MD-SOAR Maryland Shared Open Access Repository
Not a member yet
34521 research outputs found
Sort by
Scalable Deep Learning for Greenland Ice Bed Topography Prediction
No full textThe rapid evolution of deep learning and GPU-accelerated computing has revolutionized large-scale geospatial modeling, yet the application of these advances to high-resolution ice-sheet bed topography prediction remains challenging due to the sheer volume of input data and the computational demands of deep convolutional neural networks. In this thesis, we investigate the performance tradeoffs of increasing spatial extents on the Greenland Upernavik dataset. The experiments are conducted on distributed multi-node multi-GPU training pipelines on the BedTopoCNN architecture. Three spatial extents—600×600, 1200×1200, and 2400×2400 pixels—are benchmarked on both the Frontera and CHIP highperformance computing clusters via PyTorch’s DistributedDataParallel framework over up to four nodes. We measure epoch time, strong scaling efficiency, and analyze how larger input patches impact convergence speed and resource utilization. Overall, our benchmarks show that while smaller 600×600 patches deliver the best throughput, enabling full 20000-epoch runs in under 48 hours with excellent scaling and stable accuracy. After expanding to 1200×1200, it hits a practical limit on single GPUs but gains efficiency from multi-GPU parallelism. Finally, pushing to 2400×2400 overwhelms even 16-GPU configurations. But adding a bottleneck layer helps substantially to meet the 48-hour training time at the largest spatial extent
Imaging Spectroscopy for Enhancing Regional Coastal Wetland Extent Monitoring
No full textCoastal wetland habitats contribute to important ecosystem services, including improved water quality, carbon sequestration, and flood mitigation, but they are difficult to be monitored in situ due to their inaccessibility. This research aims to explore the potential and limitation of monitoring coastal wetland habitats in the Greater Cape Floristic Region (GCFR), South Africa, with the NASA Earth Mineral Dust Source Investigation (EMIT). High-resolution habitat extent derived from PlanetScope imagery was used to train random forest regression algorithms with EMIT data. We estimate the subpixel extent of three wetland habitats (salt marshes, reeds and sedges, and submerged aquatic vegetation (SAV)). We validate model performance using a testing estuary unseen by the model. The best-performing model achieved a root mean square error (RMSE) of 7.8% (281 m²) for salt marsh, 11.4% (410 m²) for reeds and sedges, and 5.6% (202 m²) for SAV. At higher tidal stages, model performance decreased with RMSE of 23.6% for salt marsh and 12.2% for SAV, underscoring the influence of tidal inundation on mapping accuracy of these habitats. These findings illustrate the need to select tidal stages when mapping these habitats, particularly the importance of imagery acquired at low tidal stages during the growing season. This study shows that when medium-resolution imaging spectroscopy is combined with machine learning, we can estimate subpixel habitat extent, addressing the spatial limitations of the EMIT imaging spectrometer. With more data, our approach could provide information on long-term trends and changes in these ecosystems. EMIT-based subpixel monitoring of coastal wetlands is possible and can provide important information on the extent and change of these ecosystems.PT and AE were supported by the University of the Witwatersrand grant, while ADC was supported by the NASA Grant 80NSSC22K1382. We appreciate the support of NASA's Biological Survey of the Cape (BioScape) for making this project successful. The authors appreciate the assistance of the administrators and professionals of the Department of Conservation Services, South African National Parks, Sedgefield, Langebaan, South Africa, especially Kyle Smith, Ruth-Marry Fisher, and Danielle Seymour, for their expertise and field guidance.https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2025JG00877
Evidence for Supermassive Black Hole Binaries
