Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Pre-training”

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

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

At least 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

A Representation Fusion Framework for Decoupling Diagnostic Information in Multimodal Learning

Modern medicine increasingly relies on multimodal data, ranging from clinical notes to imaging and genomics, to guide diagnosis and treatment. However, integrating these heterogeneous data sources in a principled and interpretable manner remains a major challenge. We present MODES (Multi-mOdal Disentangled Embedding Space), a representation fusion framework that explicitly separates shared and modality-specific factors of variation, offering a structured latent space for multimodal information that improves both prediction and interpretability. By leveraging pre-trained unimodal foundation models, MODES mitigates the dependency on extensive paired datasets, crucial in data-scarce clinical settings. We introduce a masking strategy that optimizes representation dimensionality by eliminating low-information dimensions, to achieve compact, information-rich representations. Our framework demonstrates superior performance in predicting diagnoses and phenotypes compared to unimodal and conventional fusion models. MODES also enables robust diagnostic inference in missing data scenarios, offering an opportunity toward interpretable and efficient multimodal diagnostics in personalized healthcare.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2026

Enhancing transfer learning in angle-resolved photoemission spectroscopy (ARPES) with spatially-aware representations via graph convolution

A recent application of machine learning has been to spatially-resolved angle-resolved photoemission spectroscopy (ARPES). Here we advance the state-of-the-art by applying representational learning to transform ARPES data into an embedding space of a pre-trained self-supervised learning model, thus enhancing the pipeline that improves the bandstructure classification and domain assignment/segmentation performance compared to a k-means clustering method. In the current iteration, the real-space information is entered into the domain assignment through the graph convolution method, which improves the transfer learning performance of the original self-supervised model. Lastly, an unsupervised automated tool is developed that incorporates these techniques to enable automatic domain assignment.

ARPES↗
DOE OSTI2025

Coefficient-to-Basis Network: a fine-tunable operator learning framework for inverse problems with adaptive discretizations and theoretical guarantees

We propose a Coefficient-to-Basis Network (C2BNet), a novel framework for solving inverse problems within the operator learning paradigm. C2BNet efficiently adapts to different discretizations through fine-tuning, using a pre-trained model to significantly reduce computational cost while maintaining high accuracy. Unlike traditional approaches that require retraining from scratch for new discretizations, our method enables seamless adaptation without sacrificing predictive performance. Furthermore, we establish theoretical approximation and generalization error bounds for C2BNet by exploiting low-dimensional structures in the underlying datasets. Our analysis demonstrates that C2BNet adapts to low-dimensional structures without relying on explicit encoding mechanisms, highlighting its robustness and efficiency. To validate our theoretical findings, we conducted extensive numerical experiments that showcase the superior performance of C2BNet on several inverse problems. The results confirm that C2BNet effectively balances computational efficiency and accuracy, making it a promising tool to solve inverse problems in scientific computing and engineering applications.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Make the Fastest Faster: Importance Mask Synthesis for Interactive Volume Visualization using Reconstruction Neural Networks

Visualizing a large-scale volumetric dataset with high resolution is challenging due to the substantial computational time and space complexity. Recent deep learning-based image inpainting methods significantly improve rendering latency by reconstructing a high-resolution image for visualization in constant time on GPU from a partially rendered image where only a portion of pixels go through the expensive rendering pipeline. However, existing solutions need to render every pixel of either a predefined regular sampling pattern or an irregular sample pattern predicted from a low-resolution image rendering. Both methods require a significant amount of expensive pixel-level rendering. In this work, we provide Importance Mask Learning (IML) and Synthesis (IMS) networks, which are the first attempts to directly synthesize important regions of the regular sampling pattern from the user’s view parameters, to further minimize the number of pixels to render by jointly considering the dataset, user behavior, and the downstream reconstruction neural network. Our solution is a unified framework to handle various types of inpainting methods through the proposed differentiable compaction/decompaction layers. Experiments show our method can further improve the overall rendering latency of state-of-the-art volume visualization methods using reconstruction neural network for free when rendering scientific volumetric datasets. Our method can also directly optimize the off-the-shelf pre-trained reconstruction neural networks without elongated retraining.

