Search NASA⌕ Search

SEARCH · Search NASA

Results for “classification”

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 289 records · Page 16
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Radioisotope Identification with List-Mode Gamma-Ray Data

This work explores the potential of utilizing temporal data from gamma-ray detectors, known as list-mode data, to enhance radioisotope identification. Traditional identification methods, which rely on full gamma-ray spectrum analysis, often require long dwell times and struggle with spectra containing similarly spaced spectral peaks. We hypothesize that by leveraging the probabilistic nature of nuclear decay and the time-encoded information from decay sequences and interactions with surrounding materials, we can improve classification accuracy over static spectral analysis. This research examines the temporal content of list-mode data through exploratory data analysis via correlation discovery and qualitative distribution analysis. Additionally, we propose a probabilistic classification model that can utilize spectral data, temporal data, or both to determine if the incorporation of temporal information improves radioisotope identification. Our findings suggest that the temporal information present in list-mode gamma-ray data has merit and should be further investigated to develop more robust and optimal methods for utilizing this temporal information in applications requiring radioisotope identification.

List-mode data↗
DOE OSTI2025

Particle hit clustering and identification using point set transformers in liquid argon time projection chambers

Liquid argon time projection chambers are often used in neutrino physics and dark-matter searches because of their high spatial resolution. The images generated by these detectors are extremely sparse, as the energy values detected by most of the detector are equal to 0, meaning that despite their high resolution, most of the detector is unused in a particular interaction. Instead of representing all of the empty detections, the interaction is usually stored as a sparse matrix, a list of detection locations paired with their energy values. Traditional machine learning methods that have been applied to particle reconstruction such as convolutional neural networks (CNNs), however, cannot operate over data stored in this way and therefore must have the matrix fully instantiated as a dense matrix. Operating on dense matrices requires a lot of memory and computation time, in contrast to directly operating on the sparse matrix. We propose a machine learning model using a point set neural network that operates over a sparse matrix, greatly improving both processing speed and accuracy over methods that instantiate the dense matrix, as well as over other methods that operate over sparse matrices. Compared to competing state-of-the-art methods, our method improves classification performance by 14%, segmentation performance by more than 22%, while taking 80% less time and using 66% less memory. Compared to state-of-the-art CNN methods, our method improves classification performance by more than 86%, segmentation performance by more than 71%, while reducing runtime by 91% and reducing memory usage by 61%.

calibration and fitting methods↗
DOE OSTI2024

Reliability and resilience of environmental flows under uncertainty: reconsidering water year types and inconsistent flow requirements in California

Abstract Environmental water allocation in California is a complex legal process involving various government agencies and stakeholders. E-flow requirements can be based on annual runoff typologies called water year types (WYTs), which dictate water volume, timing, and duration. In this study, we examined hydropower licensing documents of the major water and power projects in the Central Sierra Nevada to catalog e-flow requirements by WYT. In this study case, we identify how WYT classification systems and categories vary across and within different basins. Additionally, we assessed the impacts of climate change on hydrology, the frequency of WYTs identified, and the reliability and resilience of e-flows using future projections (2031–2060) of 10 Global Circulation Models (GCMs). We then propose a potential adaptation strategy using a 30 year moving percentiles approach to recalculate WYTs. We identified eight WYT classifications systems were identified, and their WYT distributions statistically significantly changes across all GCMs, even though most GCMs indicate no statistically significant change in hydrology. Disparities in future impacts are observed among and within hydropower projects, with some river reaches showing negative impacts on reliability and resilience. The adaptation strategy can generally boost resilience and improve reliability, but simply updating existing WYT thresholds without flexible regulatory frameworks reconsidering WYTs and e-flows thresholds, may not yield substantial improvements. Challenges in managing e-flows in California within regulatory and hydroclimatic contexts are intricate due to the lack of standardized approaches, leading to inconsistencies and potential conflicts among stakeholders, that will likely be exacerbated by climate change. Thus, we emphasize that targeted, site-specific, and adaptive management strategies are crucial, besides the need for a harmonized and consistent approach to defining and applying WYT categories and methods and/or e-flow assessments.

