SEARCH · Search NASA

Results for “CNN”

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.

SourceResearch connectionFrom yearThrough year

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

At least 127 records · Page 7

All sources NASA NTRS DOE OSTI

DOE OSTI2025

Fusion of Experiments and Simulations for Real-Time Identification of Pipeline Defects

In this study, we explored fusion of experiments and simulations for real time identification of pipeline defects across physical and non-physical domains. The challenges associated to data processing were addressed and a combined classification models was presented via CNN models. In addition, regression model based on XGBOOST is built to determine the defect location and defect dimension from data-driven features of guided wave signals captured by SMS fiber optic sensor.

deep learning↗

DOE OSTI2025

Web-Based Tools for Data-Informed Remedy Optimization: Software Theory and User Guide

This report documents the development and application of two web-based decision-support tools for pump-and-treat (P&T) groundwater remediation systems: PTOLEMY (Pump-and-Treat Optimized Location Evaluation to Maximize Yields) and OPTIMA (Optimization for Pump-and-Treat Implementation, Management, & Assessment). These tools enhance remedy design and management by leveraging advanced computational methods – specifically deep learning and multi-objective optimization – within a user-friendly platform. By integrating data-driven models with established hydrogeological knowledge, PTOLEMY and OPTIMA enable more efficient evaluation of well placement and operational strategies, helping site managers balance multiple remediation objectives under complex conditions. Both tools are implemented as modules within the SOCRATES (Suite Of Comprehensive Rapid Analysis Tools for Environmental Sites) web platform, which provides data access, visualization, and analytics to support remedy optimization across sites in the U.S. Department of Energy Office of Environmental Management complex. PTOLEMY is a rapid screening module designed to identify promising locations for new extraction wells. It employs a multi-channel three-dimensional convolutional neural network (MC3D-CNN) trained on high-fidelity simulation data to predict the relative performance (in terms of contaminant mass recovery) of potential well sites. Through an interactive web interface, PTOLEMY visualizes the probability of high performance across a site, highlighting areas where an extraction well is likely to yield above-threshold contaminant removal over a multi-year period. PTOLEMY’s map-based displays and exportable results support transparent communication of screening analyses. By focusing attention on the most favorable candidate locations, the tool augments traditional engineering judgment and physics-based modeling, providing a data informed basis for subsequent detailed evaluations. OPTIMA is a multi objective optimization module designed to find wellfield layouts and operating schedules that meet various cleanup goals. It quickly evaluates thousands of candidate setups – combinations of well locations, timing, and rates – and returns a small set of best trade-off options for comparison. At its core, OPTIMA uses a U-Net-based surrogate model – a deep-learning emulator of a groundwater flow and transport simulator – to dramatically accelerate scenario evaluations. Coupling this fast surrogate with the NSGA-II (Non-dominated Sorting Genetic Algorithm II) evolutionary algorithm, OPTIMA explores a wide decision space of well locations and schedules to identify Pareto-optimal solutions that trade off key objectives (e.g., minimizing cleanup time, maximizing contaminant mass removal, and minimizing plume extent). The tool outputs a family of optimal configurations and visualizes their trade-offs (Pareto frontiers of cleanup metrics and maps of optimized well placements). Site managers can use these results to understand the range of viable strategies and to select candidate designs for more detailed verification. OPTIMA is currently under active development and not yet fully released; this guide provides early documentation to support planning and gather user feedback.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2025

Investigating Spatial Variability of Aerosol, Cloud Condensation Nuclei, and Ice Nucleating Particles in Mountainous Terrain Field Campaign Report