Large-scale data↗
DOE OSTI2025

RingX: Scalable Parallel Attention for Long-Context Learning on HPC

The attention mechanism has become foundational for remarkable AI breakthroughs since the introduction of the Transformer, driving the demand for increasingly longer context to power frontier models such as large-scale reasoning language models and high-resolution image/video generators. However, its quadratic computational and memory complexities present substantial challenges. Current state-of-the-art parallel attention methods, such as ring attention, are widely adopted for long-context training but utilize a point-to-point communication strategy that fails to fully exploit the capabilities of modern HPC network architectures. In this work, we propose ringX, a scalable family of parallel attention methods optimized explicitly for HPC systems. By enhancing workload partitioning, refining communication patterns, and improving load balancing, ringX achieves up to 3.4 × speedup compared to conventional ring attention on the Frontier supercomputer. Optimized for both bi-directional and causal attention mechanisms, ringX demonstrates its effectiveness through training benchmarks of a Vision Transformer (ViT) on a climate dataset and a Generative Pre-Trained Transformer (GPT) model, Llama3 8B. Our method attains an end-to-end training speedup of approximately 1.5 × in both scenarios. To our knowledge, the achieved 38% model FLOPs utilization (MFU) for training Llama3 8B with a 1M-token sequence length on 4,096 GPUs represents one of the highest training efficiencies reported for long-context learning on HPC systems. Our code implementation is available at https://github.com/jqyin/ringX-attention.

Yin, Junqi [ORNL] (ORCID:0000000338435520)↗
DOE OSTI2025

Activation Domain Hunter (ADhunter) v2.0

ADhunter is a software program that enables accurate identification and quantification of transcriptional activation domains. Unlike previous software, ADhunter uses protein representations from a pre-trained protein language model, model ensembling, and a training dataset from a diverse sampling of protein sequence space for state-of-the-art performance. These advantages enable improved perception of transcriptional activation domains across sequence space that can be used for mapping natural genetic circuits and engineering synthetic genetic circuits. In particular, ADhunter enables fine-tuned control of gene expression through synthetic transcription factors that can be used for complex control of cellular programs.

Waldburger, Lucas [Lawrence Berkeley National Labo↗
DOE OSTI2026

Dataset for Top Model Decision Tree: Selecting Segmentation Models for Reliable Quantitative Analysis in Low- and Ultralow-Dose CryoEM

Motivation Multiple deep learning model architectures can be used to segment bacterial membranes in cryoEM images. However, an AI-based tool advancement is often presented with only a single segmentation model for broad use, and this single model may show inconsistent results across datasets from different users. Here, we present the Top Model Decision Tree, a model screening framework to screen for the best model to generate bacterial inner and outer membrane masks based on user priorities. We use pre-trained segmentation models from YOLOv11, YOLO26, U-Net, Detectron2 and SAM3 fine-tuned on bacterial inner and outer membranes imaged with cryoEM. Run the Framework This notebook must be opened in Google Colab. Mount Google Drive and run with a GPU-based runtime. Open the notebook and follow steps to git clone in folders and files within this repository. There will be a repeating top_model_decision_tree.ipynb (notebook clone) that will not be used. Save your .png binary mask files and .csv table outputs within your Google Drive or download before closing the notebook. The models and all analysis/training scripts are available at [GitHub: https://github.com/Lynnicia/CryoEM_membranes_top_model_decision_tree and https://github.com/Sireesiru/Semantic-Segmentation-of-bacterial-cell-envelope-using-U-Nets.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2026

Fidelity-preserving enhancement of ptychography with foundational text-to-image models

Ptychographic phase retrieval enables high-resolution imaging of complex samples but often suffers from artifacts such as grid pathology and multislice crosstalk, which degrade reconstructed images. We propose a plug-and-play (PnP) framework that integrates physics model-based phase retrieval with text-guided image editing using foundational diffusion models. By employing the alternating direction method of multipliers, our approach ensures consensus between data fidelity and artifact removal subproblems, maintaining physical consistency while enhancing image quality. Artifact removal is achieved using a text-guided diffusion image editing method (LEDITS++) with a pre-trained foundational diffusion model, allowing users to specify artifacts for removal in natural language. Demonstrations on simulated and experimental datasets show significant improvements in artifact suppression and structural fidelity, validated by metrics such as peak signal-to-noise ratio and diffraction pattern consistency. This work highlights the combination of text-guided generative models and model-based phase retrieval algorithms as a transferable and fidelity-preserving method for high-quality diffraction imaging.

