Dynamic Clustering Research Articles

Dynamics of paramagnetic permanent chains and self-assembled clusters under a rapidly rotating magnetic field.

We study experimentally and theoretically the dynamics of permanent paramagnetic chains and mixed clusters formed by permanent paramagnetic chains and paramagnetic particles under the influence of a time-varying magnetic field. First, we examine the dynamics of permanent chains at high frequencies (∼50 to 1000Hz). These permanent chains exhibit continuous rotational motion with a frequency several orders of magnitude lower than that of the magnetic field. We develop a theoretical model that accurately describes the dependence of the rotational dynamics of chains on their length, as well as the amplitude and frequency of the external magnetic field in this high frequency regime. Next, we examine how cluster dynamics are affected by the presence of permanent chains. We show that the rotation of clusters composed of a high proportion of permanent chains is slowed down but remains qualitatively well described by the theoretical model we developed for homogeneous clusters of isotropic particles. We propose that the decrease in angular velocity for mixed clusters is due to the hardening of the cluster's 2D elastic modulus caused by the increase of the steric interaction parameter stemming from the presence of chemical links between particles in the chains.

STAD: Ship trajectory anomaly detection in ocean with dynamic pattern clustering

With the continuous increase of maritime traffic, developing an efficient and accurate model for ship trajectory anomaly detection has become crucial for ensuring maritime transportation safety. The high complexity and variability of the marine environment lead to diverse ship trajectory patterns, making it challenging to learn effective trajectory representations for accurately identifying anomalies. We thus proposed an unsupervised deep learning model called STAD for ship trajectory anomaly detection in ocean to address this challenge. Concretely, STAD leverages offset reconstruction-based representation learning and a deep Gaussian Mixture Model (GMM) estimation network to learn the underlying complex clustering patterns of ship trajectories and utilize the learned patterns to enhance trajectory anomaly detection. Extensive experiments on multiple AIS datasets indicate that our model significantly outperforms existing methods in detecting multiple representative types of ship trajectory anomalies, including shift deviation, abnormal heading, and abnormal speeding. This study could help closely monitor the status of ship movement and detect abnormal behaviors in advance, thus benefiting maritime safety.

Dynamic cluster field modeling of collective chemotaxis

Collective migration of eukaryotic cells is often guided by chemotaxis, and is critical in several biological processes, such as cancer metastasis, wound healing, and embryogenesis. Understanding collective chemotaxis has challenged experimental, theoretical and computational scientists because cells can sense very small chemoattractant gradients that are tightly controlled by cell-cell interactions and the regulation of the chemoattractant distribution by the cells. Computational models of collective cell migration that offer a high-fidelity resolution of the cell motion and chemoattractant dynamics in the extracellular space have been limited to a small number of cells. Here, we present Dynamic cluster field modeling (DCF), a novel computational method that enables simulations of collective chemotaxis of cellular systems with O(1000)\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$${\\mathcal {O}}(1000)$$\\end{document} cells and high-resolution transport dynamics of the chemoattractant in the time-evolving extracellular space. We illustrate the efficiency and predictive capabilities of our approach by comparing our numerical simulations with experiments in multiple scenarios that involve chemoattractant secretion and uptake by the migrating cells, cell migration in confined spaces, regulation of the attractant distribution by cell motion, and interactions of the chemoattractant with an enzyme. The proposed algorithm opens new opportunities to address outstanding problems that involve collective cell migration in the central nervous system, immune response and cancer metastasis.

Transport in cellular aggregates described by fluctuating hydrodynamics

Biological functionality of cellular aggregates is largely influenced by the activity and displacements of individual constituent cells. From a theoretical perspective this activity can be characterized by hydrodynamic transport coefficients of diffusivity and conductivity. Motivated by the clustering dynamics of bacterial microcolonies we propose a model of active multicellular aggregates and use recently developed macroscopic fluctuation theory to derive a fluctuating hydrodynamics for this model system. Both semianalytic theory and microscopic simulations show that the hydrodynamic transport coefficients are affected by nonequilibrium microscopic parameters and significantly decrease inside of the clusters. We further find that the Einstein relation connecting the transport coefficients and fluctuations breaks down in the parameter regime where the detailed balance is not satisfied. This study offers valuable tools for experimental investigation of hydrodynamic transport in other systems of cellular aggregates such as tumor spheroids and organoids. Published by the American Physical Society 2024

