Search NASA⌕ Search

SEARCH · Search NASA

Results for “pruning”

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 19 records
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

“Understanding Robustness Lottery”: A Geometric Visual Comparative Analysis of Neural Network Pruning Approaches

Deep learning approaches have provided state-of-the-art performance in many applications by relying on large and overparameterized neural networks. However, such networks are very brittle and are difficult to deploy on resource-limited platforms. Model pruning, i.e., reducing the size of the network, is a widely adopted strategy that can lead to a more robust and compact model. Many heuristics exist for model pruning, but our understanding of the pruning process remains limited due to the black-box nature of a neural network model. Empirical studies show that some heuristics improve performance whereas others can make models more brittle. Here, this work aims to shed light on how different pruning methods alter the network’s internal feature representation and the corresponding impact on model performance. To facilitate a comprehensive comparison and characterization of the high-dimensional model feature space, we introduce a visual geometric analysis of feature representations. We evaluated a set of critical geometric concepts decomposed from the commonly adopted classification loss and used them to design a visualization system to compare and highlight the impact of pruning on model performance and feature representation. The proposed tool provides an environment for an in-depth comparison of pruning methods and a comprehensive understanding of how the model responds to common data corruption. By leveraging the proposed visualization, machine learning researchers can reveal the similarities between pruning methods and redundancy in robustness evaluation benchmarks, obtain geometric insights about the differences between pruned models that achieve superior robustness performance, and identify samples that are robust or fragile to model pruning and common data corruption.

Li, Zhimin [Univ. of Utah, Salt Lake City, UT (Uni↗
DOE OSTI2025

network-pruner (Neural network pruning analysis) [SWR-25-113]

This repository implements an iterative magnitude pruning algorithm for pruning neural networks in PyTorch. The pruning method involves gradually removing less significant weights from the model to achieve a specified sparsity, followed by fine-tuning the pruned model to recover performance.

Griffin, Kevin [National Renewable Energy Laborato↗
DOE OSTI2025

SymbolNet: neural symbolic regression with adaptive dynamic pruning for compression

Abstract Compact symbolic expressions have been shown to be more efficient than neural network (NN) models in terms of resource consumption and inference speed when implemented on custom hardware such as field-programmable gate arrays (FPGAs), while maintaining comparable accuracy (Tsoi et al 2024 EPJ Web Conf. 295 09036). These capabilities are highly valuable in environments with stringent computational resource constraints, such as high-energy physics experiments at the CERN Large Hadron Collider. However, finding compact expressions for high-dimensional datasets remains challenging due to the inherent limitations of genetic programming (GP), the search algorithm of most symbolic regression (SR) methods. Contrary to GP, the NN approach to SR offers scalability to high-dimensional inputs and leverages gradient methods for faster equation searching. Common ways of constraining expression complexity often involve multistage pruning with fine-tuning, which can result in significant performance loss. In this work, we propose S y m b o l N e t , a NN approach to SR specifically designed as a model compression technique, aimed at enabling low-latency inference for high-dimensional inputs on custom hardware such as FPGAs. This framework allows dynamic pruning of model weights, input features, and mathematical operators in a single training process, where both training loss and expression complexity are optimized simultaneously. We introduce a sparsity regularization term for each pruning type, which can adaptively adjust its strength, leading to convergence at a target sparsity ratio. Unlike most existing SR methods that struggle with datasets containing more than O ( 10 ) inputs, we demonstrate the effectiveness of our model on the LHC jet tagging task (16 inputs), MNIST (784 inputs), and SVHN (3072 inputs).

Tsoi, Ho Fung (ORCID:0000000225502184)↗
DOE OSTI2024

Fast Iterative Multi-site Hosting Capacity Analysis for Distribution Systems With Search Space Pruning

Interconnection studies for distributed energy resources (DERs) is a time-intensive process, primarily due to the necessity of solving large number of power flow scenarios. Hosting capacity analysis (HCA) is a time-consuming aspect of interconnection studies that is divided into single-site HCA (SHCA) and multi-site HCA (MHCA). From a computational and understandable standpoint, the industry seeks iteration-based solutions for SHCA, although it doesn't maximize the total DER hosting capacity (DERHC) of the grid, as MHCA does. While non-iterative solutions are available for MHCA, they involve a trade-off between the modeling accuracy of the distribution system, solution quality, and ease of understanding. In this work, we present a fast iterative solution for MHCA, reducing computational complexity by eliminating the need to solve power flows for a large amount of search space, thus making iterative solutions feasible. This iterative approach guarantees both a global optimal solution with sufficient time and a fast, close-to-optimal solution through efficient search space pruning. It also easily integrates with existing utility HCA tools. The results are demonstrated on select locations in the IEEE-123 bus system for community-scale interconnection studies. We highlight the benefits of skipping the need to solve millions of power flows, all while maximizing the grid's total DERHC.

