Network Clustering Research Articles (Page 1)

We present an open-source collection of scripts and programs for the setup, management and evaluation of calculations with the Vienna ab initio simulation package (VASP), called utils4VASP. It contains 14 independent Python scripts and Fortran programs, all with a unified and intuitive handling concept based on command-line arguments. A large repertoire of VASP calculations can be set up with some simple command line calls, including the generation and combination of POSCAR files for bulk and surface slab structures, the respective POTCAR and KPOINTS files and task-specific INCAR files. It further enables the management and evaluation of complex setups not covered by other utility scripts or programs so far, like split-up and parallelized frequency calculations for large structures, or the automated evaluation and visualization of core level energy or Bader partial charge calculations. Emphasis is made on surface-science related calculations, like the targeted placement of adsorbates on substrates or the visualization of scanning-tunneling microscope pictures. Finally, the generation and management of machine-learned interatomic potentials (MLIPs) based on VASP reference data is greatly simplified. Training data collected by on the fly learnings of VASP ML force fields can be effectively selected and combined, or exported into data formats used for Behler-Parrinello neural network or message-passing atomic cluster expansion (MACE) MLIPs. In this publication, all features within utils4VASP are presented concisely, giving both theoretical background and application examples.

Heterogeneous wireless sensor networks (HWSNs), comprising super nodes and normal sensors, offer a promising solution for monitoring diverse environments. However, their deployment is constrained by the limited battery life of sensors. To address this issue, clustering and routing techniques have been employed to conserve energy. Nevertheless, existing approaches often struggle with suboptimal energy distribution and weak network coverage. Additionally, they mostly failed to exploit other energy saving techniques such as sleep scheduling. This paper proposes a novel genetic algorithm (GA)-based approach to optimize sleep scheduling, routing, and clustering in HWSNs. The method comprises two phases, namely join sleep scheduling and tree construction, and clustering of normal nodes. Inspired by the concept of unequal clustering, the HWSN is split into some rings in the first phase, and the number of awake super nodes in each ring keeps the same. This approach addresses the challenges of balancing energy consumption and network lifetime. Furthermore, including network coverage and energy-related criteria in the proposed GA yields long-lasting network operation. Through rigorous simulations, we demonstrate that, on average, our algorithm reduces energy consumption and improves network coverage by 23% and 21.9%, respectively, and extends network lifetime by 501 rounds, compared to the state-of-the-art methods.

Intelligent node identification and dynamic clustering for underwater acoustic sensor networks

Despite demonstrating bistability in spin-crossover (SCO) materials, the absence of long-range magnetic order and poor electrical conductivity limit their prospect in spintronic and nanoelectronic applications. Intending to create hybrid devices made of SCO-2D architecture, here, we report an easily processable Fe-based SCO nanostructures grown on 2D reduced graphene oxide (rGO). X-ray photoelectron spectra of the hybrid clearly reveal the formation of new bonding state with possible charge transfer between rGO and SCO nanoparticles. This interfacial charge transfer enhances intermolecular interactions, resulting in increased cooperativity within the heterostructure. The temperature dependent Mössbauer spectra analysis distinctly uncovers the proportion of Fe (II) spin states within the hybrid nanocomposite samples, highlighting how the formation of a 2D network of SCO clusters enhances the cooperativity. Notably, both the thermal hysteresis and the mean spin-transition temperature are tunable through the application of a magnetic field, underscoring significant magnetic interactions. The inherently low conductivity of pristine SCO nanostructures is addressed by embedding them within a conductive rGO matrix. This facilitates the electrical detection of magnetic bistability through high-spin/low-spin conductance switching, even in the absence of an external magnetic field. As a result, spin functionality is integrated into the conductance behavior, paving the way for hybrid 2D spintronic devices. Finally, ab-inito calculations, on the experimentally motivated model systems provide insights into the microscopic mechanism confirming the enhanced magnetic interaction in the hybrid architecture facilitated by interfacial charge transfer.

Neural network topologies supporting individual variations in vividness of visual imagery.

AUV-Assisted data collection using hybrid clustering and reinforcement learning in underwater acoustic sensor networks

Intelligent clustering and routing protocol for wireless sensor networks using quantum inspired Harris Hawk optimizer and deep reinforcement learning

TGFB2/IGFBP5 activated by transcription factors STAT2 and SMAD3 initiate granulosa cell degeneration and cause follicular atresia in chickens.

Joint computation offloading and resource allocation in clustered MEC-enabled ultra-dense networks with multi-slope channels

Repairing a failed clustered network by external activation

Spanning tree-based dynamic graph routing implements a topology-aware approach for facilitating adaptive load balancing in clustered Wireless Sensor Networks (WSNs). Traditional routing methods often encounter communication constraints and energy inefficiencies resulting from fixed route choices and unequal data distribution across cluster members. The proposed framework establishes a spanning tree structure that dynamically adjusts to changes in topology, energy levels, and node density. The main goal is to enhance route selection while guaranteeing equitable workload allocation across all clusters, thereby prolonging network lifespan and reducing packet latency. The approach combines dynamic edge reconfiguration with localized decision-making, enabling real-time modifications in routing pathways according to residual energy and communication traffic. Each cluster creates an energy-weighted spanning tree that is regularly updated via distributed coordination, improving scalability and fault tolerance. Experimental assessments under diverse network settings reveal enhanced throughput, decreased transmission overhead, and stable energy usage across all nodes, confirming its appropriateness for extensive WSN implementations.