image editing↗
DOE OSTI2026

RouteE API

This is the API endpoint for RouteE energy prediction, which can be used to get both single vehicle link or route energy estimates and transportation network-wide energy consumption estimates for a variety of vehicles. This enables external researchers and transportation engineers to access and utilize NLR's growing library of pre-trained vehicle models for prediction of transportation energy consumption. This API provides three endpoints: - /route: Energy estimation of a vehicle over a planning link or sequence of links (route). - /network: Network-wide estimation of energy consumption for all vehicle traffic in the desired area. - /compass: Energy-optimal “eco-routing” between input origin and destination coordinates (Currently in beta for Denver metro area only).

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2026

Structure-Aware Unsupervised, Transformational Machine Learning for Drug Discovery (DTRA Basic Research Final Report)

The major goal of this project is to develop machine learning (ML) methods to enable improved predictive power on real drug discovery for novel targets. More specifically, we planned to demonstrate the capability and effectiveness of ML tools utilizing unlabeled large-volume protein-ligand datasets. We investigated multiple pre-training approaches for 3D protein-ligand structure-based foundation models, without relying on experimental binding data. We also addressed scenarios in which crystal structures are unavailable or binding data are limited. We also planned to develop a complete pipeline to screen novel compounds as well as to demonstrate the capability and effectiveness of the developed methods by testing on a realistic drug discovery task such as SARS-CoV-2. While the major goals and milestones remain consistent with the original proposal, certain technical details have been adjusted, based on the experimental results and related outcomes.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Scalable foundation models for numerical simulations on HPC platforms

In recent years, foundation models (FMs) have begun to reshape numerical simulations on high-performance computing (HPC) platforms. These large, pre-trained AI models enable rapid predictions across a broad range of physical domains, including Earth system modeling, fluid dynamics, materials science, as well as complex multi-modal simulations in aerospace engineering and fusion research. By training on diverse datasets, FMs learn intricate relationships and underlying physical behavior while also enabling the quantification of uncertainty in their predictions. This capability allows simulations that once required days of numerical calculation to be completed in minutes (FM inference), supporting real-time design optimization, uncertainty-aware decision making, and more comprehensive exploration of complex scenarios.

AI↗
DOE OSTI2024

Active Learning for Rapid Targeted Synthesis of Compositionally Complex Alloys

The next generation of advanced materials is tending toward increasingly complex compositions. Synthesizing precise composition is time-consuming and becomes exponentially demanding with increasing compositional complexity. An experienced human operator does significantly better than a novice but still struggles to consistently achieve precision when synthesis parameters are coupled. The time to optimize synthesis becomes a barrier to exploring scientifically and technologically exciting compositionally complex materials. This investigation demonstrates an active learning (AL) approach for optimizing physical vapor deposition synthesis of thin-film alloys with up to five principal elements. We compared AL-based on Gaussian process (GP) and random forest (RF) models. The best performing models were able to discover synthesis parameters for a target quinary alloy in 14 iterations. We also demonstrate the capability of these models to be used in transfer learning tasks. RF and GP models trained on lower dimensional systems (i.e., ternary, quarternary) show an immediate improvement in prediction accuracy compared to models trained only on quinary samples. Furthermore, samples that only share a few elements in common with the target composition can be used for model pre-training. We believe that such AL approaches can be widely adapted to significantly accelerate the exploration of compositionally complex materials.

Chemistry↗
DOE OSTI2026

Discovering Strong Gravitational Lenses in the Dark Energy Survey with Interactive Machine Learning and Crowd-sourced Inspection with Space Warps