Guddanti, Kishan Prudhvi↗
DOE OSTI2026

Locally purified maximally mixed states at scale: Entanglement pruning and symmetries

Locally Purified Density Operators (LPDOs) are state-of-the-art tensor network ansatze candidates that efficiently represent mixed quantum states at scale. However, given their non-uniqueness, their representational complexity is generally sub-optimal in practical computations. Here, in this work we perform a comprehensive numerical and analytical analysis and resolve this issue in the experimentally relevant limit where noise depolarizes the density operator into a maximally mixed state. To resolve the sub-optimality issue, we analyze two numerical tools, one analytic method, and detail the relations between them. The numerical tools used are fidelity-preserving truncations and isometric gauge transformations leveraging Riemannian optimizations over entropic objective functions. In addition, by invoking the injectivity and symmetry constraints of the maximally mixed LPDO, we also present analytical closed-form expressions for the disentangler and discuss their relation to numerical optimizers. Further, away from the maximally mixed state, our simulations highlight how the truncation threshold smoothly interpolate, as a function of depolarization, between established matrix product results and our new results. Our work shows how, by minimizing the resources required to represent key states of practical interest in experiment, the efficiency of tensor network algorithms can be substantially increased. This paves the path for uncovering tensor network’s fundamental scalability limits and latent potential in representing the wide locus of mixed quantum states that are accessible on near-term quantum devices.

Gangapuram, Amit Jamadagni [Oak Ridge National Lab↗
DOE OSTI2026

Mitigating Algorithmic Bias in Cancer Site Classification Models

Purpose Integrating artificial intelligence in cancer diagnostics has improved tumor classification beyond rule-based systems. Despite these advancements, these models may still encode demographic biases. We conducted a large-scale, applied bias-probing study of a deep learning–based cancer site classifier to quantify race information encoded in document embeddings. We then evaluated how performance changes when race-correlated embedding dimensions are removed in a post-training sensitivity analysis. Methods The cancer site classifier was trained using 3.5 million electronic cancer pathology reports from six of the National Cancer Institute's SEER registries. We trained a hierarchical self-attention network to generate 400-dimensional document embeddings. These embeddings were used to train two downstream, gradient-boosted decision tree classifiers: one to classify the cancer sites and another to predict racial categories. We identified overlapping features by intersecting the top 50 feature-importance rankings from the site and race models and computed their cumulative feature importance in each model. As a post hoc sensitivity analysis, we progressively pruned these overlapping dimensions, retrained the site model, and compared overall macro-F1 and accuracy, race-stratified macro-F1, and group fairness metrics on the basis of demographic parity and equalized odds before and after pruning. Results The analysis revealed minimal feature overlap between the cancer site and race prediction models, and the cumulative importance scores indicated a negligible influence of racial information on clinical predictions. Post-training pruning of overlapping features did not compromise the models' diagnostic accuracy, with a 0.07% loss in accuracy. Conclusion Our findings demonstrate that HiSAN-generated embeddings from SEER data can be used effectively in cancer site classification without significant demographic bias influencing the outcomes. Post-training pruning therefore functions as a practical audit and sensitivity check.