The U.S. Department of Energy Atmospheric System Research (ASR)-supported Surface Atmosphere Integrated Field Laboratory (SAIL) campaign in the East River Watershed (ERW) of the Upper Colorado River Basin in southwestern Colorado ran from fall 2021 to spring 2023. Two monitoring sites were deployed in the ERW as part of SAIL. The two sites were the Aerosol Observation System (AOS) located on Crested Butte Ski Mountain, and the second ARM Mobile Facility (AMF2), located at the Rocky Mountain Biological Laboratory in Gothic, Colorado. To gain a more comprehensive understanding of aerosols in complex, mountainous terrain, Handix Scientific deployed SAIL-Net, a distributed network of six measurement nodes spanning the domain of the SAIL research area from October 2021 to July 2023. Each node measured aerosol particles between 140 nm and 3.4 μm in diameter using a small portable optical particle spectrometer (POPS; Gao et al. 2016), cloud condensation nuclei (CNN) using a miniature CCN counter (CloudPuck), and ice nucleating particles (INP) using the time-resolved aerosol filter sampler (TRAPS; Creamean et al. 2018). Our approach was similar to other studies that aimed to better characterize and understand aerosols and gas-phase pollutants using networks of lower-cost sensors (Caubel et al. 2019, Kelly et al. 2021, Asher et al. 2022). Such studies have identified neighborhood-level variations in pollutant concentrations (Schneider et al. 2017, Popoola et al. 2018, Caubel et al. 2019). Small-scale variations such as this are poorly represented in models and poorly measured by a single monitoring system (Caubel et al. 2019). Previous work has shown the representation error (the ability of measurements to represent a larger area) increases with complex orography, leading to decreases in model accuracy (Schutgens et al. 2017). The overall goal of SAIL-Net was to improve our understanding of the variability of aerosol in the ERW, thus increasing our knowledge of aerosol-cloud interactions in this region and informing the usefulness of distributed networks of measurements for future studies. We met this goal by answering the following science questions: 1. What is the aerosol temporal variability, and how does aerosol inhomogeneity vary seasonally? Is there significant seasonal variability in sources, or are short-term meteorological conditions the most important determining factor in sources for cloud nuclei? 2. What is the aerosol spatial variability? What are the aerosol characteristics at cloud base, presumably the particles most representative of those acting as cloud nuclei? 3. How should measurement networks be designed to capture aerosol-cloud interactions, and what do they need to measure? Can a single measurement site accurately represent aerosol properties in regions of complex terrain? SAIL-Net consisted of six measurement nodes spread across the ERW near Crested Butte, Colorado. The primary objective in site placement was to select locations that captured the vertical variation in aerosol properties while also spanning the domain of the SAIL campaign. The elevation of the sites ranged from roughly 2750 m along the valley floor of the ERW to approximately 3500 m near the top of Crested Butte Mountain, which is one of the taller peaks in the ERW. The farthest distance between sites was 14 km, while the closest two sites were approximately 1 km apart. Two of the sites were collocated with the ARM SAIL sites; our instruments sat on top of one of the trailers at AOS and another one of our sites was located in a meadow just above AMF2.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2026

Prong Segmentation using Point Set Transformers in Multiple View Neutrino Detectors

NOvA is a long-baseline neutrino experiment studying neutrino oscillations by detecting neutrinos from the NuMI beam at Fermilab. Its physics analysis relies on accurate prong segmentation, which involves matching each hit to its source particle and identifying the particle type. This task has commonly been addressed using a combination of traditional clustering algorithms and convolutional neural networks (CNNs). However, NOvA’s detector design presents data as two sparse and decoupled 2D images (XZ and YZ views) rather than a native 3D representation, posing a significant challenge for traditional CNN-based models. In this talk, we propose a novel neural network based on the Point Set Transformer. By treating detector hits as sparse point clouds and implementing a cross-view attention mechanism, our model enables efficient information mixing between both views. Evaluated on NOvA simulated data, our model achieves superior accuracy while requiring significantly fewer computational resources compared to other models. Furthermore, the model demonstrates great performance when applied to Liquid Argon Time Projection Chamber (LArTPC) data, which shows its potential as a universal prong segmentation algorithm for multiple view neutrino detectors.

Liu, Jiaxi [UC, Irvine]↗

DOE OSTI2026

SBND Shower Reconstruction with SPINE

The Short-Baseline Near Detector (SBND) is a liquid argon time projection chamber (LArTPC) neutrino detector in the Short-Baseline Neutrino (SBN) program at Fermilab. SBND is designed to investigate the Low-Energy Excess (LEE), an unexplained excess of electron-like events observed by previous short-baseline neutrino experiments that may point to physics beyond the Standard Model. In LArTPC detectors, precise shower reconstruction is essential for distinguishing electrons from photons, a key requirement for testing possible explanations of the LEE and improving $\nu_e$ event selection. In this poster, the reconstruction studies using the Scalable Particle Imaging with Neural Embeddings (SPINE), a machine learning based reconstruction framework for particle imaging detectors will be presented. SPINE combines sparse convolutional neural networks (CNN) and graph neural networks (GNN) to enable detailed reconstruction and characterization of neutrino interactions in LArTPC detectors. Shower calorimetry and kinematic reconstruction are performed in dedicated post-processing stages. Strong agreement between data and Monte Carlo simulation will be demonstrated, indicating high-precision detector calibration and reconstruction performance. The agreement between reconstructed and true electron shower energy will also be discussed, emphasizing the robustness of the shower reconstruction performance. These results demonstrate the unprecedented precision achievable with SPINE in SBND, highlighting their potential for future high-resolution neutrino measurements.