Environmental Sciences & Ecology↗
DOE OSTI2024

Real-time confinement regime detection in fusion plasmas with convolutional neural networks and high-bandwidth edge fluctuation measurements

Abstract A real-time detection of the plasma confinement regime can enable new advanced plasma control capabilities for both the access to and sustainment of enhanced confinement regimes in fusion devices. For example, a real-time indication of the confinement regime can facilitate transition to the high-performing wide-pedestal (WP) quiescent H-mode, or avoid unwanted transitions to lower confinement regimes that may induce plasma termination. To demonstrate real-time confinement regime detection, we use the 2D beam emission spectroscopy (BES) diagnostic system to capture localized density fluctuations of long wavelength turbulent modes in the edge region at a 1 MHz sampling rate. BES data from 330 discharges in either L-mode, H-mode, quiescent H (QH)-mode, or WP QH-mode were collected from the DIII-D tokamak and curated to develop a high-quality database to train a deep-learning classification model for real-time confinement detection. We utilize the 6×8 spatial configuration with a time window of 1024 µ s and recast the input to obtain spectral-like features via fast Fourier transform preprocessing. We employ a shallow 3D convolutional neural network for the multivariate time-series classification task and utilize a softmax in the final dense layer to retrieve a probability distribution over the different confinement regimes. Our model classifies the global confinement state on 44 unseen test discharges with an average F 1 score of 0.94, using only ∼1 ms snippets of BES data at a time. This activity demonstrates the feasibility for real-time data analysis of fluctuation diagnostics in future devices such as ITER, where the need for reliable and advanced plasma control is urgent.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2026

NuGraph2 with explainability: post-hoc explanations for geometric neural network predictions

With the growing popularity of artificial intelligence (AI) used for scientific applications, the ability of attribute a result to a reasoning process from the network is in high demand for robust scientific generalizations to hold. In this work we aim to motivate the need for and demonstrate the use of post-hoc explainability methods when applied to AI methods used in scientific applications. To this end, we introduce explainability add-ons to the existing graph neural network (GNN) for neutrino tagging, NuGraph2. The explanations take the form of a suite of techniques examining the output of the network (node classifications) and the edge connections between them, and probing of the latent space using novel general-purpose tools applied to this network. We show how none of these methods are singularly sufficient to show network ‘understanding’, but together can give insights into the processes used in classification. While these methods are tested on the NuGraph2 application, they can be applied to a broad range of networks, not limited to GNNs. The code for this work is publicly available on GitHub at https://github.com/voetberg/XNuGraph.

Voetberg, Margaret [Fermilab] (ORCID:0009000527154↗
DOE OSTI2025

SetBERT: the deep learning platform for contextualized embeddings and explainable predictions from high-throughput sequencing

MOTIVATION: High-throughput sequencing (HTS) is a modern sequencing technology used to profile microbiomes by sequencing thousands of short genomic fragments from the microorganisms within a given sample. This technology presents a unique opportunity for artificial intelligence to comprehend the underlying functional relationships of microbial communities. However, due to the unstructured nature of HTS data, nearly all computational models are limited to processing DNA sequences individually. This limitation causes them to miss out on key interactions between microorganisms, significantly hindering our understanding of how these interactions influence the microbial communities as a whole. Furthermore, most computational methods rely on post-processing of samples which could inadvertently introduce unintentional protocol-specific bias. RESULTS: Addressing these concerns, we present SetBERT, a robust pre-training methodology for creating generalized deep learning models for processing HTS data to produce contextualized embeddings and be fine-tuned for downstream tasks with explainable predictions. By leveraging sequence interactions, we show that SetBERT significantly outperforms other models in taxonomic classification with genus-level classification accuracy of 95%. Furthermore, we demonstrate that SetBERT is able to accurately explain its predictions autonomously by confirming the biological-relevance of taxa identified by the model. AVAILABILITY AND IMPLEMENTATION: All source code is available at https://github.com/DLii-Research/setbert. SetBERT may be used through the q2-deepdna QIIME 2 plugin whose source code is available at https://github.com/DLii-Research/q2-deepdna.