No full textWe review the state of the evidence for the existence and observational appearance of supermassive black hole binaries. Such objects are expected from standard hierarchical galaxy evolution to form after two galaxies, each containing a supermassive black hole, have merged, in the centre of the merger remnant. A complex interaction is predicted to take place with stars and gas in the host galaxy, leading to observable signatures in weakly as well as actively accreting phases. Direct observational evidence is available and shows examples of dual active galactic nuclei from kpc scales down to parsec scales. Signatures of possibly closer supermassive black hole binaries may be seen in jetted black holes. The interaction with stars and gas in a galaxy significantly affects the hardening of the binary and hence contributes to uncertainties of the expected gravitational wave signal. The Laser Interferometer Space Antenna (LISA) should in the future detect actual mergers. Before the launch of LISA, pulsar timing arrays may have the best chance to detect a gravitational wave signal from supermassive black hole binaries. The first signs of the combined background of inspiralling objects might have been seen already.This review was inspired by a discussion meeting hosted by the Royal Astronomical Society in London on 14 April 2023. J.M.S.R acknowledges support from the Science and Technology Facilities Council (STFC) under grant ST/V50659X/1 (project reference 2442592) and thanks Andy Fabian, Jiachen Jiang and Dom Walton for useful discussions. M.A.B acknowledges support from a UKRI Stephen Hawking Fellowship (EP/X04257X/1) and the Science and Technology Facilities Council (STFC), grant code ST/W000997/1. H. M. acknowledges the support of the UK Space Agency, Grant No. ST/V002813/1 and ST/X002071/1. RN acknowledges funding from UKRI/EPSRC through a Stephen Hawking Fellowship (EP/T017287/1). M.G.H.K acknowledges the hospitality of the Institute of Astronomy, University of Cambridge, UK for part of the writing of this manuscript.http://arxiv.org/abs/2510.0753
In vivo analysis of active versus self-tapping implants with SLA surface treatment
No full textSurface modifications and implant thread designs play a crucial role in achieving primary stability and osseointegration. Therefore, it is of interest to compare the short-term clinical and radiographic outcomes of active versus self-tapping titanium implants with sandblasted, large-grit, acid-etched (SLA) surfaces in 20 systemically healthy patients requiring posterior implant-supported prostheses. Clinical and radiographic parameters were assessed at 4, 7, 12 and 26 weeks and statistical analysis showed significant intra-group improvements in probing depth, attachment levels, mucosal margin stability and bone gain. However, intergroup comparisons at all-time points revealed no statistically significant differences, indicating comparable performance between the two implant designs. Overall, both active and self-threaded SLA implants demonstrated favorable short-term clinical outcomes, supporting their effectiveness in promoting early osseointegration and potential use in early loading protocols.https://www.bioinformation.net/021/973206300213152.ht
XRISM/Xtend Transient Search (XTS) detected an X-ray flare from an eclipsing binary
No full textAuthors: Y. Ishihara (Chuo U.), K. Fukushima, Y. Kanemaru, S. Ogawa (JAXA), M. Audard (U. de Geneve), E. Behar (Technion), S. Inoue (Kyoto U.), T. Kohmura (TUS), Y. Maeda (JAXA), M. Mizumoto (UTEF), N. Nagashima (Chuo U.), M. Nobukawa (NUE), K. Pottschmidt (UMBC, NASA GSFC, CRESST), M. Shidatsu (Ehime U.), H. Sugai (Chuo U.), Y. Terada (Saitama U.), Y. Terashima (Ehime U.), Y. Tsuboi (Chuo U.), H. Uchida (Kyoto U.), T. Yoneyama (Chuo U.), M. Yoshimoto (Ehime U.)XRISM/Xtend Transient Search (XTS) detected an X-ray flare from an X-ray source XRISM J1746-2940 on 2025-04-12 TT. The source position is determined to be (R.A., Dec.) = (266.528, -29.671), with a systematic error of ~ 40 arcsec. A plausible counterpart is an eclipsing binary V734 Sgr located ~ 10 arcsec apart from the position of XRISM J1746-2940.
The flare started at 2025-04-12 at ~ 12:38 TT. The flare reached its peak on 2025-04-12 at ~ 13:12. The flare exponentially decayed in 10⁴ sec. The peak flux is estimated to be 2 × 10⁻¹¹ erg s⁻¹ cm⁻² (0.4 – 10.0 keV). A systematic error of roughly 20% should be added to the statistical error. Corresponding luminosity is 2 × D₉₇₈ₚ꜀ × 10³³ erg s⁻¹ by assuming the distance to XRISM J1746-2940 of D₉₇₈ₚ꜀.