Fan, Castaly [Florida U.; Fermilab] (ORCID:0000000↗

DOE OSTI2024

Accuracy optimized neural networks do not effectively model optic flow tuning in brain area MSTd

Accuracy-optimized convolutional neural networks (CNNs) have emerged as highly effective models at predicting neural responses in brain areas along the primate ventral stream, but it is largely unknown whether they effectively model neurons in the complementary primate dorsal stream. We explored how well CNNs model the optic flow tuning properties of neurons in dorsal area MSTd and we compared our results with the Non-Negative Matrix Factorization (NNMF) model, which successfully models many tuning properties of MSTd neurons. To better understand the role of computational properties in the NNMF model that give rise to optic flow tuning that resembles that of MSTd neurons, we created additional CNN model variants that implement key NNMF constraints – non-negative weights and sparse coding of optic flow. While the CNNs and NNMF models both accurately estimate the observer's self-motion from purely translational or rotational optic flow, NNMF and the CNNs with nonnegative weights yield substantially less accurate estimates than the other CNNs when tested on more complex optic flow that combines observer translation and rotation. Despite its poor accuracy, NNMF gives rise to tuning properties that align more closely with those observed in primate MSTd than any of the accuracy-optimized CNNs. This work offers a step toward a deeper understanding of the computational properties and constraints that describe the optic flow tuning of primate area MSTd.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2026

Machine Learning-Enabled Wearable Piezoelectric Acoustic Sensor for Real-Time Breast Abnormality Detection

In contemporary society, breast health has become a significant public health concern, particularly among women. According to statistics from the World Health Organization, both the incidence and mortality rates of breast tumors have steadily increased in recent years. Therefore, effective early-stage screening and postoperative monitoring are essential for maintaining breast health. However, conventional clinical diagnostic modalities are typically bulky, operationally complex, and unsuitable for continuous real-time monitoring, which limits their use in portable and everyday health management applications. To address these limitations, this study proposes a machine learning-integrated wearable piezoelectric sensing platform as an auxiliary tool for breast health assessment. The device consists of a PDMS matching layer embedded with flexible silver nanowires, a P(VDF-TrFE) piezoelectric layer, and a multi-channel low-noise signal acquisition circuit. It is capable of acquiring acoustic echo signals from tissue-mimicking environments and automatically evaluating signal validity using a convolutional neural network (CNN). By integrating piezoelectric sensing with deep learning-based signal analysis, the proposed system achieves a signal-to-noise ratio exceeding 70 dB and a real-time classification accuracy above 96% under controlled conditions. These results demonstrate that the platform provides a compact, portable, and intelligent approach for wearable sensing of mechanical heterogeneity and highlight its potential for future development in continuous biomedical monitoring technologies.

He, Shuaitong↗

DOE OSTI2025

Applying Deep Learning for Wildfire Identification: Economical and Accessible Solutions Leveraging Small Datasets

Wildfires significantly impact human health, air quality, visibility, weather, and climate change and cause substantial economic losses. While state and county-operated air quality monitors provide critical insights during wildfires, they are not available in all regions. This highlights the need for affordable, accessible tools that allow the general public to assess air quality impacts. In this study, we apply machine learning with deep neural networks to diagnose air quality rapidly from sky images taken at the Pacific Northwest National Laboratory in Richland, WA, USA. Using a convolutional neural network (CNN) framework, we trained a deep learning model to classify air quality indices based on sky images. By leveraging transfer learning, our approach fine-tunes a pre-trained model on a small dataset of sky images, significantly reducing training time while maintaining high accuracy. Our results demonstrate the potential of deep learning to provide rapid air quality diagnostics during wildfire episodes, offering early warnings to the public and enabling timely mitigation strategies, particularly for vulnerable populations. Additionally, we show that lower respiratory infections pose the highest health risk during acute smoke exposures. Reactive oxygen species (ROS) from wildfire particles further exacerbate health risks by triggering inflammation and other adverse effects.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2025

