an A/SMR. What was found was that the security implications can be separated into two cases. Autonomous operations supported by SSCs located only at the reactor site, and autonomous operations supported by SSCs outside of the reactor site. For the first case, the introduction of autonomous systems will likely not change the facility’s requirement to comply with existing cyber and physical security regulation

22 - GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2021

Interpreting the Diurnal Cycle of Clouds and Precipitation in the ARM GoAmazon Observations: Shallow to Deep Convection Transition

The Green Ocean Amazon (GoAmazon) 2014/5 field campaign data are used to study the diurnal cycle of clouds and precipitation. Through a careful classification of days with shallow cumulus, congestus and deep convection, we investigate the major differences among locally generated convection regimes and the most important environmental factors governing the shallow-to-deep convection transition. On shallow cumulus days, a greater sensible heat flux drives deeper boundary layer growth, which entrains drier free-tropospheric air and lowers the relative humidity, thus leading to a significantly higher cloud base than those on days with deeper convection. Congestus and deep convection regimes exhibit distinct cloud top height distributions with noticeable differences in the vertical wind shear in the mid-troposphere, suggesting an important role of wind shear in limiting the vertical extent of convection. On deep convection days, with preexisting nocturnal convection or cold-pools from external disturbances, the timing of peak surface precipitation (12:00–13:00 LST) tends to be in-phase with the diurnal variation in surface fluxes. However, it takes longer for local deep convection to develop without these disturbances. A plume model with thermodynamic and dynamical constraints is developed to explore the relative importance of various convection-controlling factors. Initial cloud-base vertical velocity and buoyancy are important in helping parcels ascend to the level of free convection (LFC). After parcels reach the LFC, entrainment of environmental air and lower free troposphere humidity become crucial in determining cloud top. Entrainment rate differentiates among convection regimes, which may be tied to the cloud size distribution at cloud base.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

Multivariate regression modelling for gender prediction using volatile organic compounds from hand odor profiles via HS-SPME-GC-MS

The efficacy of using human volatile organic compounds (VOCs) as a form of forensic evidence has been well demonstrated with canines for crime scene response, suspect identification, and location checking. Although the use of human scent evidence in the field is well established, the laboratory evaluation of human VOC profiles has been limited. This study used Headspace-Solid Phase Microextraction-Gas Chromatography-Mass Spectrometry (HS-SPME-GC-MS) to analyze human hand odor samples collected from 60 individuals (30 Females and 30 Males). The human volatiles collected from the palm surfaces of each subject were interpreted for classification and prediction of gender. The volatile organic compound (VOC) signatures from subjects’ hand odor profiles were evaluated with supervised dimensional reduction techniques: Partial Least Squares-Discriminant Analysis (PLS-DA), Orthogonal-Projections to Latent Structures Discriminant Analysis (OPLS-DA), and Linear Discriminant Analysis (LDA). The PLS-DA 2D model demonstrated clustering amongst male and female subjects. The addition of a third component to the PLS-DA model revealed clustering and minimal separation of male and female subjects in the 3D PLS-DA model. The OPLS-DA model displayed discrimination and clustering amongst gender groups with leave one out cross validation (LOOCV) and 95% confidence regions surrounding clustered groups without overlap. The LDA had a 96.67% accuracy rate for female and male subjects. The culminating knowledge establishes a working model for the prediction of donor class characteristics using human scent hand odor profiles.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Dual X-ray computed tomography-aided classification of melt pool boundaries and flaws in crept additively manufactured parts

In metal additive manufacturing (AM), understanding the process-structure-performance relationships requires a combination of multi-scale characterization techniques that allows for the measurement of the melt pool shape and boundary and classifying various defects and flaws in the AM parts. Such approaches can be destructive, only 2D in nature, or have a small field of view and can be complex to co-register and analyze. Here, in this work, we present a non-destructive 3D inspection technique that employs dual-energy X-ray computed tomography (XCT) along with a model-based iterative reconstruction (MBIR) and a new segmentation algorithm. The proposed approach and algorithm are not only capable of classifying and quantifying flaws such as pores, cracks, and inclusions, but they also allow for the extraction of microstructural features such as melt pool boundaries (MPB) and melt pool regions (MPR), that can help understand process-structure-performance relationships for alloys under study. As an exemplar application, we employed the method for characterization of an additively manufactured aluminum alloy crept under tensile stress at 300 °C for 1064 h. Our results demonstrate high quality segmentation and classification of various flaws and MPB and MPR, for the first time, using 3D X-ray CT inspection. The delineated MPB and MPR in the crept samples reveal the preferential growth paths of cracks that formed during creep deformation. The technique was used for successfully quantifying the characteristics (number of defects, their density, volume fraction, etc.) of the manufacturing-induced pores and creep-induced cracks, which is necessary to better understand the creep failure mechanisms of the material.

