Search NASA⌕ Search

Engineering topics

Sun, Yu

Publications and source records attributed to Sun, Yu.

At least 19 records
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Demystifying the Resilience of Large Language Models: An End-to-End Perspective

Deep neural networks are known to be resilient to random bit-wise faults in their parameters. However, this resilience has primarily been established through evaluations of classification models. The extent to which this claim holds for large-language models remains underexplored. In this work, we conduct an extensive measurement study on the impact of random bitwise faults in commercial-scale language models. We perform an in-depth analysis of the resulting generation outputs. We first expose that these language models are not truly resilient to random bit-flips. While aggregate metrics such as accuracy may suggest resilience, an in-depth inspection of the generated outputs shows significant degradation in text quality. Our analysis also shows that tasks requiring more complex reasoning suffer more from performance and quality degradation. Moreover, we extend our analysis to models with augmented reasoning capabilities, such as Chain-of-Thought or Mixture of Experts architectures, and characterize their failure scenarios under random bit-flips.

Sun, Yu↗
DOE OSTI2025

Spin Frustration and Unconventional Surface Spin Canting State in Van der Waals Ferromagnet/Antiferromagnet Heterostructures

Atomically flat surfaces of van der Waals (vdW) materials pave an avenue for addressing a long-standing fundamental issue of how a compensated antiferromagnet (AFM) surface frustrates a ferromagnetic (FM) overlayer in FM/AFM heterostructures. We investigate Fe5GeTe2/NiPS3 vdW heterostructures by characterizing AFM and FM spins separately. We find that in-plane zig-zag AFM NiPS3 develops three equivalent AFM domains, which are robust against external magnetic field and magnetic coupling with Fe5GeTe2. Moreover, evidence is provided of in-plane-AFM-induced perpendicular magnetic anisotropy (PMA) in adjacent Fe5GeTe2, and an unconventional out-of-plane surface spin canting state with the Fe5GeTe2 spins spatially turn from out-of-plane direction near the interface to in-plane direction away from the interface in Fe5GeTe2/NiPS3. The out-of-plane surface spin canting is a unique property of spin frustration in vdW magnetic heterostructures.

Wang, Tianye↗
DOE OSTI2023

End-to-End Pipeline for Trigger Detection on Hit and Track Graphs

There has been a surge of interest in applying deep learning in particle and nuclear physics to replace labor-intensive offline data analysis with automated online machine learning tasks. This paper details a novel AI-enabled triggering solution for physics experiments in Relativistic Heavy Ion Collider and future Electron-Ion Collider. The triggering system consists of a comprehensive end-to-end pipeline based on Graph Neural Networks that classifies trigger events versus background events, makes online decisions to retain signal data, and enables efficient data acquisition. Here, the triggering system first starts with the coordinates of pixel hits lit up by passing particles in the detector, applies three stages of event processing (hits clustering, track reconstruction, and trigger detection), and labels all processed events with the binary tag of trigger versus background events. By switching among different objective functions, we train the Graph Neural Networks in the pipeline to solve multiple tasks: the edge-level track reconstruction problem, the edge-level track adjacency matrix prediction, and the graph-level trigger detection problem. We propose a novel method to treat the events as track-graphs instead of hit-graphs. This method focuses on intertrack relations and is driven by underlying physics processing. As a result, it attains a solid performance (around 72% accuracy) for trigger detection and outperforms the baseline method using hit-graphs by 2% higher accuracy.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Rutin is a potent senomorphic agent to target senescent cells and can improve chemotherapeutic efficacy