Shivanna, Abhishek [ORNL] (ORCID:0009000665228593)↗
DOE OSTI2026

Unsupervised atomic data mining via multi-kernel graph autoencoders for machine learning force fields

Constructing a chemically diverse dataset while avoiding sampling bias is critical to training efficient and generalizable force fields. However, in computational chemistry and materials science, many common dataset generation techniques are prone to oversampling regions of the potential energy surface. Furthermore, these regions can be difficult to identify and isolate from each other or may not align well with human intuition, making it challenging to systematically remove bias in the dataset. While traditional clustering and pruning (down-sampling) approaches can be useful for this, they can often lead to information loss or a failure to properly identify distinct regions of the potential energy surface due to difficulties associated with the high dimensionality of atomic descriptors. In this work, we introduce the Multi-kernel Edge Attention-based Graph Autoencoder (MEAGraph) model, an unsupervised approach for analyzing atomic datasets. MEAGraph combines multiple linear kernel transformations with attention-based message passing to capture geometric sensitivity and enable effective dataset pruning without relying on labels or extensive training. Demonstrated applications on niobium, tantalum, and iron datasets show that MEAGraph efficiently groups similar atomic environments, allowing for the use of basic pruning techniques for removing sampling bias. This approach provides an effective method for representation learning and clustering that can be used for data analysis, outlier detection, and dataset optimization.

Materials science↗
DOE OSTI2025

Uncontrolled Learning: Codesign of Neuromorphic Hardware Topology for Neuromorphic Algorithms

Neuromorphic computing has the potential to revolutionize future technologies and our understanding of intelligence, yet it remains challenging to realize in practice. The learning-from-mistakes algorithm, inspired by the brain's simple learning rules of inhibition and pruning, is one of the few brain-like training methods. This algorithm is implemented in neuromorphic memristive hardware through a codesign process that evaluates essential hardware trade-offs. While the algorithm effectively trains small networks as binary classifiers and perceptrons, performance declines significantly with increasing network size unless the hardware is tailored to the algorithm. This work investigates the trade-offs between depth, controllability, and capacity—the number of learnable patterns—in neuromorphic hardware. This highlights the importance of topology and governing equations, providing theoretical tools to evaluate a device's computational capacity based on its measurements and circuit structure. The findings show that breaking neural network symmetry enhances both controllability and capacity. Additionally, by pruning the circuit, neuromorphic algorithms in all-memristive circuits can utilize stochastic resources to create local contrasts in network weights. Through combined experimental and simulation efforts, the parameters are identified that enable networks to exhibit emergent intelligence from simple rules, advancing the potential of neuromorphic computing.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2025

DESIVAST: Catalogs of Low-redshift Voids Using Data from the DESI Data Release 1 Bright Galaxy Survey

We present three separate void catalogs created using a volume-limited sample of the DESI Data Release 1 Bright Galaxy Survey. We use the algorithms VoidFinder and V 2 to construct void catalogs out to a redshift of z = 0.24. Excluding voids affected by the boundaries of the survey, we obtain 1489 voids with VoidFinder, 389 with V 2 using REVOLVER pruning, and 297 with V 2 using VIDE pruning. Comparing our catalogs with overlapping Sloan Digital Sky Survey void catalogs, we find generally consistent void properties but significant differences in the void volume overlap, which we attribute to differences in the galaxy selection and survey masks. These catalogs are suitable for studying the variation in galaxy properties with cosmic environment and for cosmological studies.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2025

Neural architecture codesign for fast physics applications

We develop a pipeline to streamline neural architecture codesign for physics applications to reduce the need for ML expertise when designing models for novel tasks. Our method employs neural architecture search and network compression in a two-stage approach to discover hardware efficient models. This approach consists of a global search stage that explores a wide range of architectures while considering hardware constraints, followed by a local search stage that fine-tunes and compresses the most promising candidates. We exceed performance on various tasks and show further speedup through model compression techniques such as quantization-aware-training and neural network pruning. We synthesize the optimal models to high level synthesis code for FPGA deployment with the hls4ml library. Additionally, our hierarchical search space provides greater flexibility in optimization, which can easily extend to other tasks and domains. We demonstrate this with two case studies: Bragg peak finding in materials science and jet classification in high energy physics, achieving models with improved accuracy, smaller latencies, or reduced resource utilization relative to the baseline models.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2026

