MD-SOAR Maryland Shared Open Access Repository
Not a member yet
34521 research outputs found
Sort by
White Dwarfs as Dark Matter Collectors: A Study of Elemental Capture Rates
No full textThis paper investigates the capture of dark matter in white dwarfs, focusing on the interaction between dark matter particles and the ions within the white dwarf’s dense core. The study models the dark matter capture rate using an optically-thin approximation, where each dark matter particle undergoes a single scattering event before either being captured or escaping the white dwarf’s gravitational influence. Key assumptions in the model include a zero core temperature for the white dwarf, a uniform core composition, and the exclusion of multi-scattering events. The paper examines the equations governing the interaction between dark matter and ions within the white dwarf’s core. The results of the calculations indicate that an increase in the mass of the white dwarf, as well as a higher concentration of heavier elements in its core, enhances the dark matter capture rate. The work also highlights potential future refinements, aiming to provide an introduction to the ongoing research to model dark matter interactions in stellar remnants as a potential detection method
Deep Reinforcement Learning-Based Computation Computational Offloading for Space–Air–Ground Integrated Vehicle Networks
No full textIn remote or disaster areas, where terrestrial networks are difficult to cover and Terrestrial Edge Computing (TEC) infrastructures are unavailable, solving the computation computational offloading for Internet of Vehicles (IoV) scenarios is challenging. Current terrestrial networks have high data rates, great connectivity, and low delay, but global coverage is limited. Space–Air–Ground Integrated Networks (SAGIN) can improve the coverage limitations of terrestrial networks and enhance disaster resistance. However, the rising complexity and heterogeneity of networks make it difficult to find a robust and intelligent computational offload strategy. Therefore, joint scheduling of space, air, and ground resources is needed to meet the growing demand for services. In light of this, we propose an integrated network framework for Space-Air Auxiliary Vehicle Computation (SA-AVC) and build a system model to support various IoV services in remote areas. Our model aims to maximize delay and fair utility and increase the utilization of satellites and Autonomous aerial vehicles (AAVs). To this end, we propose a Deep Reinforcement Learning algorithm to achieve real-time computational computational offloading decisions. We utilize the Rank-based Prioritization method in Prioritized Experience Replay (PER) to optimize our algorithm. We designed simulation experiments for validation and the results show that our proposed algorithm reduces the average system delay by 17.84%, 58.09%, and 58.32%, and the average variance of the task completion delay will be reduced by 29.41%, 48.74%, and 49.58% compared to the Deep Q Network (DQN), Q-learning and RandomChoose algorithms.This work was supported in part by the National Natural Science Foundation of China under Grant 62072360 and Grant 62172438 in part by the Key Research and Development Plan of Shaanxi Province under Grant 2021ZDLGY02 09 Grant 2023 GHZD 44 and Grant 2023 ZDLGY 54 in part by the National Key Laboratory Foundation under Grant 2023 JCJQ LB 007 in part by the Natural Science Foundation of Guangdong Province of China under Grant 2022A1515010988 in part by the Key Project on Artificial Intelligence of Xi an Science and Technology Plan under Grant 23ZDCYJSGG0021 2022 Grant 23ZDCYYYCJ0008 and Grant 23ZDCYJSGG0002 2023 and in part by the Proof of Concept Fund from Hangzhou Research Institute of Xidian University under Grant GNYZ2023QC0201 Grant GNYZ2024QC004 and Grant GNYZ2024QC015https://ieeexplore.ieee.org/abstract/document/1094763
xIDS-EnsembleGuard: An Explainable Ensemble Learning-based Intrusion Detection System
No full text23rd IEEE International Conference on Trust, Security and Privacy in Computing and Communications (TrustCom-2024)In this paper, we focus on addressing the challenges of detecting malicious attacks in networks by designing an advanced Explainable Intrusion Detection System (xIDS). The existing machine learning and deep learning approaches have invisible limitations, such as potential biases in predictions, a lack of interpretability, and the risk of overfitting to training data. These issues can create doubt about their usefulness, transparency, and a decrease in trust among stakeholders. To overcome these challenges, we propose an ensemble learning technique called "EnsembleGuard." This approach uses the predicted outputs of multiple models, including tree-based methods (LightGBM, GBM, Bagging, XGBoost, CatBoost) and deep learning models such as LSTM (long short-term memory) and GRU (gated recurrent unit), to maintain a balance and achieve trustworthy results. Our work is unique because it combines both tree-based and deep learning models to design an interpretable and explainable meta-model through model distillation. By considering the predictions of all individual models, our meta-model effectively addresses key challenges and ensures both explainable and reliable results. We evaluate our model using well-known datasets, including UNSW-NB15, NSL-KDD, and CIC-IDS-2017, to assess its reliability against various types of attacks. During analysis, we found that our model outperforms both tree-based models and other comparative approaches in different attack scenarios.This work was supported in part by the Guangdong Provincial Key Laboratory of Human Digital Twin (Grant 2022B1212010004), Guangzhou Basic Research Program (Grant SL2023A04J00930), and the Shenzhen Holdfound Foundation Endowed Professorshiphttp://arxiv.org/abs/2503.0061