We conduct a search for strong gravitational lenses in the Dark Energy Survey (DES) Year 6 imaging data. We implement a pre-trained Vision Transformer (ViT) for our machine learning (ML) architecture and adopt interactive machine learning to construct a training sample with multiple classes to address common types of false positives. Our ML model reduces ∼236 million DES cutout images to 22,564 targets of interest, including ∼85% of previously reported galaxy–galaxy lens candidates discovered in DES. These targets were visually inspected by citizen scientists, who ruled out ∼90% as false positives. Of the remaining 2618 candidates, 149 were expert-classified as “definite” lenses and 516 as “probable” lenses, for a total of 665 systems, with 147 of these candidates being newly identified. Additionally, we trained a second ViT to find double-source plane lens systems, finding at least one double-source system. Our main ViT excels at identifying galaxy–galaxy lenses, consistently assigning high scores to candidates with high expert assessments. The top 800 ViT-scored images include ∼100 of our “definite” lens candidates. This selection is an order of magnitude higher in purity than previous convolutional neural-network-based lens searches and demonstrates the feasibility of applying our methodology for discovering large samples of lenses in future surveys.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2026

A Fast, Accurate Prediction for System-Wide Damage Due to Dynamic Wind Loading

The complex relationship between photovoltaic (PV) hardware configurations, overall system dynamics, and turbulent aerodynamic phenomena generates highly unsteady, non-uniform loads that can lead to damaging instabilities. These effects may result in glass breakage, cell cracking, and structural failures in frames and mounting systems, even under moderate wind conditions. Addressing industry concerns about premature system failures in field conditions deemed survivable, our research aims to develop a fast and accurate predictive model for system damage. This model integrates configurable hardware choices with advanced simulation tools to represent the overall system-specific dynamics effectively. Using this model, we predict responses under varying weather conditions and hardware setups, translating these predictions into pre-trained surrogate models capable of accurately identifying failure risks and rapidly testing new system hardening measures. In this presentation, we will showcase preliminary results in capturing system dynamics through our customizable library of PV hardware configurations. Additionally, we will highlight how these new tools build upon PVade's established wind load modeling capabilities and foster the development of advanced AI/ML surrogates for improving system robustness.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2026

Incremental Learning for Passive Microwave Precipitation Retrievals using Advanced Technology Microwave Sounder

Spaceborne passive microwave (PMW) radiometry is central to global precipitation monitoring, yet retrieval uncertainties remain substantial, particularly for cross-track sounders whose variable footprints and channel configurations are optimized for atmospheric temperature and moisture profiling rather than precipitation. Consequently, existing operational products often exhibit angular-dependent biases, limited effective swath utilization, unrealistic rainfall probability distributions, and systematic misclassification of precipitation phase. These limitations are further compounded by the scarcity of globally accurate and representative precipitation observations, as training data from the Dual-frequency Precipitation Radar (DPR) and the Cloud Profiling Radar (CPR) are spatially sparse, lack uniform global coverage, and exhibit heterogeneous error characteristics across precipitation regimes. To address these challenges, this study presents a supervised retrieval algorithm that incrementally trains an ensemble of extreme gradient-boosted decision trees by augmenting base learners with pre-training on reanalysis data and post-training on coincident DPR and CPR observations matched with the Advanced Technology Microwave Sounder (ATMS). By transferring prior information from reanalysis to posterior constraints from radar observations and adopting a sequential detection–estimation strategy for precipitation phase and rate retrieval, the proposed approach yields retrievals across the full ATMS swath that are largely free from persistent deficiencies in current Global Precipitation Measurement (GPM) passive microwave operational products. In particular, the method resolves bimodal artifacts in rainfall retrievals and mitigates systematic high-latitude snowfall biases, including overestimation across the Arctic and underestimation across the Antarctic. Validation against independent Multi-Radar Multi-Sensor (MRMS) data over the Contiguous United States (CONUS) further demonstrates improved performance in precipitation phase detection and rate estimation relative to both reanalysis and current GPM PMW products.

Mahyar Garshasbi↗
DOE OSTI2026

Artificial Intelligence for Event Reconstruction and Higgs Physics at CMS and Future Colliders