Flexible AI Models for Grid Resilience

The rapid growth in size and complexity of artificial intelligence (AI) and machine learning (ML) models has led to increased energy demands, posing a threat to the reliability of the existing power grid. This project addresses the challenge of highly intermittent and energy-intensive inference workloads by (1) developing fidelity-adaptive neural networks capable of dynamic response to grid conditions and (2) integrating these networks with power flow simulations to assess their impact on power grid reliability. We will explore both top-down and bottom-up approaches to create hierarchies of submodels that provide a controlled trade-off between power draw and prediction accuracy. The top-down method utilizes NN pruning to reduce a flagship model into progressively smaller, energy-efficient variants. The bottom-up approach employs geometrically principled weight setting strategies to construct depth-efficient models from the ground up. A real-time hardware-in-the-loop (HIL) platform will be developed to simulate a scaled AC power grid, integrating live AI workload power draw and enabling dynamic model switching in response to grid feedback. This work will provide a novel framework for evaluating the impact of flexible AI/ML workloads on grid performance and establish new methodologies for energy-aware computing in data centers. The outcomes will demonstrate that adaptive AI/ML can play a critical role in improving grid stability while advancing NREL's leadership in energy-efficient computing research.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2024

Improved Statistics for F-theory Standard Models

Much of the analysis of F-theory-based Standard Models boils down to computing cohomologies of line bundles on matter curves. By varying parameters one can degenerate such matter curves to singular ones, typically with many nodes, where the computation is combinatorial and straightforward. The question remains to relate the (a priori possibly smaller) value on the original curve to the singular one. In this work, we introduce some elementary techniques (pruning trees and removing interior edges) for simplifying the resulting nodal curves to a small collection of terminal ones that can be handled directly. When applied to the QSMs, these techniques yield optimal results in the sense that obtaining more precise answers would require currently unavailable information about the QSM geometries. This provides us with an opportunity to enhance the statistical bounds established in earlier research regarding the absence of vector-like exotics on the quark-doublet curve.

Bies, Martin↗
DOE OSTI2026

Learning interpretable surface elasticity properties from bulk properties via neural network equation learners

Surface elasticity is central to understanding the mechanics and stability of surfaces and interfaces. It is characterized by quantities such as surface tension, residual surface stress, and surface stiffness. However their analytical expressions are typically difficult to derive from atomistic data, and depend strongly on modeling choices. This work presents a neural network-based equation learner which combines customized activation functions and connection-based pruning to discover parsimonious, closed-form equations for surface elasticity from atomistic simulations. Applying the method to seven face-centered cubic (FCC) metals, our equation learner uncovers interpretable equations that describe both low-Miller index and high-Miller index surface properties, capturing long-tail property distributions accurately. The discovered expressions are decoupled into two components: a universal, geometry-driven orientation function, and material-specific baseline coefficients. We find that lower-order properties such as surface tension are fundamentally geometry dependent, while higher-order properties such as surface stress and elasticity show more complex geometry and material dependence. We also relate material dependent coefficients to bulk properties, forming a clear map from bulk material properties to surface elasticity. Overall, this approach demonstrates that interpretable neurosymbolic machine learning can bridge the gap between atomistic simulations and physical laws, enabling the discovery of generalizable structure–property relationships for materials science phenomena such as surface elasticity.

Equation learning↗
DOE OSTI2025

Physics-Driven Construction of Compact Primitive Gaussian Density Fitting Basis Sets

We present a model-assisted density fitting (MADF) basis set generator, an algorithm for generating primitive atomic Gaussian density fitting (DF) basis sets (DFBSs) from a contracted Gaussian orbital basis set (OBS). The MADF algorithm produces DFBSs suitable for accurate robust DF approximation of 2-particle interactions in mean-field and correlated electronic structures. The algorithm is designed to (a) saturate the OBS product space by a large regularized set of primitive solid-harmonic Gaussian shells with nonuniform distribution of exponents, followed by (b) pruning of the shells according to their contributions to the 2- body energy of a correlated atomic ensemble. Building the DFBS generator model almost exclusively on mathematical and physical principles allows one to limit the number of parameters that control the density fitting error to three, with a single set of parameters sufficient for computations with all basis cardinal numbers, with and without correlation of core electrons, with and without scalar and spin-dependent relativistic effects, spanning almost all of the Periodic Table. Performance assessment included basis sets up to quadruple-ζ quality from several major basis set families, using molecules composed of main-group, d-block, and f-block elements. The resulting DF errors in Hartree−Fock and second-order MP2 energies (with relativistic all-electron treatments, when appropriate) were on the order of 20 and 10 μE h per electron, respectively.