From Guessing to Asking: An Approach to Resolving the Persona Knowledge Gap in LLMs during Multi-Turn Conversations
No full text2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics, Albuquerque, New Mexico, April 29–May 4, 2025In multi-turn dialogues, large language models (LLM) face a critical challenge of ensuring coherence while adapting to user-specific information. This study introduces the persona knowledge gap, the discrepancy between a model's internal understanding and the knowledge required for coherent, personalized conversations. While prior research has recognized these gaps, computational methods for their identification and resolution remain underexplored. We propose Conversation Preference Elicitation and Recommendation (CPER), a novel framework that dynamically detects and resolves persona knowledge gaps using intrinsic uncertainty quantification and feedback-driven refinement. CPER consists of three key modules: a Contextual Understanding Module for preference extraction, a Dynamic Feedback Module for measuring uncertainty and refining persona alignment, and a Persona-Driven Response Generation module for adapting responses based on accumulated user context. We evaluate CPER on two real-world datasets: CCPE-M for preferential movie recommendations and ESConv for mental health support. Using A/B testing, human evaluators preferred CPER's responses 42% more often than baseline models in CCPE-M and 27% more often in ESConv. A qualitative human evaluation confirms that CPER's responses are preferred for maintaining contextual relevance and coherence, particularly in longer (12+ turn) conversations.https://arxiv.org/abs/2503.1255
Long-Tailed Federated Learning in Internet of Medical Things Based on Ensemble Distillation and Imbalanced Calibration
No full textThe Internet of Medical Things (IoMT) has a promising future, as its devices can monitor vital signs, offer treatment guidance, and perform real-time diagnostics using AI and wireless communication technologies. However, due to the difficulty of collecting patient data on a large scale and potential privacy risks, traditional centralized machine learning methods are often challenging to apply in IoMT devices. Federated learning, as a privacy-preserving technology, aims to build high-quality deep learning models across distributed clients while protecting data privacy. However, current popular federated learning methods exhibit suboptimal performance when dealing with non-IIDness data, especially in the case of long-tail class distributions, leading to unsatisfactory results. Additionally, due to privacy constraints on distributed clients, these methods cannot leverage traditional deep learning techniques to handle long-tail data, which is often characterized by long-tail heterogeneous distributions in IoMT. To address these challenges, this paper proposes a solution of Privacy-preserving Computing Client Scoring and Knowledge Distillation (FedLT+SKD). The method uses privacy protection computation to provide prior knowledge of global data class distribution while ensuring data privacy. Based on this prior knowledge, it employs a points-based sampling strategy to identify clients that perform well on long tail data and uploads their local model to the server. On the server side, the robustness of the global model is enhanced by collection distillation and imbalance correction. We verify the effectiveness of this method on the medical datasets ISIC, ChestX-ray14, MRI and also on the traditional datasets CIFAR-10-LT and CIFAR-100-LT, and the experimental results show that the method is superior to the popular federation and long-tail learning methods.This work was supported in part by National Natural Science Foundation of China under Grant 62102264, Taishan Scholar Project under Grant tsqnz20230602, Natural Science Foundation of Shandong Province under Grant ZR2024MF115 and ZR2023LZH010, and Youth Innovation University Team Project in Shandong under Grant 2022KJ062.https://ieeexplore.ieee.org/abstract/document/10869336
I Hate the News Jan 28
No full textThe weekly news analysis from I Hate Politics: The regional power transmission organization covering New Jersey to Illinois goes to Annapolis and finds upset legislators. Montgomery County Department of Permitting Services can’t seem to spend the money it has to improve service. A state program to give renters on the verge of eviction legal assistance seems to have worked. City of Rockville eases height restrictions and eliminates some off-street parking requirements in the Town Center. Stay with us. Newly in public domain music from the 1920s: The Benson Orchestra of Chicago, the Paul Whiteman band, Carl Fenton, and Jan Garber.https://open.spotify.com/episode/1l25FWCcGC4ajxWwNGuQe