Identification and Classification of Fungal GPCR Gene Families

G protein-coupled receptors (GPCRs) are transmembrane proteins crucial for signal transduction in eukaryotes, responding to diverse extracellular signals. Researchers have found and systematically summarized 14 distinct types of GPCRs in fungi but their distribution among numerous fungal species remained largely unexamined. Additionally, three families of mammalian homologs (Rhodopsin, Glutamate, and Frizzled) have been found in previous studies, but they are not included in the systematic classification of fungal GPCRs. Our study establishes a unified classification of 17 GPCR classes in fungi, combining 14 fungal and 3 mammalian previously recognized groups, and classifies 28,294 GPCRs across 1357 fungal species, significantly expanding the scale of GPCRs in fungi and demonstrating their broader distribution. We found that mammalian homologs are notably more prevalent in Early Diverging Fungi (EDF), whereas the previous 14 classes are predominantly found in Ascomycota and Basidiomycota. The most abundant class detected in fungi was Pth11-like GPCRs, exclusively found in Pezizomycotina and involved in fungal pathogenicity. Our analysis suggested that Pezizomycotina ancestor possessed an extensive array of Pth11-like GPCRs, but over time, some species underwent considerable reductions in these GPCRs in conjunction with genome contractions. Utilizing a custom-built convolutional neural network (CNN) for the identification of fungal GPCRs, we identified several putative novel fungal GPCRs. Predicted interactions between these prospective new GPCRs and G-alpha proteins, as simulated by AlphaFold Multimer, provided additional support for their functional relevance. In conclusion, our work defines the first large-scale, unified classification of fungal GPCRs, reveals lineage-specific expansions and contractions, and uncovers previously unrecognized GPCR candidates with potential functional roles in fungal signaling.

G protein-coupled receptors↗

DOE OSTI2024

Emerging Trends and Technologies Used for the Identification, Detection, and Characterisation of Plant-Parasitic Nematode Infestation in Crops

Accurate identification and estimation of the population densities of microscopic, soil-dwelling plant-parasitic nematodes (PPNs) are essential, as PPNs cause significant economic losses in agricultural production systems worldwide. This study presents a comprehensive review of emerging techniques used for the identification of PPNs, including morphological identification, molecular diagnostics such as polymerase chain reaction (PCR), high-throughput sequencing, meta barcoding, remote sensing, hyperspectral analysis, and image processing. Classical morphological methods require a microscope and nematode taxonomist to identify species, which is laborious and time-consuming. Alternatively, quantitative polymerase chain reaction (qPCR) has emerged as a reliable and efficient approach for PPN identification and quantification; however, the cost associated with the reagents, instrumentation, and careful optimisation of reaction conditions can be prohibitive. High-throughput sequencing and meta-barcoding are used to study the biodiversity of all tropical groups of nematodes, not just PPNs, and are useful for describing changes in soil ecology. Convolutional neural network (CNN) methods are necessary to automate the detection and counting of PPNs from microscopic images, including complex cases like tangled nematodes. Remote sensing and hyperspectral methods offer non-invasive approaches to estimate nematode infestations and facilitate early diagnosis of plant stress caused by nematodes and rapid management of PPNs. This review provides a valuable resource for researchers, practitioners, and policymakers involved in nematology and plant protection. It highlights the importance of fast, efficient, and robust identification protocols and decision-support tools in mitigating the impact of PPNs on global agriculture and food security.

Plant Sciences↗

DOE OSTI2025

Flood Susceptibility Mapping Using Machine Learning and Geospatial-Sentinel-1 SAR Integration for Enhanced Early Warning Systems

This study presents a comprehensive framework for flood susceptibility mapping by integrating geospatial factors with both statistical and machine learning models. Thirteen Flood-related factors, including DEM, slope, TWI, NDVI, etc., are extracted as features of models, and historical flood data derived from Sentinel-1 SAR from 2018 to 2023 are used as the target variables of the models. These datasets are analyzed using a frequency-based statistical model and three machine learning models, including Random Forest, XGBoost, and CNN, to generate flood susceptibility maps. The performance of each model is evaluated through AUC; and SHAP scores are separately generated for Machine learning (ML) models to explain each feature contribution in the ML model. The generated susceptibility maps are validated by high-flood-risk locations monitored by flood sensors, BLE inundation models, and flood-prone areas suggested by the Local Community Task Force. The results indicate that the XGBoost model outperforms all other models, with an AUC of 0.92 and demonstrates the highest alignment with recommended high-flood-risk locations, while the frequency-based statistical model showed the weakest performance with an AUC of 0.65. SHAP value graphs highlight the elevation, slope, and TWI as the most influential features across all models. The susceptibility maps generated by the machine learning model show strong agreement with the BLE map and high-flood-risk areas identified by the local Community Task Force.