Ludwig, David W↗
DOE OSTI2024

Comparative Performance of Three Eye-Tracking Devices in Detection of Mild Traumatic Brain Injury in Acute Versus Chronic Subject Populations

ABSTRACT Introduction Presently, traumatic brain injury (TBI) triage in field settings relies on symptom-based screening tools such as the updated Military Acute Concussion Evaluation. Objective eye-tracking may provide an alternative means of neurotrauma screening due to sensitivity to neurotrauma brain-health changes. Previously, the US Army Medical Research and Development Command Non-Invasive NeuroAssessment Devices (NINAD) Integrated Product Team identified 3 commercially available eye-tracking devices (SyncThink EYE-SYNC, Oculogica EyeBOX, NeuroKinetics IPAS) as meeting criteria toward being operationally effective in the detection of TBI in service members. We compared these devices to assess their relative performance in the classification of mild traumatic brain injury (mTBI) subjects versus normal healthy controls. Materials and Methods Participants 18 to 45 years of age were assigned to Acute mTBI, Chronic mTBI, or Control group per study criteria. Each completed a TBI assessment protocol with all 3 devices counterbalanced across participants. Acute mTBI participants were tested within 72 hours following injury whereas time since last injury for the Chronic mTBI group ranged from months to years. Discriminant analysis was undertaken to determine device classification performance in separating TBI subjects from controls. Area Under the Curves (AUCs) were calculated and used to compare the accuracy of device performance. Device-related factors including data quality, the need to repeat tests, and technical issues experienced were aggregated for reporting. Results A total of 63 participants were recruited as Acute mTBI subjects, 34 as Chronic mTBI subjects, and 119 participants without history of TBI as controls. To maximize outcomes, poorer quality data were excluded from analysis using specific criteria where possible. Final analysis utilized 49 (43 male/6 female, mean [x̅] age = 24.3 years, SD [s] = 5.1) Acute mTBI subjects, and 34 (33 male/1 female, x̅ age = 38.8 years, s = 3.9) Chronic mTBI subjects were age- and gender-matched as closely as possible with Control subjects. AUCs obtained with 80% of total dataset ranged from 0.690 to 0.950 for the Acute Group and from 0.753 to 0.811 for the Chronic mTBI group. Validation with the remaining 20% of dataset produced AUCs ranging from 0.600 to 0.750 for Acute mTBI group and 0.490 to 0.571 for the Chronic mTBI group. Conclusions Potential eye-tracking detection of mTBI, per training model outcomes, ranged from acceptable to excellent for the Acute mTBI group; however, it was less consistent for the Chronic mTBI group. The self-imposed targeted performance (AUC of 0.850) appears achievable, but further device improvements and research are necessary. Discriminant analysis models differed for the Acute versus Chronic mTBI groups, suggesting performance differences in eye-tracking. Although eye-tracking demonstrated sensitivity in the Chronic group, a more rigorous and/or longitudinal study design is required to evaluate this observation. mTBI injuries were not controlled for this study, potentially reducing eye-tracking assessment sensitivity. Overall, these findings indicate that while eye-tracking remains a viable means of mTBI screening, device-specific variability in data quality, length of testing, and ease of use must be addressed to achieve NINAD objectives and DoD implementation.

General & Internal Medicine↗
DOE OSTI2024

AstroCLIP: a cross-modal foundation model for galaxies

ABSTRACT We present AstroCLIP, a single, versatile model that can embed both galaxy images and spectra into a shared, physically meaningful latent space. These embeddings can then be used – without any model fine-tuning – for a variety of downstream tasks including (1) accurate in-modality and cross-modality semantic similarity search, (2) photometric redshift estimation, (3) galaxy property estimation from both images and spectra, and (4) morphology classification. Our approach to implementing AstroCLIP consists of two parts. First, we embed galaxy images and spectra separately by pre-training separate transformer-based image and spectrum encoders in self-supervised settings. We then align the encoders using a contrastive loss. We apply our method to spectra from the Dark Energy Spectroscopic Instrument and images from its corresponding Legacy Imaging Survey. Overall, we find remarkable performance on all downstream tasks, even relative to supervised baselines. For example, for a task like photometric redshift prediction, we find similar performance to a specifically trained ResNet18, and for additional tasks like physical property estimation (stellar mass, age, metallicity, and specific-star-formation rate), we beat this supervised baseline by 19 per cent in terms of R2. We also compare our results with a state-of-the-art self-supervised single-modal model for galaxy images, and find that our approach outperforms this benchmark by roughly a factor of two on photometric redshift estimation and physical property prediction in terms of R2, while remaining roughly in-line in terms of morphology classification. Ultimately, our approach represents the first cross-modal self-supervised model for galaxies, and the first self-supervised transformer-based architectures for galaxy images and spectra.