We derived the above systematic error for the flux by comparing our derived values for the sources detected with XTS in several observations with those for the corresponding X-ray counterparts. We estimated the systematic error for the source position from the separations between the detected sources with the corresponding counterparts in the same field of view.https://www.astronomerstelegram.org/?read=1714
FedMentor: Domain-Aware Differential Privacy for Heterogeneous Federated LLMs in Mental Health
No full textPrivacy-preserving adaptation of Large Language Models (LLMs) in sensitive domains (e.g., mental health) requires balancing strict confidentiality with model utility and safety. We propose FedMentor, a federated fine-tuning framework that integrates Low-Rank Adaptation (LoRA) and domain-aware Differential Privacy (DP) to meet per-domain privacy budgets while maintaining performance. Each client (domain) applies a custom DP noise scale proportional to its data sensitivity, and the server adaptively reduces noise when utility falls below a threshold. In experiments on three mental health datasets, we show that FedMentor improves safety over standard Federated Learning without privacy, raising safe output rates by up to three points and lowering toxicity, while maintaining utility (BERTScore F1 and ROUGE-L) within 0.5% of the non-private baseline and close to the centralized upper bound. The framework scales to backbones with up to 1.7B parameters on single-GPU clients, requiring < 173 MB of communication per round. FedMentor demonstrates a practical approach to privately fine-tune LLMs for safer deployments in healthcare and other sensitive fields.The Second Workshop on GenAI for Health Potential, Trust, and Policy Compliance,GenAI4Health @NeurIPS 2025, December 6, 2025,California, USAhttp://arxiv.org/abs/2509.1427
Measurement of Cr and Ti fluxes and sub-iron/iron flux ratios with CALET on the International Space Station
No full text39th International Cosmic Ray Conference (ICRC2025), July 15–24 2025,Geneva, SwitzerlandAuthors: CALET Collaboration, O. Adriani, Y. Akaike, K. Asano, Y. Asaoka, E. Berti, P. Betti, G. Bigongiari, W.R. Binns, M. Bongi, P. Brogi, A. Bruno, N. Cannady, G. Castellini, C. Checchia, M.L. Cherry, G. Collazuol, G.A. de Nolfo, K. Ebisawa, A. W. Ficklin, H. Fuke, S. Gonzi, T.G. Guzik, T. Hams, K. Hibino, M. Ichimura, M.H. Israel, K. Kasahara, J. Kataoka, R. Kataoka, Y. Katayose, C. Kato, N. Kawanaka, Y. Kawakubo, K. Kobayashi, K. Kohri, H.S. Krawczynski, J.F. Krizmanic, P. Maestro, P.S. Marrocchesi, M. Mattiazzi, A.M. Messineo, J.W. Mitchell, S. Miyake, A.A. Moiseev, M. Mori, N. Mori, H.M. Motz, K. Munakata, S. Nakahira, J. Nishimura, M. Negro, S. Okuno, J.F. Ormes, S. Ozawa, L. Pacini, P. Papini, B.F. Rauch, S.B. Ricciarini, K. Sakai, T. Sakamoto, M. Sasaki, Y. Shimizu, A. Shiomi, P. Spillantini, F. Stolzi, S. Sugita, A. Sulaj, M. Takita, T. Tamura, T. Terasawa, S. Torii, Y. Tsunesada, Y. Uchihori, E. Vannuccini, J.P. Wefel, K. Yamaoka, S. Yanagita, A. Yoshida, K. Yoshida, and W. V. ZoberThe analysis of cosmic ray nuclei provides critical insights for a theoretical understanding of the acceleration and propagation mechanisms of charged particles in our Galaxy. A unique source of information on the average path length that cosmic rays travel before reaching Earth can be provided by the elements lying just below iron in the periodic table (sub-iron). Most of these elements are believed to be produced by the spallation of heavier nuclei as they propagate in the interstellar medium. The Calorimetric Electron Telescope (CALET), which has been operational on the International Space Station since 2015, has collected a substantial dataset comprising cosmic-ray (CR) iron and sub-iron events across a broad energy spectrum. In this contribution we present the measurements of the energy dependence of the titanium and chromium fluxes in cosmic rays, as their flux ratios to iron, in the energy interval from 10 GeV/n to 250 GeV/n. The measurements, based on data collected during eight years of operation, are reported with significantly enhanced precision compared to existing measurements, including a detailed assessment of systematic uncertainties. In addition to the sub-iron fluxes, an update of CALET’s analysis of the iron flux has been carried out up to 1.6 TeV/n.We gratefully acknowledge JAXA’s contributions to the development of CALET and to the operations aboard the JEM-EF on the ISS. This work was supported in part by JSPS Grant-in-Aid for Scientific Research (S) No. 26220708, No. 19H05608, and No. 24H00025, JSPS Grant-in-Aid for Scientific Research (B) No. 24K00665, and by the MEXTSupported Program for the Strategic Research Foundation at Private Universities (2011-2015) (No. S1101021) at Waseda University. The CALET effort in Italy is supported by ASI under Agreement No. 2013-018-R.0 and its amendments. The CALET effort in the United States is supported by NASA through Grants No. NNX16AB99G, No. NNX16AC02G, and No. NNH14ZDA001N-APRA-0075.https://pos.sissa.it/501/136
Cross Calibration of Galaxy Cluster Temperatures Measured with NuSTAR, XMM-Newton, and Chandra