Predictive analytics is redefining how organizations interpret complex market dynamics and execute strategic decisions. The convergence of big data, advanced analytics, and machine learning (ML) enables firms to transform raw market signals into actionable growth strategies. This study explores how predictive analytics contributes to strategic decision-making by integrating machine learning models into corporate foresight and execution systems. Using simulated data and model-based illustrations, the paper demonstrates how predictive algorithms such as regression analysis, clustering, and neural networks can forecast demand, identify emerging market opportunities, and enhance strategic agility. The findings suggest that organizations applying predictive analytics within decision architectures achieve measurable improvements in market responsiveness, forecasting accuracy, and growth realization. The research proposes a framework linking data-driven foresight to strategic execution, offering implications for corporate leaders seeking to operationalize analytics-driven strategy in volatile business environments.

Distributed PV systems offer decentralized renewable generation, while EVs bring about flexible yet unpredictable demand profiles. By coordinating these technologies, we have the potential to decrease carbon emissions, improve grid reliability, and optimize energy usage on a local level. However, impediments such as the intermittent nature of PV generation, the uncertainty of EV charging behaviors, and the limitations of current grid infrastructures continue to pose significant challenges that must be addressed in order to fully realize the potential benefits of these technologies. In this paper, we undertake an in-depth examination of the existing research landscape pertaining to the integration of photovoltaic (PV) systems and electric vehicles (EVs) within local networks. Our analysis spans across technical, operational, environmental, and economic facets. We delve into the technological attributes of distributed PV systems and the distinctive load dynamics exhibited by EV clusters, emphasizing the potential for enhanced synergies through the implementation of smart grid solutions. Key areas of focus include advanced modeling techniques, coordinated scheduling strategies, and optimization frameworks, all aimed at bolstering system performance. Furthermore, we explore the substantial environmental advantages and cost-effectiveness associated with PVEV integration, while simultaneously shedding light on the policy and market hurdles that impede widespread adoption. By amalgamating insights gleaned from recent scholarly endeavors, this paper offers a holistic perspective on PVEV integration and serves as a valuable resource for researchers, policymakers, and stakeholders in the energy industry who aspire to design sustainable local energy systems.

ABSTRACTBackground and AimsEpstein–Barr virus (EBV) has been implicated in autoimmune diseases (AIDs), yet a comprehensive analysis of global research trends, knowledge gaps, and translational opportunities remains lacking. Therefore, we aimed to study the research output of EBV‐associated AIDs globally.MethodsAll publications related to EBV‐associated AIDs from 1993 to 2023 were collected from the Science Citation Index‐Expanded of Web of Science. Subsequently, the data were evaluated using the bibliometric methodology. Bibliometrix package in R software was used for data retrieval. VOSviewer and CiteSpace were used to visualize the research focus and trend regarding the effect of EBV‐associated AIDs.ResultsWe analyzed 1589 publications to explore the global scientific landscape on EBV‐associated AIDs. Growth in publications exhibited two peaks, with post‐2020 acceleration coinciding with increased interest in EBV's immunological role. The USA exhibited the highest publications with 543 publications, many of which investigated molecular pathways such as lipid metabolism in EBV‐associated AIDs. Then, Italy (n = 161) and Japan (n = 140) took the second and third places, respectively. Among the institutions involved, Tel Aviv University provided the biggest nodes in each cluster of the cooperation network. The most frequently cited author in the field, according to our results, was Shoenfeld Y. Finally, the results of keyword co‐occurrence analysis showed that systemic lupus erythematosus and rheumatoid arthritis are the most extensively investigated topics in this study area.ConclusionThis study highlights pivotal milestones in EBV‐AIDs research and proposes future directions, including genetic–host immune system interaction, prevention trials, and collaborative mechanisms. Prioritizing these emerging hotspots could advance therapeutic strategies and interdisciplinary synergies.

Metabolic bariatric surgery has become a central strategy for obesity management, showing potential to improve both physical and mental health outcomes. This study aimed to analyze publication trends in metabolic bariatric surgery and mental health, identify leading countries, institutions, authors, and collaborations, and reveal key hotspots and future research directions. A bibliometric analysis of publications spanning 1998 to 2024 was performed. Key bibliometric indicators including publication, citation counts, and collaborative networks. Network structures and cluster clarity were measured using modularity Q and weighted mean silhouette. Data visualization tools mapped emerging research themes and highlight leading contributors. Our study revealed a steady increase in publications and citations related to metabolic bariatric surgery and mental health from 1998 to 2024. The USA led in document count, citations, and international collaboration, with prominent institutions like the University of Pittsburgh contributing significantly to the field. The top journals were Obesity Surgery, and Surgery for Obesity and Related Diseases, while the most cited articles were by Adams TD (2007) and Schauer PR (2017). Keyword trends shifted from surgical techniques to the psychological impacts of metabolic bariatric surgery, with terms like morbid obesity, gastric bypass, and mental health remaining central, while newer areas like cardiovascular risks and long-term outcomes emerged as new hotspots. Bibliometric trends demonstrate growth but underscore challenges in global research access. Future research should address methodological issues and healthcare disparities, with a focus on standardizing training, clinical trials, and outcomes. Promising study areas include eating and alcohol-related disorders, cardiovascular, and long-term psychological health.