Parker, Liam (ORCID:0009000749521674)↗
DOE OSTI2024

Evaluating cosmological biases using photometric redshifts for Type Ia Supernova cosmology with the Dark Energy Survey Supernova Program

Cosmological analyses with Type Ia Supernovae (SNe Ia) have traditionally been reliant on spectroscopy for both classifying the type of supernova and obtaining reliable redshifts to measure the distance–redshift relation. While obtaining a host-galaxy spectroscopic redshift for most SNe is feasible for small-area transient surveys, it will be too resource intensive for upcoming large-area surveys such as the Vera Rubin Observatory Legacy Survey of Space and Time, which will observe on the order of millions of SNe. Here, we use data from the Dark Energy Survey (DES) to address this problem with photometric redshifts (photo-z) inferred directly from the SN light curve in combination with Gaussian and full p(z) priors from host-galaxy photo-z estimates. Using the DES 5-yr photometrically classified SN sample, we consider several photo-z algorithms as host-galaxy photo-z priors, including the Self-Organizing Map redshifts (SOMPZ), Bayesian Photometric Redshifts (BPZ), and Directional-Neighbourhood Fitting (DNF) redshift estimates employed in the DES 3 × 2 point analyses. With detailed catalogue-level simulations of the DES 5-yr sample, we find that the simulated w can be recovered within ±0.02 when using SN+SOMPZ or DNF prior photo-z, smaller than the average statistical uncertainty for these samples of 0.03. With data, we obtain biases in w consistent with simulations within ~1σ for three of the five photo-z variants. We further evaluate how photo-z systematics interplay with photometric classification and find classification introduces a subdominant systematic component. This work lays the foundation for next-generation fully photometric SNe Ia cosmological analyses.

(cosmology:) dark energy↗
DOE OSTI2024

Preserving isohydricity: vertical environmental variability explains Amazon forest water-use strategies

Abstract Increases in hydrological extremes, including drought, are expected for Amazon forests. A fundamental challenge for predicting forest responses lies in identifying ecological strategies which underlie such responses. Characterization of species-specific hydraulic strategies for regulating water-use, thought to be arrayed along an ‘isohydric–anisohydric’ spectrum, is a widely used approach. However, recent studies have questioned the usefulness of this classification scheme, because its metrics are strongly influenced by environments, and hence can lead to divergent classifications even within the same species. Here, we propose an alternative approach positing that individual hydraulic regulation strategies emerge from the interaction of environments with traits. Specifically, we hypothesize that the vertical forest profile represents a key gradient in drought-related environments (atmospheric vapor pressure deficit, soil water availability) that drives divergent tree water-use strategies for coordinated regulation of stomatal conductance (gs) and leaf water potentials (ΨL) with tree rooting depth, a proxy for water availability. Testing this hypothesis in a seasonal eastern Amazon forest in Brazil, we found that hydraulic strategies indeed depend on height-associated environments. Upper canopy trees, experiencing high vapor pressure deficit (VPD), but stable soil water access through deep rooting, exhibited isohydric strategies, defined by little seasonal change in the diurnal pattern of gs and steady seasonal minimum ΨL. In contrast, understory trees, exposed to less variable VPD but highly variable soil water availability, exhibited anisohydric strategies, with fluctuations in diurnal gs that increased in the dry season along with increasing variation in ΨL. Our finding that canopy height structures the coordination between drought-related environmental stressors and hydraulic traits provides a basis for preserving the applicability of the isohydric-to-anisohydric spectrum, which we show here may consistently emerge from environmental context. Our work highlights the importance of understanding how environmental heterogeneity structures forest responses to climate change, providing a mechanistic basis for improving models of tropical ecosystems.

Forestry↗
DOE OSTI2026

Soft symmetries of topological orders

(2+1)D topological orders possess emergent symmetries given by a group $\text{Aut}(\mathcal{C})$, which consists of the braided tensor autoequivalences of the modular tensor category $\mathcal{C}$ that describes the anyons. In this paper we discuss cases where $\text{Aut}(\mathcal{C})$ has elements that neither permute anyons nor are associated to any symmetry fractionalization but are still non-trivial, which we refer to as soft symmetries. We point out that one can construct topological defects corresponding to such exotic symmetry actions by decorating with a certain class of gauged SPT states that cannot be distinguished by their torus partition function. This gives a physical interpretation to work by Davydov on soft braided tensor autoequivalences. This has a number of important implications for the classification of gapped boundaries, non-invertible spontaneous symmetry breaking, and the general classification of symmetry-enriched topological phases of matter. Here, we also demonstrate analogous phenomena in higher dimensions, such as (3+1)D gauge theory with gauge group given by the quaternion group $Q_8$.