Tscluster: A python package for the optimal temporal clustering framework

Temporal clustering extends the conventional task of data clustering by grouping time series data according to shared temporal trends across sociospatial units, with diverse applications in the social sciences, especially urban science. The two dominant methods are as follows: Time Series Clustering (TSC), with dynamic cluster centres but static labels for each entity, and Sequence Label Analysis (SLA), with static cluster centres but dynamic labels. To implement the universe of models spanning the design space between TSC and SLA, we present tscluster, an open-source Python framework. tscluster offers: (1) several innovative techniques, such as Bounded Dynamic Clustering (BDC), that are not available in existing libraries, allowing users to set an upper bound on the number of label changes and identify the most dynamically evolving time series; (2) a user-friendly interface for applying and comparing these methods; (3) globally optimal solutions for the clustering objective by employing a mixed-integer linear programming formulation, enhancing the reproducibility and robustness of the results in contrast to existing methods based on initialization-sensitive local optimization; and (4) a suite of visualization tools for interpretability and comparison of clustering results. We present our framework using a case study of neighbourhood change in Toronto, comparing two methods available in tscluster. Supplemental materials provide an additional case study of local business development in Chicago and a detailed mathematical exposition of our framework. tscluster can be installed via PyPI (pypi.org/project/tscluster), and the source code is accessible on Github (github.com/tscluster-project/tscluster). Documentation is available online at the tscluster website (tscluster.readthedocs.io).

Comparison of Matrix Product State and Multiconfiguration Time-Dependent Hartree Methods for Nonadiabatic Dynamics of Exciton Dissociation.

The excited-state dynamics of organic molecules, molecular aggregates, and donor-acceptor clusters is typically governed by the interplay of electronic excitations and, due to their flexibility and soft bonding, by the interaction with their vibrations. This interaction in these systems can be characterized by a few relevant electronic states that are coupled to numerous vibrational normal modes, encompassing a vast configurational space of the molecules. The full quantum simulation of these type of systems has been long dominated by the multiconfiguration time-dependent Hartree (MCTDH) approach and its multilayer variants, which are considered the gold standard in the presence of electron-vibration coupling with a large number of modes. Recently, also the matrix product state ansatz (MPS) with appropriate time-evolution schemes has been applied to these types of Hamiltonians. In this article, we provide a numerical comparison of excited-state dynamics between the MCTDH and MPS approaches for two electron-vibration coupled systems. Notably, we consider two models for exciton dissociation at a P3HT:PCBM heterojunction, comprising two electronic states and 100 vibrational modes, and 26 electronic states and 113 vibrational modes, respectively. While both methods agree very well for the first model, more pronounced deviations are found for the second model. We trace back the divergence between the methods to the different way entanglement is treated.

Changes in Language Learners’ Affect: A Complex Dynamic Systems Theory Perspective

AbstractThis study investigated changes in motivation, self‐efficacy beliefs, and a range of emotions, including enjoyment, hope, pride, curiosity, anxiety, boredom, apathy, confusion, and shame, from a complex dynamic systems theory (CDST) perspective over a 2‐year period in the Hungarian English as a Foreign Language (EFL) context. Using the same questionnaire, we collected data four times throughout 4 semesters from 101 participants studying English in two Hungarian high schools. For data analysis, we used latent growth curve modeling (LGCM) to detect the group‐level changes in learners’ motivation, self‐efficacy, and emotions. We also employed dynamic cluster analysis to identify trends in learners’ trajectories regarding these variables. In our panel data, linear models described the data well concerning the ought‐to second language (L2) self, language learning experience, boredom, apathy, and confusion, and for enjoyment, curiosity, anxiety, and shame, nonlinear models had the best fit. We could also identify trajectories depicting attractor states and learner paths that featured influences of perturbations.

Dynamical clustering and wetting phenomena in inertial active matter

Dynamical clustering is a key feature of active matter systems composed of self-propelled agents that convert environmental energy into mechanical motion. At the micron scale, where overdamped dynamics dominate, particles with opposite motility can obstruct each other’s movement, leading to transient dynamical arrest. This arrest can promote cluster formation and motility-induced phase separation. However, in macroscopic agents, where inertia plays a significant role, clustering is heavily influenced by bounce-back effects during collisions, which can impede cluster growth. Here we present an experiment based on active granular particles, in which inertia can be systematically tuned by changing the shaker frequency. As a result, a set of phenomena driven and controlled by inertia emerges. Before the suppression of clustering, inertia induces a transition in the cluster’s inner structure. For small inertia, clusters are characterized by the crystalline order typical of overdamped particles, while for large inertia clusters with liquid-like order are observed. In addition, in contrast to microswimmers, where active particles wet the boundary by primarily forming clusters attached to the container walls, in an underdamped inertial active system, walls do not favor cluster formation and effectively annihilate motility-induced wetting phenomena. As a consequence, inertia suppresses cluster nucleation at the system boundaries.