ABSTRACT Ultra-high-strength steel (UHSS) has extremely strict requirements on mechanical properties. Frequent process transitions in the continuous annealing process (CAP) significantly impact quality stability, and effective process evaluation techniques are currently lacking. In this study, a deep clustering network with attention (ADCN) based on multidimensional time series is proposed. Deep learning networks such as Gate Recurrent Unit and attention mechanisms are introduced to mine the temporal curve features of transition states and the coupling relationship between multi-dimensional process parameters, while the matrix traces is used to improve the optimisation objectives in traditional clustering tasks. The proposed method innovatively provides a comprehensive evaluation result, including modal categories, membership degrees, and significant transition regions. The proposed ADCN achieves an average 19.1% improvement in clustering accuracy compared to existing methods and ablation structures. Its effectiveness is further verified by transition state evaluation results and typical case analysis. The proposed evaluation method can be further applied to process optimisation and production scheduling, which is of great significance to improve the quality stability of UHSS.

ABSTRACT Western homicide typologies (e.g., instrumental–expressive) poorly capture weapon restrictions and rural–urban divides in non-firearm societies like China. This study examined 254 adjudicated single-victim homicide cases (2013–2022) using PAM clustering and similarity network analysis to uncover latent behavioural patterns. Three distinct clusters emerged: Sharp-Weapon/Over-kill (knife-dominated, 79% over-kill), Strangulation/Control (high coercive control, low over-kill), and Mixed Method (heterogeneous tactics). These diverge from Western dichotomies, reflecting China’s limited firearm access (61% bladed tools vs. < 3% firearms) and rural–urban disparities. Logistic regression showed that female victims were six times more likely to experience over-kill (OR = 6.05), while sharp-weapon use reduced its odds (OR = 0.31). Findings reveal the interplay between method rationality (weapon efficiency) and emotional context (gendered violence) and underscore structural factors such as rural socioeconomic precarity. The study advocates culturally grounded criminological frameworks integrating offender profiles with crime scene dynamics for improved violence prevention and investigation.

Federated graph-level clustering network with adaptive knowledge compensation.

Nonsyndromic cleft lip and palate (NSCLP) is a common craniofacial malformation increasingly recognized to involve neurodevelopmental abnormalities, though the molecular basis remains unclear. Here, single-nucleus RNA sequencing of the superior temporal plane from mid-gestation NSCLP fetuses is performed, and profound alterations in cell-type composition, intercellular communication, and transcriptional programs are uncovered. Integrative analyses with weighted gene co-expression network analysisand single-cell regulatory network interference and clustering based on single-nucleus transcriptomes identify myocyte enhancer factor 2C (MEF2C) as a shared transcriptional regulator consistently downregulated in excitatory and inhibitory neurons across mid-term gestation, which is validated in NSCLP fetal brain tissues. MEF2C expression is negatively correlated with synaptophysinimmunofluorescence intensity. In MEF2C-deficient primary cortical neurons, impaired synaptic formation, reduced postsynaptic density protein-95expression, and weakened excitatory postsynaptic transmission without altering intrinsic excitability are found. Upstream regulators of MEF2C are enriched for pathways controlling neuronal differentiation, synaptic plasticity, and epigenetic regulation, suggesting broad disruption of neurodevelopmental programs. Together, this study provides molecular evidence of disrupted brain development in NSCLP and implicates MEF2C as a potential mediator of neurodevelopmental impairments.

An overarching objective in contemporary statistical network analysis is extracting salient information from datasets consisting of multiple networks. To date, considerable attention has been devoted to node and network clustering, while comparatively less attention has been devoted to downstream connectivity estimation and parsimonious embedding dimension selection. Given a sample of potentially heterogeneous networks, this article proposes a method to simultaneously estimate a latent matrix of connectivity probabilities and its embedding dimensionality or rank after first pre-estimating the number of communities and the node community memberships. The method is formulated as a convex optimization problem and solved using an alternating direction method of multipliers algorithm. We establish estimation error bounds under the Frobenius norm and nuclear norm for settings in which observable networks have blockmodel structure, even when node memberships are imperfectly recovered. When perfect membership recovery is possible and dimensionality is much smaller than the number of communities, the proposed method outperforms conventional averaging-based methods for estimating connectivity and dimensionality. Numerical studies empirically demonstrate the accuracy of our method across various scenarios. Additionally, analysis of a primate brain dataset demonstrates that posited connectivity is not necessarily full rank in practice, illustrating the need for flexible methodology.