Aging is a major risk factor for most chronic disorders, for which cellular senescence is one of the central hallmarks. Senescent cells develop the pro-inflammatory senescence-associated secretory phenotype (SASP), which significantly contributes to organismal aging and age-related disorders. Development of senotherapeutics, an emerging class of therapeutic agents to target senescent cells, allows to effectively delay aging and alleviate chronic pathologies. Here we report preliminary outputs from screening of a natural medicinal agent (NMA) library for senotherapeutic candidates and validated several agents with prominent potential as senomorphics. Rutin, a phytochemical constituent found in a number of plants, showed remarkable capacity in targeting senescent cells by dampening expression of the full spectrum SASP. Further analysis indicated that rutin restrains the acute stress-associated phenotype (ASAP) by specifically interfering with the interactions of ATM with HIF1α, a master regulator of cellular and systemic homeostasis activated during senescence, and of ATM with TRAF6, part of a key signaling axis supporting the ASAP development toward the SASP. Conditioned media produced by senescent stromal cells enhanced the malignant phenotypes of prostate cancer cells, including in vitro proliferation, migration, invasion, and more importantly, chemoresistance, while rutin remarkably downregulated these gain-of-functions. Although classic chemotherapy reduced tumor progression, the treatment outcome was substantially improved upon combination of a chemotherapeutic agent with rutin. Our study provides a proof of concept for rutin as an emerging natural senomorphic agent, and presents an effective therapeutic avenue for alleviating age-related pathologies including cancer.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2023

Coordinate-Based Seismic Interpolation in Irregular Land Survey: A Deep Internal Learning Approach

Physical and budget constraints often result in irregular sampling, which complicates accurate subsurface imaging. Preprocessing approaches, such as missing trace or shot interpolation, are typically employed to enhance seismic data in such cases. Recently, deep learning has been used to address the trace interpolation problem at the expense of large amounts of training data to adequately represent typical seismic events. Nonetheless, most research in this area has focused on trace reconstruction, with little attention having been devoted to shot interpolation. Furthermore, existing methods assume regularly spaced receivers/sources failing in approximating seismic data from real (irregular) surveys. This work presents a novel shot gather interpolation approach which uses a continuous coordinate-based representation of the acquired seismic wavefield parameterized by a neural network. The proposed unsupervised approach, which we call coordinate-based seismic interpolation (CoBSI), enables the prediction of specific seismic characteristics in irregular land surveys without using external data during neural network training. Importantly, experimental results on real and synthetic 3-D data validate the ability of the proposed method to estimate continuous smooth seismic events in the time-space and frequency-wavenumber domains, improving sparsity or low-rank-based interpolation methods.

58 GEOSCIENCES↗
DOE OSTI2023

Trigger Detection for the sPHENIX Experiment via Bipartite Graph Networks with Set Transformer

Trigger (interesting events) detection is crucial to high-energy and nuclear physics experiments because it improves data acquisition efficiency. It also plays a vital role in facilitating the downstream offline data analysis process. The sPHENIX detector, located at the Relativistic Heavy Ion Collider in Brookhaven National Laboratory, is one of the largest nuclear physics experiments on a world scale and is optimized to detect physics processes involving charm and beauty quarks. Furthermore, these particles are produced in collisions involving two proton beams, two gold nuclei beams, or a combination of the two and give critical insights into the formation of the early universe. This paper presents a model architecture for trigger detection with geometric information from two fast silicon detectors. Transverse momentum is introduced as an intermediate feature from physics heuristics. We also prove its importance through our training experiments. Each event consists of tracks and can be viewed as a graph. A bipartite graph neural network is integrated with the attention mechanism to design a binary classification model. Compared with the state-of-the-art algorithm for trigger detection, our model is parsimonious and increases the accuracy and the AUC score by more than 15%.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

The flavonoid procyanidin C1 has senotherapeutic activity and increases lifespan in mice

Ageing-associated functional decline of organs and increased risk for age-related chronic pathologies is driven in part by the accumulation of senescent cells, which develop the senescence-associated secretory phenotype (SASP). Here we show that procyanidin C1 (PCC1), a polyphenolic component of grape seed extract (GSE), increases the healthspan and lifespan of mice through its action on senescent cells. By screening a library of natural products, we find that GSE, and PCC1 as one of its active components, have specific effects on senescent cells. At low concentrations, PCC1 appears to inhibit SASP formation, whereas it selectively kills senescent cells at higher concentrations, possibly by promoting production of reactive oxygen species and mitochondrial dysfunction. In rodent models, PCC1 depletes senescent cells in a treatment-damaged tumour microenvironment and enhances therapeutic efficacy when co-administered with chemotherapy. Intermittent administration of PCC1 to either irradiated, senescent cell-implanted or naturally aged old mice alleviates physical dysfunction and prolongs survival. We identify PCC1 as a natural senotherapeutic agent with in vivo activity and high potential for further development as a clinical intervention to delay, alleviate or prevent age-related pathologies.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Scalable Plug-and-Play ADMM with Convergence Guarantees

