A key challenge in free space optics/underwater wireless optical communication (FSO/UWOC) hybrid systems is maintaining the adequate power budget and signal to noise ratio (SNR), which necessitates wavelength translation (WT) between visible and infrared spectra at the water-air interface, thus increasing both the cost and complexity. In this work, we propose and simulate a hybrid free space optics/single-mode fiber (FSO/SMF) ring network supporting simultaneous downstream (DS) and upstream (US) transmissions to multiple terrestrial and underwater nodes without requiring local optical sources and WT scheme. The proposed architecture employs differential phase shift keying modulation for downstream channels to enhance turbulence tolerance and reuses the same optical carriers for US on-off keying channels through a dual-drive Mach-Zehnder modulator (DD-MZM) and sinusoidal radio frequency source based pulse carving scheme. The performance of the DS and US channels is analyzed using Bit-error rate (BER) results obtained for different values of refractive index structure parameter (Cn2) and atmospheric attenuation coefficient (αatm) using Gamma-Gamma channel model. Simulation results clearly show that forward-error correction target Bit-error rate of 3.8×10−3 is achieved both for DS and US channels under different turbulence strengths and weather conditions. Node survivability and link inflation analysis are performed for different scenarios. These findings reflect that the proposed ring topology based hybrid FSO/SMF system is flexible and resilient to the adverse channel effects, making it a promising solution for high-speed, long-range future Internet of underwater things (IoUTs) applications.
To address the difficulty in rapidly and accurately identifying excessive nitrate nitrogen and ammonium nitrogen in nitrogen-containing farmland leachate against the complex background of soil leachate, near-infrared (NIR) and ultraviolet-visible (UV-Vis) spectra were acquired, and a nitrogen exceedance discrimination method based on two-trace two-dimensional correlation spectroscopy (2T2D-COS) and N-way partial least squares discriminant analysis (NPLS-DA) was developed. The results showed that 2T2D-COS effectively revealed weak spectral features and their dynamic variation relationships that were difficult to distinguish in one-dimensional spectra of complex systems. For nitrate nitrogen, synchronous spectra achieved better modeling performance for both NIR and UV-Vis data, with test accuracies of 94.55% and 88.46%, respectively. For ammonium nitrogen, asynchronous spectra performed slightly better in NIR, whereas synchronous spectra still performed better in UV-Vis, with both test accuracies reaching 87.50%. Overall, NIR spectroscopy showed better stability and greater robustness to background interference than UV-Vis. These results provide a new methodological reference for the rapid exceedance discrimination of nitrogen-containing farmland leachate.
Plastic pollution of the environment and water is a persistent global concern. Microplastic contamination has impacted aquatic environments, demanding the development of efficient remediation techniques to address this issue. In this work, N-TiO2–supported corncob activated (ZnCl2) carbon (CCAC/N-TiO2) composites (CT13) (CT11), and (CT31) were synthesized through a simple wet-impregnation approach. The prepared composite materials were characterized using Fourier transform infrared (FTIR) spectroscopy, XRD (X-ray diffraction), SEM (Scanning electron microscopy), EDS (Electron dispersive spectroscopy), XPS (X-ray photoelectron spectroscopy), thermogravimetric analysis (TGA), Photoluminescence (PL) spectroscopy, UV-visible spectroscopy and Dynamic Light Scattering (DLS) techniques. Tauc’s method was applied to evaluate the optical band gap, and the results indicated that the composite material’s spectral response extended into the visible-light region, accompanied by a significant reduction in band gap energy. The removal performance of the CCAC/N-TiO2 composites toward PVC-NPs was systematically evaluated under different pH conditions (4, 7, and 10), varying contact times, and various light conditions. CT13 composite demonstrated exceptional performance, achieving a 94% degradation efficiency after 180 min of exposure to tungsten light. The removal of PVC-NPs was determined to occur via a photocatalytic pathway and was confirmed by quenching experiments. Additionally, SEM, FTIR, DLS, and fluorescence microscopy verified the presence of PVC-NPs on the composite surfaces under both dark and light conditions. The major photodegradation products were identified using gas chromatography-mass spectrometry (GC-MS). The addition of CCAC to N-TiO2 significantly improved its ability to remove PVC-NPs. This is because the CCAC addition increased the number of active sites for adsorption. The CT13 composite’s surface attracts and captures the PVC-NPs through a variety of interactions, including hydrophobic interactions, electrostatic attractions, π-π interactions, halogen bonding, and hydrogen bonding. This strong adsorption increases the number of available reaction sites, which in turn boosts the photocatalytic removal of the PVC-NPs. This study sheds light on the use of biomass-derived materials for water purification, providing a long-term solution to pollution and agricultural waste issues.