Google Engine↗

DOE OSTI2026

Quasars Acting as Strong Lenses Found in DESI DR1

Quasars acting as strong gravitational lenses offer a rare opportunity to probe the redshift evolution of scaling relations between supermassive black holes and their host galaxies, particularly the M$_{BH}$–M$_{host}$ relation. Using these powerful probes, the mass of the host galaxy can be precisely inferred from the Einstein radius θ$_{E}$. Using 812,118 quasars from DESI DR1 (0.03 ≤ z ≤ 1.8), we searched for quasars lensing higher-redshift galaxies by identifying background emission-line features in their spectra. To detect these rare systems, we trained a convolutional neural network (CNN) on mock lenses constructed from real DESI spectra of quasars and emission-line galaxies (ELGs), achieving a high classification performance (AUC = 0.99). We also trained a regression network to estimate the redshift of the background ELG. Applying this pipeline, we identified seven high-quality (Grade A) lens candidates, each exhibiting a strong [O II] doublet at a higher redshift than the foreground quasar; four candidates additionally show Hβ, [O III] λ4959, and [O III] λ5007 emission. These results significantly expand the sample of quasar lens candidates beyond the 12 identified and 3 confirmed in previous work and demonstrate the potential for scalable, data-driven discovery of quasars as strong lenses in upcoming spectroscopic surveys.

McArthur, Everett [Stanford U., Phys. Dept.; KIPAC↗

DOE OSTI2026

RU Net for Automatic Characterization of TRISO Fuel Cross Sections

TRistructural ISOtropic (TRISO) particle fuel is a type of nuclear fuel known for its high-temperature and high-burnup performance. Each sub-millimeter diameter TRISO particle consists of uranium-oxycarbide (UCO) or UO2 fuel kernel, coated with buffer, inner pyrolytic carbon (IPyC), silicon carbide (SiC), and outer pyrolytic carbon (OPyC) layers. The SiC layer acts as the main containment barrier for the TRISO particle to retain the fission products, while the IPyC and OPyC layers provide additional barriers to the release of fission products, especially fission gases. During irradiation, phenomena like kernel swelling, buffer densification, and IPyC fracture may impact fuel performance. Post-irradiation microscopy on entire compact cross sections or samples of individual particles deconsolidated from compacts is often used to identify these irradiation-induced changes in morphology. However, each fuel compact generally contains thousands of TRISO particles. To get statistical information on these phenomena, it is cumbersome work if done manually. For example, to get information about swelling/densification behaviors of different layers or kernels after irradiation, researchers previously manually measured the perimeter of each TRISO layer in hundreds of particles after four rounds of iterative grinding and polishing encompassing more than 2000 cross-section images for a total of four fuel compacts. To attempt to reduce the subjectivity inherent in that process and accelerate data analysis, we conducted a study on the automatic TRISO layer segmentation on cross-sectional microscopic images using Convolutional Neural Networks (CNNs). CNNs are a class of machine learning algorithms specifically designed for processing structured grid data that have gained popularity in recent years due to their remarkable performance in various computer vision tasks, including image classification, object detection, and image segmentation. In this research, we have generated the large irradiated TRISO layer dataset with more than 2000 cross-section TRISO microscopic images and the corresponding annotated images. Based on these annotated images, we have employed different CNNs for automatic segmentation of different TRISO layers. These include RU-Net (developed in this study), as well as three existing architectures: U-Net, Residual Network (ResNet), and Attention U-Net. The preliminary results show that the model based on RU-Net has the best performance in terms of intersection-over-union (IoU). Through the aid of these CNN models, we can expedite the analysis of TRISO particle cross-sections, significantly reducing the manual labor involved and improving the objectivity of the segmentation results.

Convolutional Neural Networks↗

DOE OSTI2025

Testing convolutional neural network based deep learning systems: a statistical metamorphic approach

Machine learning technology spans many areas and today plays a significant role in addressing a wide range of problems in critical domains,i.e., healthcare, autonomous driving, finance, manufacturing, cybersecurity,etc. Metamorphic testing (MT) is considered a simple but very powerful approach in testing such computationally complex systems for which either an oracle is not available or is available but difficult to apply. Conventional metamorphic testing techniques have certain limitations in verifying deep learning-based models (i.e., convolutional neural networks (CNNs)) that have a stochastic nature (because of randomly initializing the network weights) in their training. In this article, we attempt to address this problem by using a statistical metamorphic testing (SMT) technique that does not require software testers to worry about fixing the random seeds (to get deterministic results) to verify the metamorphic relations (MRs). We propose seven MRs combined with different statistical methods to statistically verify whether the program under test adheres to the relation(s) specified in the MR(s). We further use mutation testing techniques to show the usefulness of the proposed approach in the healthcare space and test two CNN-based deep learning models (used for pneumonia detection among patients). The empirical results show that our proposed approach uncovers 85.71% of the implementation faults in the classifiers under test (CUT). Furthermore, we also propose an MRs minimization algorithm for the CUT, thus saving computational costs and organizational testing resources.