Plug-and-play priors (PnP) is a broadly applicable methodology for solving inverse problems by exploiting statistical priors specified as denoisers. Recent work has reported the state-of-the-art performance of PnP algorithms using pre-trained deep neural nets as denoisers in a number of imaging applications. However, current PnP algorithms are impractical in large-scale settings due to their heavy computational and memory requirements. This work addresses this issue by proposing an incremental variant of the widely used PnP-ADMM algorithm, making it scalable to problems involving a large number measurements. Here, we theoretically analyze the convergence of the algorithm under a set of explicit assumptions, extending recent theoretical results in the area. Additionally, we show the effectiveness of our algorithm with nonsmooth data-fidelity terms and deep neural net priors, its fast convergence compared to existing PnP algorithms, and its scalability in terms of speed and memory.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

SGD-Net: Efficient Model-Based Deep Learning with Theoretical Guarantees

Deep unfolding networks have recently gained popularity for solving imaging inverse problems. However, the computational and memory complexity of data-consistency layers within traditional deep unfolding networks scales with the number of measurements, limiting their applicability to large-scale imaging inverse problems. We propose SGD-Net as a new methodology for improving the efficiency of deep unfolding through stochastic approximations of the data-consistency layers. Our theoretical analysis shows that SGD-Net can be trained to approximate batch deep unfolding networks to an arbitrary precision. Our simulations on intensity diffraction tomography and sparse-view computed tomography show that SGD-Net can match the performance of the traditional batch network at a fraction of training and testing complexity.Deep unfolding networks have recently gained popularity for solving imaging inverse problems. However, the computational and memory complexity of data-consistency layers within traditional deep unfolding networks scales with the number of measurements, limiting their applicability to large-scale imaging inverse problems. We propose SGD-Net as a new methodology for improving the efficiency of deep unfolding through stochastic approximations of the data-consistency layers. Our theoretical analysis shows that SGD-Net can be trained to approximate batch deep unfolding networks to an arbitrary precision. Our simulations on intensity diffraction tomography and sparse-view computed tomography show that SGD-Net can match the performance of the traditional batch network at a fraction of training and testing complexity.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Boosting the Performance of Plug-and-Play Priors via Denoiser Scaling

Plug-and-play priors (PnP) is an image reconstruction framework that utilizes an image denoiser as an imaging prior. Unlike traditional regularized inversion, PnP does not require the prior to be expressible in the form of a regularization function. This flexibility enables PnP algorithms to exploit the most effective image denoisers, leading to their state-of-the-art performance in various imaging tasks. However, many powerful denoisers, such as the ones based on convolutional neural networks (CNNs), do not have tunable parameters that would allow controlling their influence within PnP. To address this issue, in this paper, we introduce a scaling parameter that adjusts the magnitude of the denoiser input and output. We theoretical justify the denoiser scaling from the perspectives of proximal optimization, statistical estimation, and consensus equilibrium. Finally, we provide numerical experiments demonstrating the ability of denoiser scaling to systematically improve the performance of PnP for denoising CNN priors that do not have explicitly tunable parameters.