36 MATERIALS SCIENCE↗
DOE OSTI2020

An Information Theory Approach on Deciding Spectroscopic Follow-ups

Classification and characterization of variable phenomena and transient phenomena are critical for astrophysics and cosmology. These objects are commonly studied using photometric time series or spectroscopic data. Given that many ongoing and future surveys are conducted in a time domain, and given that adding spectra provides further insights but requires more observational resources, it would be valuable to know which objects we should prioritize to have a spectrum in addition to a time series. We propose a methodology in a probabilistic setting that determines a priori which objects are worth taking a spectrum of to obtain better insights, where we focus on the insight of the type of the object (classification). Objects for which we query their spectrum are reclassified using their full spectral information. We first train two classifiers, one that uses photometric data and another that uses photometric and spectroscopic data together. Then for each photometric object we estimate the probability of each possible spectrum outcome. We combine these models in various probabilistic frameworks (strategies), which are used to guide the selection of follow-up observations. The best strategy depends on the intended use, whether it is obtaining more confidence or accuracy. For a given number of candidate objects (127, equal to 5% of the data set) for taking spectra, we improve the class prediction accuracy by 37% as opposed to 20% of a non-naive (non-random) best-baseline strategy. Our approach provides a general framework for follow-up strategies and can be extended beyond classification to include other forms of follow-ups beyond spectroscopy.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2020

Are we modeling the properties of unconventional shales correctly?

Technological advancements have led to impressive growth in hydrocarbon (HC) production from unconventional shale reservoirs. However, major limiting factors in shale gas development are i) low recovery of gas in place (<20%), (ii) a rapid decline in well productivity, iii) release of contaminants, despite the use of advanced hydraulic fracturing fluids and multiple fracturing stages. This is mainly due to a lack of understanding of the nanoscale physicochemical properties of shales, especially that of kerogen, the macromolecule that is not only the source of the majority of the HC’s in shales but also holds most of the HC’s in adsorbed state. Over the last few years, a large number of studies have been published on simulating the physicochemical properties of shale utilizing molecular models of kerogen. However, the molecular models of kerogen input into these simulations are based on the kerogen “type” derived from a very limited number of shale samples. In this paper, we examine the variations that can exist in kerogen structure within a particular kerogen “type” across different shale basins and single shale basin at similar thermal maturity levels. We propose that using kerogen “type” based structural models for molecular simulations could lead to inaccurate estimation of HC reserves, HC recovery, HC production, frackability, and quantity and quality of produced water. Our study highlights the need for developing a better classification of kerogen based on its “molecular structure” instead of “type” for a more accurate prediction of physicochemical properties of shales.

04 OIL SHALES AND TAR SANDS↗
DOE OSTI2023

Identifying common stored product insects using automated deep learning methods

Monitoring stored product insect pests is a common practice for post-harvest management of stored grain and grain-based commodities, which helps ensure product quality from harvest to final consumer. Current methods of sampling and monitoring can be time-consuming, labor-intensive, expensive and require expertise in insect identification. Therefore, this study aims to develop an image-based automated identification system for common stored product insect species using deep-learning methods. Top-down images of the common stored product adult insect species of Rhyzopertha dominica, Cryptolestes ferrugineus, Tribolium castaneum, Sitophilus oryzae, and Oryzaephilus surinamensis were acquired and analyzed. Deep learning-based, state-of-the-art Convolutional Neural Networks (CNN) models (ResNet-50, MobileNet-v2, DarkNet-53, and EfficientNet-b0) were fine-tuned with a transfer learning approach to classify the insect species. All models were able to correctly identify the insect species with at least 96% accuracy and with few misclassifications. One issue with trained CNNs is that they do not explain the reasoning for the classification and are often called a “black box”. Therefore, visualization methods called Gradient-weighted Class Activation Mapping (Grad-CAM) were implemented to explore the black box network. The Grad-CAM uses heat maps to highlight the major image features that the network focused on to make insect species predictions. The Grad-CAM verifies the network's prediction and also helps improve network performance. This study contributes to the overall goal of developing a camera-based system for monitoring stored grain insects. As a result, the developed system would empower warehouse, flour mills, and other food facilities with a tool to quickly and accurately identify insect species in stored product environments and could be implemented as part of a close to real-time monitoring system.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2022