IntroductionBased on the two-dimensional coupled nonlinear Schrödinger equation, we systematically investigate the collapse dynamics of vector optical fields (VOFs) in a chiral-Kerr medium, where chirality is induced by the unequal refractive indices of left- and right-circularly polarized components.MethodsThe critical power for VOF collapse is derived analytically using the moment method.Results and DiscussionNumerical simulations reveal that the number, locations, propagation distances, and profiles of collapse events are determined by the real and imaginary parts of the chiral refractive indices, the initial power, and the polarization topological charge m. In an achiral Kerr medium, collapse occurs symmetrically. Circular birefringence (CB), i.e., a difference in the real parts of the refractive indices, induces rotation of the state of polarization (SoP) while preserving symmetric collapse positions about the coordinate axes. For a locally linearly polarized VOF, the number of collapse points equals 4×⌈m/2⌉. For a hybrid polarized VOF under CB, the beam differentiates into 4m collapse points. In contrast, circular dichroism (CD), arising from a disparity in the imaginary parts, triggers linear-to-elliptical/circular polarization conversion and shifts the collapse positions, with the rotation direction dictated by the relative magnitudes of n+ and n−. Under low and high initial powers, the hybrid polarized VOF ultimately splits into 2m beamlets concentrated at regions of dominant handedness under CD. The rotation angle of the collapse positions with respect to the Y-axis is π/(4m). These findings establish chiral-Kerr media as a promising platform for controlling vector field collapse and polarization dynamics, with potential applications in chiral sensing, nonlinear optics, and structured light manipulation.
This paper investigates whether a two-phase management strategy—investing in trust before demanding performance—can outperform static policies, and under what cultural conditions this approach succeeds or fails. We develop an agent-based model inspired by the Ising framework to formalize the Care-Dare dilemma in organizational trust dynamics: agents in a well-mixed population interact through trust-building “Care” actions and growth-demanding “Dare” actions, with behavioral tendencies updated via an action-feedback update rule. Unlike the standard Ising model, agents adapt their behavioral tendencies based on the care and dare actions they receive from others, reflecting the social constraint that individuals observe behavior but not latent dispositions. We systematically compare static policies, trust-based dynamic strategies, and fixed-timing controls under individualistic (J = 0) and collectivistic (J = 10) cultural contexts (n = 30 replications per condition). Three principal findings emerge. First, the two-phase Care→Dare structure is the primary driver of performance gains: switching from Care-emphasis to strong Dare intervention (H = −0.5) at 40%–50% of the simulation horizon yields 27%–35% higher cumulative performance than the best static policy (Welch p < 10-4), with performance peaking at this ratio and declining sharply for longer Care phases. Second, trust-based switching provides a state-dependent heuristic that, while switching earlier than optimal (∼20% of the horizon vs. the optimal 40%–50%), still outperforms the best static policy by approximately 25% (300,278 ± 1,511 vs. 239,487 ± 831; Welch p < 10-34) and safely avoids the catastrophic late-switching regime—demonstrating practical value through robustness rather than optimality. Third, cultural conformity creates a sharp regime crossover in a narrow window, J ∈ (1.25, 1.5): moderate conformity (J = 0.5–1.0) amplifies the two-phase strategy’s advantage by 14%, while stronger conformity (J ≥ 1.5) catastrophically reverses it, trapping the organization in an irreversible low-performance state. We connect these findings to the Emotional Bank Account metaphor and discuss implications for management in individualistic and collectivistic organizational cultures.