A Utility-Aware and Holistic Approach for Privacy Preserving Distributed Mining with Worst Case Privacy Guarantee
No full textOrganizations often want to predict some attribute values collaboratively. However, they are often unwilling or not allowed to directly share their private data. Thus there is great need for distributed privacy preserving techniques. There exists a rich body of work based on Secure MultiParty Computation techniques. However, most such techniques are tied to a specific mining algorithm and users have to run a different protocol for each mining algorithm. A holistic approach was proposed in which all parties first use a SMC protocol to generate a synthetic data set and then share this data for different mining algorithms. However, this approach has two major drawbacks: 1) it provides no worst case privacy guarantee, 2) parties involved in the mining process often know what attribute to predict, but the holistic approach does not take this into account. In this paper, we propose a method that addresses these shortcomings. Experimental results demonstrate the benefits of the proposed solution.https://userpages.umbc.edu/~zhchen/papers/chen-skm-privacy.pd
Improvisational Storytelling Agents
No full text31st Conference on Neural Information Processing Systems (NIPS 2017), Long Beach, CA, USA.The problem of improvisational story generation involves one or more agents collaborating in order to create a story without any advance notice of topic. We present a pipeline for an artificial agent that is capable of improvisational storytelling while collaborating with a human agent. Starting with story corpora, we “eventify” sentences, which creates a simplified and abstracted representation. The rest of the pipeline–the agent’s response–is broken into three parts: generating successive events (event-to-event), translating of events back into natural language (event-to-sentence), and plugging the specifics of the story back into the generated sentences (slot filling). We discuss techniques for each of these sub-problems.This work was supported by the Defense Advanced Research Projects Agency (DARPA) under
Contract No. W911NF-15-C-0246.https://laramartin.net/pub/NeurIPS-Creativity-17-improvisational-storytelling-agents.pd
Distributed urban forest patch sampling detects edge effects and woodland condition for monitoring and management
No full textUrban forest patches, including woodland interiors and bounding edge habitat, result from secondary succession and fragmentation of more extensive forested landscapes in the eastern United States. Management regimes, surrounding land use, and successional processes lead to distinct environments and contribute to local and regional heterogeneity. However, many woodlands are degraded due to frequent disturbance, aggressive exotic species, and heavy browsing, which stress canopies, reduce regeneration, and may reduce ecosystem services. Effective management requires rapid, repeatable assessment of forest composition, structure, and condition at the scale of local decision-making. We present and apply a protocol for characterizing urban woodlands that generates new insight into the status of urban woodlands and baseline data for change detection over time. Samples of overstory composition, ground cover, surface soil measurements, and the Schumacher Vine Encroachment Index were collected at 845 points across each of 47 patches across Baltimore, Maryland. Simple citywide summaries allowed characterization of Baltimore's urban overstories as overwhelmingly native, though dominated by a range of successional conditions. By contrast, we found that ground layers were predominantly exotic, with abundant invasives or ruderal native species benefiting from disturbed conditions. Seven overstory types were distinguished, the majority under threat from aggressive vines. Most soils showed little evidence of compaction, but variable organic content. Distributed data allowed cross-patch comparison as well as within-patch analyses along edge-to-interior gradients. Species diversity, nativity, and overstory basal area all increased toward woodland interiors, whereas soil compaction and vine encroachment decreased. Structural and compositional shifts in both overstory and ground layer species revealed indicators of edge (15.2–18.7 m) and interior (>41.5 m) conditions, as well as evidence of transitional zones with distinct patterns of biodiversity. Despite high levels of fragmentation and disturbance that challenge municipal land managers operating with limited resources, rapid, low-cost sampling enabled comparison across multiple scales, encouraging repeated sampling and adaptive response to changing forest conditions. Qualitative and quantitative analysis as well as specific examples illustrated the generic utility of the protocol for a range of applications and its ability to produce new insight enabling management action and informed conservation planning.Funding information UMBC Baltimore Office of Sustainability USDA Forest Service Northern ResearchStation Abell Foundation National Fishand Wildlife Foundation Ms RaniFischerhttps://onlinelibrary.wiley.com/doi/abs/10.1002/ecs2.7023
Bio-thermal medical devices, methods, and models: new developments and advances