Dictionary Learning with Accumulator Neurons

The Locally Competitive Algorithm (LCA) uses local competition between non-spiking leaky integrator neurons to infer sparse representations, allowing for potentially real-time execution on massively parallel neuromorphic architectures such as Intel's Loihi processor. Here, we focus on the problem of inferring sparse representations from streaming video using dictionaries of spatiotemporal features optimized in an unsupervised manner for sparse reconstruction. Non-spiking LCA has previously been used to achieve unsupervised learning of spatiotemporal dictionaries composed of convolutional kernels from raw, unlabeled video. We demonstrate how unsupervised dictionary learning with spiking LCA (\hbox{S-LCA}) can be efficiently implemented using accumulator neurons, which combine a conventional leaky-integrate-and-fire (\hbox{LIF}) spike generator with an additional state variable that is used to minimize the difference between the integrated input and the spiking output. We demonstrate dictionary learning across a wide range of dynamical regimes, from graded to intermittent spiking, for inferring sparse representations of both static images drawn from the CIFAR database as well as video frames captured from a DVS camera. On a classification task that requires identification of the suite from a deck of cards being rapidly flipped through as viewed by a DVS camera, we find essentially no degradation in performance as the LCA model used to infer sparse spatiotemporal representations migrates from graded to spiking. We conclude that accumulator neurons are likely to provide a powerful enabling component of future neuromorphic hardware for implementing online unsupervised learning of spatiotemporal dictionaries optimized for sparse reconstruction of streaming video from event based DVS cameras.

artificial intelligence↗
DOE OSTI2024

Machine Learning-Based Identification of the Interface Regions for Coupling Local and Nonlocal Models

Local-nonlocal coupling approaches provide a means to combine the computational efficiency of local models and the accuracy of nonlocal models. However, the coupling process can be challenging, requiring expertise to identify the interface between local and nonlocal regions. Here, this study introduces a machine learning-based approach to automatically detect the regions in which the local and nonlocal models should be used. The method uses loading functions evaluated at grid points to decide the model selection at those points. Training of the networks is based on datasets provided by classes of loading functions for which reference coupling configurations are computed using accurate coupled solutions, where accuracy is measured in terms of the relative error between the solution to the coupling approach and the solution to the nonlocal model. We study two approaches that vary in data structure. The first, the full-domain input data approach, uses the entire load vector and outputs a complete label vector, performing a global classification. The second, a window-based approach, processes loads into windows and addresses the problem as a node-wise classification where each window's central point is classified individually. The classification problems are solved via deep learning algorithms based on convolutional neural networks. The performance of these approaches is studied on one-dimensional numerical examples using F1-scores and accuracy metrics. Notably, the windowing approach achieves an accuracy of 0.96 and an F1-score of 0.97, highlighting its potential to automate coupling processes effectively and enhance computational efficiency in material science applications.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Validation of the NOνA experiment 2023-tuning on simulated neutrino-matter interactions

NOνA is a long-baseline neutrino oscillation experiment that utilizes a two-detector design to study the oscillations of muon neutrinos into electron neutrinos over a baseline of 810 km. The Near Detector (ND) measures the neutrino beam spectrum and composition before oscillation, which is then compared to the oscillated neutrino energy spectrum observed in the Far Detector (FD). In the ND, the neutrinos are detected through their interactions with the heavy target nuclei within the detector. NOνA employs the GENIE neutrino event generator for simulating these neutrino-nucleus interactions. However, the default GENIE prediction does not adequately reproduce the ND data. To address this, NOνA developed a tune of the neutrino interaction models within GENIE version 3:0:6 to minimize discrepancies between the simulated predictions and the observed data in the ND. This dissertation tests the NOνA’s 2023 tune of the GENIE neutrino cross-section simulations by performing a data/simulations comparison for the ND. The analysis employed datasets comprising $2.55\times10^{21}$ protons-on-target (POT) in neutrino beam mode and $1.14\times10^{21}$ POT in antineutrino beam mode. The NOνA tuning of neutrino-matter interaction simulations matches with ND data within the $1\sigma$ error band, overestimating muon neutrino and antineutrino charged current interactions by approximately 6 % and 9 %, respectively. Discrepancies were observed in the energy region dominated by Quasi-Elastic-like interactions. Systematic uncertainties associated with the modeling of the neutrino cross-section, particularly those pertaining Quasi-Elastic like interactions, contributed considerably to the overall error in the simulations. Furthermore, the reconstruction algorithm used in NOνA for particle classification demonstrated significant misidentifications between charged pions and protons, as well as a tendency to overlook additional pions or protons in multi-particle simulated events.

Cortés Parra, Camilo Andrés↗