Gauge theory techniques↗
DOE OSTI2025

Energy Shifts and Broadening of Excitonic Resonances in Electrostatically Doped Semiconductors

Tuning the density of resident electrons or holes in semiconductors through electrostatic doping provides crucial insight into the composition of the many excitonic complexes that are routinely observed as absorption or photoluminescence resonances in optical studies. Moreover, we can change the way these resonances shift and broaden in energy by controlling the quantum numbers (e.g., spin and valley) of the resident carriers with applied magnetic fields and doping levels, as well as by selecting the quantum numbers of the photoexcited or recombining electron-hole (𝑒−ℎ) pair through optical polarization. Here, we discuss the roles of distinguishability and optimality of excitonic complexes, showing them to be key ingredients that determine the energy shifts and broadening of optical resonances in charge-tunable semiconductors. A distinguishable 𝑒−ℎ pair means that the electron and hole undergoing photoexcitation or recombination have quantum numbers that are not shared by any of the resident carriers. An optimal excitonic complex refers to a complex whose particles come with all available quantum numbers of the resident carriers. Based on the carrier density, magnetic field, and light polarization, all optical resonances may be classified as either distinct or indistinct depending on the distinguishability of the 𝑒−ℎ pair, and the underlying excitonic complex can be classified as either optimal or suboptimal. The universality of these classifications, inherited from the fundamental Pauli exclusion principle, allows us to understand how optical resonances shift in energy and whether they should broaden as doping is increased. This understanding is supported by conclusive evidence that the broadening and decay of optical resonances cannot be simply attributed to enhanced screening when resident carriers are added to a semiconductor. Finally, applying the classification scheme in either monolayer or moiré heterobilayer semiconductor systems, we relate the energy shift and amplitude of the neutral-exciton resonance to the compressibility of the resident carrier gas.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2024

Differentiable vertex fitting for jet flavor tagging

This work explores the use of differentiable programming to integrate domain knowledge, in the form of domain specific software, into neural networks to develop scientific machine learning systems. We propose a differentiable vertex fitting algorithm that estimates the crossing point of multiple curves. In the high energy physics setting, these curves are defined by particle equations of motion and the crossing point represents the origin of particle production. This differentiable vertex fitting algorithm can be seamlessly integrated into neural networks, and we show its utility and efficacy in the high energy physics application of the classification of jets, i.e., collimated streams of particles in particle detectors whose originating parent particle we aim to classify. We demonstrate how differentiable vertex fitting can be integrated into larger transformer-based models for jet flavor tagging and show improvements in heavy flavor jet classification when compared to baseline models. Published by the American Physical Society 2024

Smith, Rachel E. C. (ORCID:0000000335851262)↗
DOE OSTI2024

SIGHT: Stacked Integration of Geospatial Hierarchical Typologies for Inferring Building Characteristics

Building characteristics are often absent in building stock datasets, particularly in regions most vulnerable to climate change and requiring effective disaster management strategies. Traditional machine learning approaches, while widely used to predict building attributes, typically neglect the spatial context of the data, leading to less accurate and reliable outcomes. To address these challenges, this paper introduces a novel algorithm, the Stacked Integration of Geospatial Hierarchical Typologies. This algorithm adapts a meta-learning framework to incorporate geospatial context into the predictive modeling process. We demonstrate the utility of the algorithm through two primary use cases: building use type classification and building height prediction. The algorithm consistently achieved or exceeded a 0.94 macro average F1 score across five geographically distinct countries for building use type classification. For building height prediction, it accurately predicted heights with a root mean square error of 3.01 in a comprehensive study using roughly 3.6 million buildings in Japan. These results underscore the benefits of integrating spatial hierarchies into machine learning models, enhancing both predictive accuracy and reliability in geospatial modeling. This work introduces a new algorithm to address the pervasive data sparsity issue in existing building stock datasets.