Computer Science↗

DOE OSTI2024

Machine Learning for DUNE Supernova Trigger

One of the major scientific goals of the Deep Underground Neutrino Experiment (DUNE) is to detect and measure the neutrino flux originating from galactic core-collapse supernovae. These neutrinos provide an opportunity to study the end of life evolution of massive stars, and reveal information about the structure of core-collapse that is not visible in observations of the electromagnetic spectrum. Because of the rarity of these events, it is crucial that DUNE is able to detect supernova neutrino interactions when they occur. However, this will require sifting through a large quantity of data, motivating the development of a trigger algorithm to identify significant events and discard irrelevant data. Machine learning provides a potential approach to building this trigger. This project generates ADC and ground truth images of simulated neutrino interactions in a LArTPC detector to be used for machine learning, and uses them to train a sparse Convolutional Neural Network (C NN). The performance of this model when applied to the task of pixel classification based on interaction type is examined. This project found that the sparse CNN approach has the potential to have high accuracy in pixel classification, meaning it may be highly relevant to the development of a supernova neutrino trigger for the DUNE far detector.

Damish, S.↗

DOE OSTI2024

A Method for Producing Hierarchical and Statistically Calibrated Predictions of Nuclear Material Properties from Existing Models

Computer vision-based analysis of micrographs of nuclear materials is an emerging technique for property prediction, synthetic route identification, and other material analysis tasks. These analysis tasks play a pivotal role in many material characterization applications such as signature development for treaty verification, process optimization, etc. The backbone in many of the recent computer vision-based techniques is a deep learning model, which takes a fixed-size set of pixels and provides a class prediction for that set of pixels. For example, previous work developed a deep convolutional neural network (CNN) to predict the synthetic route from a 256 px x 256 px patch taken from a larger image of uranium ore concentrates. In this work, we present several methods for first calibrating these models in a manner that they can provide accurate probabilities of their predictions’ veracity, and several methods of combining these probabilities. Overall, the combination of these two steps into a pipeline allows for full-image and even full-sample (where a sample has many images) predictions with associated confidence values. Finally, we show that one can also use the patch predictions and confidence to produce a visualization to map predicted constituents through the image. Results and examples for predicting and mapping uranium ore concentrates’ synthetic process from imagery will be presented.

artificial intelligence↗

DOE OSTI2024

Symmetry-Based Structured Matrices for Efficient Approximately Equivariant Networks

There has been much recent interest in designing symmetry-aware neural networks (NNs) exhibiting relaxed equivariance. Such NNs aim to interpolate between being exactly equivariant and being fully flexible, affording consistent performance benefits. In a separate line of work, certain structured parameter matrices -- those with displacement structure, characterized by low displacement rank (LDR) -- have been used to design small-footprint NNs. Displacement structure enables fast function and gradient evaluation, but permits accurate approximations via compression primarily to classical convolutional neural networks (CNNs). In this work, we propose a general framework -- based on a novel construction of symmetry-based structured matrices -- to build approximately equivariant NNs with significantly reduced parameter counts. Our framework integrates the two aforementioned lines of work via the use of so-called Group Matrices (GMs), a forgotten precursor to the modern notion of regular representations of finite groups. GMs allow the design of structured matrices -- resembling LDR matrices -- which generalize the linear operations of a classical CNN from cyclic groups to general finite groups and their homogeneous spaces. We show that GMs can be employed to extend all the elementary operations of CNNs to general discrete groups. Further, the theory of structured matrices based on GMs provides a generalization of LDR theory focussed on matrices with cyclic structure, providing a tool for implementing approximate equivariance for discrete groups. We test GM-based architectures on a variety of tasks in the presence of relaxed symmetry. We report that our framework consistently performs competitively compared to approximately equivariant NNs, and other structured matrix-based compression frameworks, sometimes with a one or two orders of magnitude lower parameter count.

Samudre, Ashwin↗

DOE OSTI2025

Fusion of Experiments and Simulations for Real-Time Identification of Pipeline Defects

deep learning↗