No full textRecent advancements in medical imaging techniques have greatly enhanced the ability to capture anatomically precise and highly detailed vascular structures within biological tissues Singh 2024. This progress is particularly significant for bioheat transfer modeling, where an accurate representation of the vascular network is essential for understanding heat exchange, blood perfusion dynamics, and thermal responses in both healthy and pathological conditions. The integration of high-resolution, three-dimensional geometries extracted from medical imaging data, often at the voxel level, enables more precise simulations, improving the predictive accuracy of thermal treatments and physiological responses Singh 2024. More recently, research efforts have continued to develop anatomically accurate models from medical imaging and develop physics and physiology-based models Singh et al., 2024. On contrary, voxel-based domains generated from medical image are crucial for bioheat transfer modeling; however, a key challenge lies in voxel resolution limitations. Due to the small dimensional scale of blood vessels, not all vessels are captured within a given voxel resolution, resulting in discontinuities in vascular segmentation. Also, pre-capillary vessels such as arterioles, which play a critical role in regulating blood flow resistance, are often modeled within the tissue as a porous domain. Such simplification leads to a loss of critical vascular information, potentially affecting the accuracy of bioheat transfer simulations. Additionally, magnetic particle imaging (MPI) has emerged as a powerful tool for tracking magnetic nanoparticles used in hyperthermia-based cancer treatments. By combining mathematical modeling with MPI, researchers are optimizing nanoparticle induced hyperthermia to improve therapeutic outcomes while minimizing unintended thermal damage to surrounding healthy tissues Singh 2020; Singh et al., 2021; Singh 2023. In this Research Topic, Pawar et al. conducted a sensitivity analysis to assess the impact of the spatial distribution of magnetic iron oxide nanoparticles (MIONs) on tumor temperature. Their study utilized co-registered magnetic resonance (MR)/computed tomography (CT) imaging alongside magnetic particle imaging (MPI) to derive in vivo MION distribution, which was then compared to mathematically generated uniform and Gaussian distributions. Theoretical predictions were based on the Pennes bioheat transfer equation, incorporating the dynamic influence of temperature on blood perfusion. To enhance accuracy, they employed a piecewise function to model the degree of vascular stasis (collapse of vasculature), as previously quantified by Singh 2022 in the context of magnetic hyperthermia. This approach provided valuable insights into optimizing MION distribution for more effective magnetic hyperthermia treatments. In another article of this Research Topic, Amare et al. highlighted the challenges involved in extracting the small blood vessels due to limited resolution of voxels obtained from image data. Their approach clearly provides evidence that mathematical representations of unsegmented blood vessels can approximate the thermal resistance and reduced the need for high-resolution imaging. In addition, their proposed methodology provides a computationally efficient alternative to high-resolution imaging, making it a valuable tool for future applications in biomedical modeling and thermal therapy planning. Besides the above numerical work, Pioletti presented an intriguing and innovative perspective on the role of self-heating in soft tissues, specifically in cartilage, because of mechanical stimulation induced heat effect. The core idea discussed in this work is that temperature changes induced by mechanical activity might be necessary for cartilage maintenance-introduces a potential paradigm shift in how we think about the physiological effects of mechanical loading on musculoskeletal tissues. In addition to the perspective article, Li et al. conducted a bibliometric analysis to assess studies on hypothermia-related injuries, treatment strategies, and underlying mechanisms. This study provides a comprehensive summary of hypothermia's impact on human health and the therapeutic applications of moderate hypothermia. By mapping research trends, frontiers, and key focus areas, the analysis offers valuable insights into the current landscape and future directions of hypothermia research. Additionally, it highlights the distinctions and interconnections between therapeutic and severe hypothermia, offering a clearer understanding of advancements and emerging trends in the field. This Research Topic presents a collection of two research articles, a perspective paper, and a review paper, each showcasing novel discoveries, state-of-the-art advancements, and future directions in the interdisciplinary field of computational modeling in biomedical engineering. These studies emphasize multiscale, multiphysics, and medical imaging-assisted approaches, highlighting their integration and applications. We believe that the insights shared in this collection will pave the way for groundbreaking research in bioheat transfer, accelerating innovations in medical device development.https://www.frontiersin.org/journals/thermal-engineering/articles/10.3389/fther.2025.1591428/ful