No full textThe use of galaxy clusters to constrain cosmology is limited in part due to uncertainties in derived cluster masses, which often depend on the gas temperature. Unfortunately, there exists a longstanding discrepancy in temperature measurements of the same galaxy clusters made by the two most sensitive X-ray observatories, Chandra and XMM-Newton. The NuSTAR X-ray Observatory’s greater sensitivity to the exponential turnover in the bremsstrahlung continuum allows for more precise and potentially more accurate galaxy cluster temperature estimates, especially given its unique ability to independently calibrate its optics in orbit. We present new NuSTAR spectra of 10 relaxed (5 keV 2 keV effective area (∼5% at 5 keV) is found to explain the trend reasonably well. These results demonstrate the potential for NuSTAR data to address the two-decade-old temperature discrepancy between Chandra and XMM-Newton.F.L., D.R.W., and C.P. acknowledge support from the NuSTAR NASA JPL subcontract RSA1688622, NASA grant 80NSSC20K1000, and the NSF-supported REU program at the University of Utah. B.J.M. acknowledges support from Science and Technology Facilities Council grants ST/ V000454/1 and ST/Y002008/1. This work made use of data from the NuSTAR mission, which is led by the California Institute of Technology, managed by the Jet Propulsion Laboratory, and funded by the National Aeronautics and Space Administration. In addition, it employs a list of Chandra data sets, obtained by the Chandra X-ray Observatory, contained in the Chandra Data Collection DOI: 10.25574/ cdc.418.https://iopscience.iop.org/article/10.3847/1538-4357/adee8
Advancing Object Detection in Transportation with Multimodal Large Language Models (MLLMs): A Comprehensive Review and Empirical Testing
No full textThis study aims to comprehensively review and empirically evaluate the application of multimodal large language models (MLLMs) and Large Vision Models (VLMs) in object detection for transportation systems. In the first fold, we provide a background about the potential benefits of MLLMs in transportation applications and conduct a comprehensive review of current MLLM technologies in previous studies. We highlight their effectiveness and limitations in object detection within various transportation scenarios. The second fold involves providing an overview of the taxonomy of end-to-end object detection in transportation applications and future directions. Building on this, we proposed empirical analysis for testing MLLMs on three real-world transportation problems that include object detection tasks, namely, road safety attribute extraction, safety-critical event detection, and visual reasoning of thermal images. Our findings provide a detailed assessment of MLLM performance, uncovering both strengths and areas for improvement. Finally, we discuss practical limitations and challenges of MLLMs in enhancing object detection in transportation, thereby offering a roadmap for future research and development in this critical area.https://www.mdpi.com/2079-3197/13/6/13
Beyond parenting behaviors: Considering parent and child attributions in parenting processes and child social-emotional development
No full textGuided by the social information processing model and parenting frameworks, the objective of the present dissertation was to contribute to the literature by contextualizing attribution processes within the dynamic and interdependent nature of parent-child interactions. Across three papers, this dissertation revealed how parental and child attribution patterns influenced parenting processes and child socialemotional development in families of Chinese heritage across developmental stages. The first paper examined how Chinese American parents’ attributions of parenting challenges to causes beyond their control were related to their parenting practices with their preschool-aged children. When Chinese American mothers ascribed parenting failures to sources outside their control, they experienced poor psychological well-being, which led them to be less warm and more psychologically controlling towards their children. Importantly, children’s behavioral difficulties contemporaneously disrupted the link between Chinese American mothers’ attributions and well-being. The second paper investigated the reciprocal associations among Chinese American parents’ attributions of parenting failures to their children, power-assertive parenting, and children’s reactive aggression using a three-wave longitudinal design, each six months apart. Mothers’ attributions to child causes and power-assertive parenting were reciprocally related across three waves. Further, child reactive aggression positively predicted maternal hostile attributions six months later, which in turn, led to more power-assertive parenting one year later. Finally, the third paper explored the age-varying associations between psychologically controlling parenting and mainland Chinese children’s depressive symptoms from childhood to early adolescence and the moderating effects of children’s attributions and gender. Parents increased their use of psychological control from late childhood, putting children at risk of developing depressive symptoms. While children were less likely to make positive interpretations of parents’ psychological control with increasing age, high levels of positive attributions protected children against depressive symptoms during the transition to early adolescence. Further, parents of boys engaged in higher levels of psychological control than parents of girls. Boys and girls were differentially impacted by such parenting across ages. Together, this dissertation emphasized the critical role of parents’ and children’s attributions in shaping parenting processes and children’s adjustment within families of Chinese heritage. Further, this dissertation revealed the dynamic interplay between parenting and children’s characteristics