Dynamic clustering based risk aware congestion control technique for vehicular network

In order to improve road safety and offer extra services to cars and their users, vehicular ad hoc networks, or VANETs, are essential parts of intelligent transportation systems (ITS). The primary aim of this research is to suggest and assess a new approach for mitigating network congestion in VANET technology through the implementation of dynamic grouping of vehicles for safety (DGVS). To do this, virtual regions are created around cars using the DBSCAN and K Mean method. This allows vehicles to communicate directly with others within the same DGVS, eliminating the need to broadcast messages across the entire network.The ultimate goal is to drastically decrease traffic while preserving vital data transfer for VANET traffic control and safety. The suggested technique determines the optimal transmission rate in accordance with the existing channel circumstances, yielding a balanced performance concerning both packet delivery and channel congestion. With this novel method, cars may only talk to other vehicles that are in the same DGVS, thus there’s no need to broadcast messages to the whole network. With the use of this technique, the efficacy of DGVS in reducing VANET congestion and enhancing network performance will be thoroughly assessed. The study’s conclusions demonstrate how well the suggested dynamic grouping of vehicles for safety (DGVS) technique works to reduce VANET congestion. It was shown through simulation-based studies that, in comparison to current approaches, DGVS dramatically decrease network congestion, resulting in notable gains in network performance and decreased communication delay. Additionally, the study found a number of possible uses for DGVS in the transportation industry, such as emergency response, traffic management, and accident prevention.

A study of the coupled dynamics of asymmetric absorbing clusters in a photophoretic trap

Abstract We report a study on the dynamics of absorbing asymmetric carbon clusters trapped by a loosely focused Gaussian beam using photophoretic force. At high laser powers, all the trapped clusters display rotation coupled with oscillation along the axial direction, with a majority spinning about a body fixed axis, while the rest display dual spin as well as orbital motion about a fixed point in space. The spinning and orbiting frequency is inversely proportional to the amplitude of the axial oscillation - with one growing at the expense of the other. Further, the frequencies of these rotations are not proportional to the laser power, but to the trap stiffnesses inferred from the corresponding natural frequencies. The clusters also stop rotating below a certain laser power, and execute random thermal fluctuations. Our work suggests that the dynamics of clusters trapped with photophoretic force are largely dependent on the cluster size and morphology, which could, in principle, be tuned to obtain various motional responses, and help in the design of rotating micromachines in air.

Groundwater dynamics clustering and prediction based on grey relational analysis and LSTM model: A case study in Beijing Plain, China

Study regionBeijing Plain, China Study focusThe traditional zoning of groundwater systems relies on lithological characteristics and hydraulic connections. However, the influence of climate change and human activities has led to the emergence of diverse dynamic patterns within these systems. Additionally, the differences in groundwater dynamic characteristics are often ignored when modeling with machine learning methods. In this study, a new clustering method named GRA-CLU was proposed to classify dynamic types of groundwater. Then, regional models are built using LSTM based on the dynamic zoning, aiming to incorporate hydrogeological significance into the pure data-driven model.New hydrological insights for the region: The GRA-CLU method classified the study area into six groundwater dynamic types. Compared with the traditional hydrogeological units, the new zoning result identified two dynamic types that have never been considered: urban influence zone and mountain-plain junction zone where surface water interacts with groundwater. Through analysis, the significant rise of groundwater levels in 2021 was more influenced by heavy rainfall events rather than human activities. This study further explored the performance of LSTM regional models based on the six dynamic zones. The results indicated that the NSE of models increased by 5.7–50.0 %, the improvement was more obvious in zones with irregular fluctuations. The RMSE decreased by 31.5–59.8 %, particularly noticeable in zones with regular annual fluctuations and large samples.