Adams, Daniel [ORNL] (ORCID:0000000196950577)↗
DOE OSTI2025

Free-Space Optical Channel Turbulence Prediction: A Machine Learning Approach

Channel turbulence is a formidable obstacle for freespace optical (FSO) communication. Anticipation of turbulence levels is highly important for mitigating disruptions but has not been demonstrated without dedicated, auxiliary hardware. We show that machine learning (ML) can be applied to raw FSO data streams to rapidly predict channel turbulence levels with no additional sensing hardware. FSO was conducted through a controlled channel in the lab under six distinct turbulence levels, and the efficacy of using ML to classify turbulence levels was examined. ML-based turbulence level classification was found to be > 98% accurate with multiple ML training parameters. Classification effectiveness was found to depend on the timescale of changes between turbulence levels but converges when turbulence stabilizes over about a one minute timescale.

Zobaer Islam, Md [School of Electrical and Compute↗
DOE OSTI2024

Transformer Masked Autoencoders for RF Device Fingerprinting

Machine learning methods for RF device fingerprinting typically rely on CNN-based models. Transformer-based models have outperformed CNNs for modulation classification tasks, but there are few implementations for device fingerprinting. We train a transformer for device fingerprinting with the largest device count to date and explore several variations of the architecture. Additionally, we demonstrate that pre-training an RF transformer as a Masked Autoencoder improves classification accuracy, as has been observed for CNN fingerprinting models and vision transformers.

artificial intelligence↗
DOE OSTI2026

Feedback and Oscillations: Constructing Feedback Systems for Root Cause Analysis of Oscillations in Power Grids

Dynamic phenomena linked to inverter-based resources (IBRs) have gained global attention. Several IBR-induced dynamics have caused bulk power system-connected wind or solar power plants to trip, and some have even led to widespread outages. In addition, many oscillations have been observed involving IBR power plants. In 2023, the IEEE Power & Energy Society (PES) IBR Subsynchronous Oscillations (SSO) task force published a journal article, “Real-World Subsynchronous Oscillation Events in Power Grids With High Penetrations of Inverter-Based Resources,” in which 19 IBR oscillation events were examined for their causation. Earlier in 2020, another PES task force article, “Definition and Classification of Power System Stability-Revisited & Extended,” authored by prominent academics, introduced converter-driven stability as a new category of stability. The international power grid industry community also took action by publishing the CIGRE Green Book, Power System Dynamic Modelling and Analysis in Evolving Networks (led by Babak Badrzadeh and Zia Emin) in 2024. In August 2024, the Energy Systems Integration Group (ESIG) released a practical guide led by Nick Miller, “Diagnosis and Mitigation of Observed Oscillations in IBR-Dominant Power System: A Practical Guide.” The goal of the guide is to assist practicing engineers in making initial judgments and conducting detailed analyses about oscillations. Finally, when addressing the classification of stability and oscillations, the guide emphasizes a causality-based taxonomy for grouping, such as voltage control-induced oscillations, synchronization-induced oscillations, and frequency or active power control-induced oscillations.

Fan, Lingling [Univ. of South Florida, Tampa, FL (↗
DOE OSTI2024

IRIS-GNN: Leveraging Graph Neural Networks for Scheduling on Truly Heterogeneous Runtime Systems

The diversity of accelerators in computer systems poses significant challenges for software developers, such as managing vendor-specific compiler toolchains, code fragmentation requiring different kernel implementations, and performance portability issues. To address these, the Intelligent Runtime System (IRIS) was developed. IRIS works across various systems, from smartphones to supercomputers, enabling automatic performance scaling based on available accelerators. It introduces abstract tasks for seamless execution transitions between accelerators while ensuring memory consistency and task dependencies. Although IRIS simplifies system details, optimal dynamic scheduling still requires user input to understand workload structures. To address this, we introduce a new scheduling policy for IRIS, termed IRIS-GNN, which is the first IRIS hybrid policy that operates in conjunction with the dynamic policies. This policy employs a Graph-Neural Network (GNN) to conduct Graph Classification of any task graphs submitted to IRIS. This GNN analyzes the structure and attributes of the task graph, categorizing it as either locality, concurrency, or mixed. This classification subsequently guides the selection of the dynamic policy used by IRIS. We provide a comparison of the performance of IRIS-GNN against the complete spectrum of IRIS’s dynamic policies, assess the overhead introduced by the GNN within this scheduling framework, and ultimately explore its practical application in real-world scenarios.

Johnston, Beau↗