Approximation↗
DOE OSTI2025

Global River Topology (GRIT): A Bifurcating River Hydrography

Existing global river networks underpin a wide range of hydrological applications but do not represent channels with divergent river flows (bifurcations, multi‐threaded channels, canals), as these features defy the convergent flow assumption that elevation‐derived networks (e.g., HydroSHEDS, MERIT Hydro) are based on. Yet, bifurcations are important features of the global river drainage system, especially on large floodplains and river deltas, and are also often found in densely populated regions. Here we developed the first raster and vector‐based Global RIver Topology that not only represents the tributaries of the global drainage network but also the distributaries, including multi‐threaded rivers, canals and deltas. We achieve this by merging a 30 m Landsat‐based river mask with elevation‐generated streams to ensure a homogeneous drainage density outside of the river mask for rivers narrower than approximately 30 m. Crucially, we employ the new 30 m digital terrain model, FABDEM, based on TanDEM‐X, which shows greater accuracy over the traditionally used SRTM derivatives. After vectorization and pruning, directionality is assigned by a series of elevation, flow angle and continuity approaches. The new global network and its attributes are validated using gauging stations, comparison with existing networks, and randomized manual checks. The new network represents 19.6 million km of streams and rivers with drainage areas greater than 50 km 2 and includes 67,495 bifurcations. With the advent of hyper‐resolution modeling and artificial intelligence, GRIT is expected to greatly improve the accuracy of many river‐based applications such as flood forecasting, water availability and quality simulations, or riverine habitat mapping.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

Maximizing efficiency of dataset compression for machine learning potentials with information theory

Machine learning interatomic potentials (MLIPs) balance high accuracy and lower costs compared to density functional theory calculations, but their performance often depends on the size and diversity of training datasets. Large datasets improve model accuracy and generalization but are computationally expensive to produce and train on, while smaller datasets risk discarding rare but important atomic environments and compromising MLIP accuracy/reliability. Here, we develop an information-theoretical framework to quantify the efficiency of dataset compression methods and propose an algorithm that maximizes this efficiency. By framing atomistic dataset compression as an instance of the minimum set cover (MSC) problem over atom-centered environments, our method identifies the smallest subset of structures that contains as much information as possible from the original dataset while pruning redundant information. The approach is extensively demonstrated on the GAP-20 and TM23 datasets and validated on 64 varied datasets from the ColabFit repository. Across all cases, MSC consistently retains outliers, preserves dataset diversity, and reproduces the long-tail distributions of forces even at high compression rates, outperforming other subsampling methods. Furthermore, MLIPs trained on MSC-compressed datasets exhibit reduced error for out-of-distribution data even in low-data regimes. We explain these results using an outlier analysis and show that such quantitative conclusions could not be achieved with conventional dimensionality reduction methods. The algorithm is implemented in the open-source QUESTS package and can be used for several tasks in atomistic modeling, from data subsampling, outlier detection, and training improved MLIPs at a lower cost.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Spatial Correlations of the Poisson Model for Radiation Transport

Characterizing the relationship between bulk physical properties and mixing in randomly heterogeneous media is a central challenge across many areas of science and engineering. A benchmark model for such studies is the Poisson model, a random tessellation of space by a Poisson process of hyperplanes. In radiation transport studies, the lack of exact expressions for the Poisson model’s spatial multipoint functions has led to approximate methods being used, introducing unquantified sources of error. Here, we recently introduced an exact solution for the Poisson model’s multipoint functions and closely related conditional probability functions (CPFs), providing a new opportunity to understand and reduce these sources of error. In this paper, we enable a more rigorous investigation of radiation transport in stochastic media by applying the recently introduced exact solution for the Poisson model’s CPFs. This paper consists of three main contributions. First, we introduce a unified framework for CPFs of the Poisson model, encompassing the recently introduced exact CPFs as well as the previously introduced atomic mix, nearest-neighbor, and combination CPFs. This framework also includes existing pruning techniques for the approximate CPFs, such as angular exclusion, as well as a novel form of angular exclusion suitable for the exact CPFs. Second, we use the exact CPFs to characterize the spatial regions where each approximate three-point CPF is most accurate, thereby explaining the observed hierarchy of accuracy among the approximate models. Finally, we evaluate material transmittance, reflectance, and flux in a three-dimensional test problem using conditional point sampling, demonstrating the relationship between CPF accuracy and transport simulation accuracy.

Poisson model↗