Theia 456: Tidally Shredding an Open Cluster

The application of clustering algorithms to the Gaia astrometric catalog has revolutionized our census of stellar populations in the milky Way, including the discovery of many new dispersed structures. We focus on one such structure, Theia 456 (COIN-Gaia-13), a loosely bound collection of ∼320 stars spanning ∼120 pc that has previously been shown to exhibit kinematic, chemical, and gyrochronal coherency, indicating a common origin. We obtain follow-up radial velocities and supplement these with Gaia astrometry to perform an in-depth dynamical analysis of Theia 456. By integrating stellar orbits through a Milky Way potential, we find the currently dispersed structure coalesced into a small cluster in the past. Via Bayesian modeling, we derive a kinematic age of 245 ± 3 Myr (statistical), a half-mass–radius of 9 ± 2 pc, and an initial one-dimensional velocity dispersion of 0.14 ± 0.02 km s−1. Our results are entirely independent of model isochrones, details of stellar evolution, and internal cluster dynamics, and the statistical precision in our age derivation rivals that of the most precise age-dating techniques known today, though our imperfect knowledge of the Milky Way potential and simple spherical model for Theia 456 at birth add additional uncertainties. Using posterior predictive checking, we confirm these results are robust under reasonable variations to the Milky Way potential. Such low-density structures that are disrupted by the Galactic tides before virializing may be ubiquitous, signifying that Theia 456 is a valuable benchmark for studying the dynamical history of stellar populations in the Milky Way.

LRRTM2 controls presynapse nano-organization and AMPA receptor sub-positioning through Neurexin-binding interface

Synapses are organized into nanocolumns that control synaptic transmission efficacy through precise alignment of postsynaptic neurotransmitter receptors and presynaptic release sites. Recent evidence show that Leucine-Rich Repeat Transmembrane protein LRRTM2, highly enriched and confined at synapses, interacts with Neurexins through its C-terminal cap, but the role of this binding interface has not been explored in synapse formation and function. Here, we develop a conditional knock-out mouse model (cKO) to address the molecular mechanisms of LRRTM2 regulation, and its role in synapse organization and function. We show that LRRTM2 cKO specifically impairs excitatory synapse formation and function in mice. Surface expression, synaptic clustering, and membrane dynamics of LRRTM2 are tightly controlled by selective motifs in the C-terminal domain. Conversely, the N-terminal domain controls presynapse nano-organization and postsynapse AMPAR sub-positioning and stabilization through the recently identified Neurexin-binding interface. Thus, we identify LRRTM2 as a central organizer of pre- and post- excitatory synapse nanostructure through interaction with presynaptic Neurexins.

Fast Collective Hydrogen-Bond Dynamics in HexafluoroisopropanolRelated to its Chemical Activity.

Using fluorinated mono-alcohols, in particular hexafluoro-isopropanol (HFIP), as a solvent can enhance chemical reaction rates in a spectacular manner. Previous work has shown evidence that this enhancement is related to the hydrogen-bond structure of these liquids. Here, we investigate the hydrogen-bond dynamics of HFIP and compare it to thatof its non-fluorinated analog, isopropanol. Ultrafast infrared spectroscopy show that the dynamics of individual hydrogen-bonds is about twice as slow in HFIP as in isopropanol. Surprisingly, from dielectric spectroscopy we find the opposite behavior for the dynamics of hydrogen-bonded clusters: collective rearrangements are3 times faster in HFIP than in isopropanol. This difference indicates that the hydrogen-bonded clusters in HFIP are smaller than in isopropanol. The differences in cluster size can be traced to changes in the hydrogen-bond donor and acceptor strengths upon fluorination. The smaller cluster size canboostreaction rates in HFIP by increasing the concentration of reactive, terminal OH-groups of the clusters, whereas the fast collective dynamics can increase the rate of formation of hydrogen bonds with the reactants. The longer lifetime of the individual hydrogen bonds in HFIP can enhance the stability of the hydrogen-bonded clusters, and so increase the probability of reactant-solvent hydrogen bonding.

From unbound to bound states: Ab initio molecular dynamics of ammonia clusters with an excess electron.

Ab initio molecular dynamics simulations of negatively charged clusters of 2-48 ammonia molecules were performed to elucidate the electronic stability of the excess electron as a function of cluster size. We show that while the electronic stability of finite temperature clusters increases with cluster size, as few as 5-7 ammonia molecules can bind an excess electron, reaching a vertical binding energy slightly less than half of the bulk value for the largest system studied. These results, which are in agreement with previous studies wherever available, allowed us to analyze the excess electron binding patterns in terms of its radius of gyration and shape anisotropy and provide a qualitative interpretation based on a particle-in-a-spherical-well model.