This dissertation charts a trajectory in which advances in artificial intelligence (AI) play a central role in pushing the high-energy physics frontier, complementing progress driven by higher collision energies and larger colliders. The discovery potential of the LHC and future colliders relies on accurate reconstruction of increasingly complex particle collision events. In the CMS experiment, this task is performed by the particle-flow (PF) algorithm. This dissertation presents the first implementation of a machine-learning-based particle-flow (MLPF) reconstruction in the CMS detector based on transformer architectures. In simulated top quark--antiquark pair (ttbar) events under LHC Run~3 (2023--2024) conditions, MLPF improves jet energy resolution by 10--20\% compared to standard PF for jets with transverse momentum between 30--100\GeV. Runtime performance is evaluated using simulated multijet events, with a median inference time of 20\unit{ms} per event on an NVIDIA L4 GPU, compa red to approximately 110\unit{ms} for standard PF. The MLPF algorithm is also validated on Run~3 collision data, representing the first data-validated ML-based reconstruction pipeline at any LHC experiment. We then extend MLPF toward future electron--positron colliders and introduce the first full-simulation cross-detector transfer learning workflow for PF reconstruction. The model is pre-trained on simulated events from the Compact Linear Collider detector (CLICdet) and fine-tuned on the CLIC-like detector (CLD) proposed for the Future Circular Collider (FCC). This approach achieves up to a 40\% improvement in jet energy resolution over rule-based reconstruction while reducing the required training dataset size by an order of magnitude, demonstrating the potential of AI to accelerate detector development and optimization. This dissertation also demonstrates how modern AI techniques enhance the sensitivity of LHC physics analyses. A CMS search for highly Lorentz-boosted Higgs bosons decaying to \textrm{W} boson pairs is presented, focusing on the single-lepton final state. A dedicated fine-tuning strategy for \ParT yields an approximately 70\% increase in expected sensitivity relative to the baseline model. The analysis uses proton--proton collision data at a center-of-mass energy of \ensuremath{\sqrt{s}=13\TeV} collected by CMS between 2016 and 2018, corresponding to an integrated luminosity of 138\ensuremath{\ \mathrm{fb}^{-1}}. The expected significance of the search is $1.86\sigma$, with an observed signal strength of $-0.19^{+0.48}_{-0.46}$. Finally, explainable AI techniques are applied to the MLPF and \ParticleNet algorithms using layerwise relevance propagation, showing that both models base their predictions on physically meaningful features consistent with our physics intuition. Together, these results demonstrate how advanced AI methods can enhance reconstruction, analysis sensitivity, and interpretability, shaping the next era of experimental parti cle physics.

Mokhtar, Farouk [UC, San Diego]↗
DOE OSTI2025

A geometric framework for momentum-based optimizers for low-rank training

Low-rank pre-training and fine-tuning have recently emerged as promising techniques for reducing the computational and storage costs of large neural networks. Training low-rank parameterizations typically relies on conventional optimizers such as heavy ball momentum methods or Adam. In this work, we identify and analyze potential difficulties that these training methods encounter when used to train low-rank parameterizations of weights. In particular, we show that classical momentum methods can struggle to converge to a local optimum due to the geometry of the underlying optimization landscape. To address this, we introduce novel training strategies derived from dynamical low-rank approximation, which explicitly account for the underlying geometric structure. Our approach leverages and combines tools from dynamical low-rank approximation and momentum-based optimization to design optimizers that respect the intrinsic geometry of the parameter space. We validate our methods through numerical experiments, demonstrating faster convergence, and stronger validation metrics at given parameter budgets.

Schotthoefer, Steffen [ORNL] (ORCID:00000002156965↗
DOE OSTI2026

Neural Scaling Laws for Jet Generation

Recently observed empirical scaling laws describe the performance of foundation-type models as three independent key quantities -- dataset size, compute, and model parameters -- are modified. Extracting these scaling laws informs the training of large complex models for which the tuning of hyperparameters in traditional ways is not feasible. This work for the first time explores if scaling laws can also be observed for the task of particle jet generation -- both relevant as a pre-training objective for foundation models and as in-situ simulation by itself. We indeed replicate the key logarithmic scaling law behavior for model-size scaling. Beyond studying the next token prediction validation loss of the generative model, we also study the sliced Wasserstein distance of five physical quantities that are not immediately available to the model during training. Our study shows that this quantity is monotonically related to the next token prediction validation loss, meaning that this loss is indeed a good proxy for the physics performance. For the scaling with dataset size and compute, we observe substantially weaker scaling behavior of both the loss and the sliced Wasserstein distance. We analyze this behavior by introducing the concept of a learnable window, and argue that autoregressive next token prediction on jet constituents exhibits comparatively rapid saturation relative to language-model studies. We discuss possible origins of this behavior, including the stochastic nature of QCD radiation and differences between generative and supervised learning tasks in collider physics.

Amram, Oz [Fermilab]↗