Boosting↗
DOE OSTI2020

Provable Convergence of Plug-and-Play Priors With MMSE Denoisers

Plug-and-play priors (PnP) is a methodology for regularized image reconstruction that specifies the prior through an image denoiser. While PnP algorithms are well understood for denoisers performing maximum a posteriori probability (MAP) estimation, they have not been analyzed for the minimum mean squared error (MMSE) denoisers. Here we address this gap by establishing the first theoretical convergence result for the iterative shrinkage/thresholding algorithm (ISTA) variant of PnP for MMSE denoisers. We show that the iterates produced by PnP-ISTA with an MMSE denoiser converge to a stationary point of some global cost function. We validate our analysis on sparse signal recovery in compressive sensing by comparing two types of denoisers, namely the exact MMSE denoiser and the approximate MMSE denoiser obtained by training a deep neural net.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2019

Cancer cell membrane-coated rare earth doped nanoparticles for tumor surgery navigation in NIR-II imaging window

With the attractive characteristics of low toxicity, narrow emission spectra, long fluorescence lifetime, and no photobleaching, rare-earth doped nanoparticles (RENPs) have become promising imaging nanoplatforms in the second near-infrared window (NIR-II, 1000–1700 nm). Many studies have been devoted to different surface modification strategies of RENPs to achieve good NIR-II tumor imaging property. However, the tumor imaging efficiency of these approaches remains disappointing. Herein, two methods were pursued to modify RENPs with either polyethylene glycol (PEG) or cancer-cell-membrane to generate PEGylated-RENPs and cancer-cell-membrane camouflaged RENPs (CC-RENPs), respectively. Furthermore, this study demonstrated that cancer-cell-membrane coating of RENPs greatly improved tumor imaging performance in NIR-II window than PEG coating. Besides, CC-RENPs uptakes in liver and spleen significantly decreased. Importantly, CC-RENPs was found to be capable to aid us to pursue precise resection of the tumor. In conclusion, CC-RENPs are capable of targeting tumors and show high tumor NIR-II imaging quality. They are highly promising nanomaterials for future medical applications.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2019

Targeting amphiregulin (AREG) derived from senescent stromal cells diminishes cancer resistance and averts programmed cell death 1 ligand (PD-L1)-mediated immunosuppression

Aging is characterized by a progressive loss of physiological integrity, while cancer represents one of the primary pathological factors that severely threaten human lifespan and healthspan. In clinical oncology, drug resistance limits the efficacy of most anticancer treatments, and identification of major mechanisms remains a key to solve this challenging issue. Here, we highlight the multifaceted senescence-associated secretory phenotype (SASP), which comprises numerous soluble factors including amphiregulin (AREG). Production of AREG is triggered by DNA damage to stromal cells, which passively enter senescence in the tumor microenvironment (TME), a process that remarkably enhances cancer malignancy including acquired resistance mediated by EGFR. Furthermore, paracrine AREG induces programmed cell death 1 ligand (PD-L1) expression in recipient cancer cells and creates an immunosuppressive TME via immune checkpoint activation against cytotoxic lymphocytes. Targeting AREG not only minimized chemoresistance of cancer cells, but also restored immunocompetency when combined with classical chemotherapy in humanized animals. Our study underscores the potential of in vivo SASP in driving the TME-mediated drug resistance and shaping an immunosuppressive niche, and provides the proof of principle of targeting major SASP factors to improve therapeutic outcome in cancer medicine, the success of which can substantially reduce aging-related morbidity and mortality.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2019

Kinetics of Recycling CF/EP Composites by Thermal Excitation of Cr{sub 2}O{sub 3}

Recycling high performance carbon fibers (CFs) by thermal excitation of Cr{sub 2}O{sub 3} semiconductor was proposed. The kinetic model of decomposition of epoxy resin was established and the method for solving kinetic parameters was proposed. The influences of temperature and treatment time on the decomposition rate of epoxy resin were investigated. The kinetic equation was established and the accuracy was verified experimentally. The morphology of the recycled CF was observed and monofilament tensile strength was determined. Results indicated that temperature and treatment time had positive correlation with decomposition rate of the epoxy resin. The reaction order n was 1.5, with excitation energy E of 85.03 kJ·mol{sup −1} and pre-exponential factor k{sub 0} of 1.26 × 10{sup 6} min{sup −1}. The established kinetic equation could be used to estimate the temperature and treatment time. Prolongation of treatment time could improve the monofilament tensile strength. The monofilament tensile strength of the recycled CF at 480 °C for 30 min was 102.72% of that of the original CF, indicating that surface heat treatment was performed on CF during the recycling process.

36 MATERIALS SCIENCE↗