A Short term Electricity Load Forecasting for Community Residents Based on Federated Learning and Considering Privacy Protection

INTRODUCTION: As the penetration rate of renewable energy increases and patterns of energy demand evolve, fluctuations on both the supply and demand sides of electricity are becoming more pronounced. Consequently, accurate forecasting of community residential electrical loads has become crucial. OBJECTIVES: Although the widespread adoption of smart meters among residents provides abundant data for model training, strict challenges arise during the training process due to the need for privacy protection and data security. METHODS: This paper proposes a privacy-preserving community residential short-term electric load forecasting method based on federated learning. Initially, the method applies shared random masking encryption to the sensitive data of community residents, ensuring data privacy while maintaining consistency with the original data after preprocessing. Subsequently, a private data aggregation scheme is established to perform dynamic clustering of the community’s electrical load. RESULTS: The clustered model then serves as the basis for establishing individual load forecasting models for each category of community residents to predict short-term electrical loads. Finally, an empirical analysis is performed using the electrical load data from 120 households across 6 communities in a city in Southern China. CONCLUSION: The analysis demonstrates that the proposed method can achieve the prediction of community residential electrical loads without sharing residents’ data, thus verifying the effectiveness of this approach.

Underlying principles of Ge-induced early cluster-to-layer transition for ultrathin Ag layer formation on oxides

The fabrication of continuous Ag layers with thicknesses of <10 nm are challenging. Moreover, information on the crucial factors responsible for early cluster-to-layer transition is limited. Hence, this study focuses on the effects of thin Ge interlayers in forming Ag layers on SiOx substrates. Experimental and numerical analyses demonstrate the active migration and intermixing of atomic Ge in the Ag and SiOx matrices during the early Ag layering stages. The incorporation of Ge into Ag matrices facilitates a faster cluster-to-layer transition than those with Al and Cu. Ge-mediated Ag clustering reduces the rearrangement momenta of extremely small Ag clusters densely concentrated on SiOx substrates. This is attributed to the Ge-induced reduction in the diffusion of atomic Ag in the surfaces of volatile Ag clusters and consequently, the suppression of cluster coalescence. These findings provide insights into the unconventional role of Ge in Ag clustering dynamics. Thus, this study presents a crucial framework for optimizing Ag wetting on oxide substrates with ultralow optical and electrical losses via extreme dissipation of the residual amounts of metalloid wetting-inducing elements.

The role of irradiation-enhanced interstitial diffusion in over-pressurizing fission gas bubbles in UO2

Fission gas bubbles in UO2 nuclear fuel have been observed to exhibit pressures in excess of the equilibrium bubble pressure; however, the cause of bubble over-pressurization has not yet been demonstrated. The mechanical interaction between a bubble and the surrounding matrix or grain boundary depends on the internal pressure of the bubble and local stress state, such that over-pressurized bubbles are thought to be responsible for fragmentation and pulverization, when exposed to a temperature ramp. Here, we investigate the role of U interstitials, produced through irradiation, in over-pressurizing bubbles by using a combined molecular dynamics (MD) and cluster dynamics approach. Firstly, the energies for the capture of interstitials and vacancies by bubbles have been determined from MD as a function of the ratio of gas atoms to vacancies that make up the bubble. Secondly, these reaction energies have been implemented in the cluster dynamics code Centipede to predict bubble over-pressurization as a function of temperature for typical fission rates. It was found that there is a transition from low pressure bubbles (at high temperatures) to high pressure bubbles (at lower temperatures). The cause of this behavior was shown to be the creation of irradiation-induced interstitials that are highly mobile relative to vacancies at low temperature; whereas, vacancies are sufficiently mobile at high temperatures to limit bubble pressures. This result supports the hypothesis that over-pressurized bubbles form during steady-state operation and that this behavior is highly sensitive to the local pellet temperature.

Computing whole embryo strain maps during gastrulation

Gastrulation is a critical process during embryonic development that transforms a single-layered blastula into a multilayered embryo with distinct germ layers, which eventually give rise to all the tissues and organs of the organism. Studies across species have uncovered the mechanisms underlying the building blocks of gastrulation movements, such as localized in-plane and out-of-plane epithelial deformations. The next challenge is to understand dynamics on the scale of the embryo: this requires quantifying strain tensors, which rigorously describe the differences between the deformed configurations taken on by local clusters of cells at time instants of observation and their reference configuration at an initial time. We present a systematic strategy for computing such tensors from the local dynamics of cell clusters, which are chosen across the embryo from several regions whose morphogenetic fate is central to viable gastrulation. As an application of our approach, we demonstrate a strategy of identifying distinct Drosophila morphological domains using strain tensors.

Dynamic graph attention-guided graph clustering with entropy minimization self-supervision

Dynamic graph attention-guided graph clustering with entropy minimization self-supervision