IntroductionAutomated tyre defect classification is critical for ensuring vehicle safety and manufacturing quality in the automotive industry. Although deep learning has made significant strides in industrial visual inspection, existing convolutional networks struggle to simultaneously capture fine-grained local defect patterns and long-range contextual dependencies in tyre surface images.MethodsTo address this limitation, we propose UniConvNet, an adaptive large-kernel convolutional network that integrates the OverLoCK Context-Mixing Dynamic Convolution (ContMix) with a novel Gaussian-modulated spatial calibration mechanism. The proposed architecture progressively expands the receptive field through a dual-branch design: a multi-scale channel split branch extracts hierarchical local features, while an OverLoCK ContMix dynamic kernel branch adaptively models long-range dependencies. A Gaussian Distribution Modulation module enhances feature robustness against illumination variations, followed by Spatial Feature Calibration to suppress background clutter. Furthermore, a Progressive Large Kernel Regulation strategy 9×9→11×11 balances local detail preservation with global context aggregation. Multi-scale feature fusion aligns and compresses multi-stage outputs for the final classification head.ResultsExtensive experiments on the TyreNet dataset (1,698 images) and a public Roboflow tyre defect dataset (3,069 images) demonstrate that the proposed method achieves 96.45% classification accuracy on TyreNet with only 4.2 GFLOPs and 31.4 M parameters, outperforming ResNet-50, ConvNeXt-T, and OverLoCK-T while maintaining competitive efficiency.DiscussionCross-dataset validation confirms strong generalization, highlighting the practical applicability of the proposed approach for industrial deployment.
Laser-driven ion acceleration has emerged as a promising technique for producing high-flux, energetic proton beams for applications in high-energy-density science, radiography, and inertial fusion energy concepts like ion fast ignition. Recent advances in target fabrication, notably the use of 3D-printed microstructured arrays with graded density profiles, have demonstrated significant improvements in proton energy and conversion efficiency compared to conventional flat foils, particularly in the highly relativistic laser regime. In this work, we present a detailed experimental scaling study of laser-driven ion acceleration using log-pile microstructured targets in the quasi-relativistic, multi-picosecond regime. Experiments were conducted at the OMEGA-EP facility, utilizing short-pulse laser beams with energies up to 1250 J, pulse durations ranging from 0.6 to 10 ps, and focal spot sizes from 14 to 50 μm, corresponding to normalized vector potentials a0 from approximately 0.7–9.2. Target parameters, including micro-wire diameter (0.5-1 μm) and total thickness (10-50 μm), were systematically varied. Proton spectra and beam profiles were characterized using radiochromic film stacks, while electron spectra were measured with a magnetic spectrometer. Our results show that ion performance metrics such as the maximum proton energy and conversion efficiency scale strongly with target parameters including the target average density. These findings provide important scaling relationships for optimizing target design and laser parameters in future high-flux, laser-driven ion acceleration experiments.
Understanding time-dependent deformation in cemented backfill is critical for green mining stability. This study proposes a fractional-order creep constitutive model that couples hardening and damage effects across the full creep process. A hardening function is introduced into the deformation modulus, while a Caputo fractional dashpot is integrated into a modified generalized Kelvin framework. This enables the model to capture both decelerating and accelerating creep phases. Numerical simulations show that low stress levels induce hardening-dominated creep with decreasing rates, whereas high stress levels trigger damage-driven acceleration and eventual failure. Validation against experimental data under three stress levels shows excellent agreement. Comparative analysis demonstrates clear advantages over classical fractional Nishihara models in describing accelerated creep. Parameter sensitivity analysis confirms model robustness and clarifies the distinct roles of fractional order and hardening factors. Overall, this work offers a reliable theoretical tool for predicting creep and assessing long-term stability in cemented backfill structures, with meaningful implications for sustainable mining engineering.
During intergroup confrontations, agitating stimuli such as opponents’ threats and provocations can trigger collective violence, even without the usual mechanisms of ingroup cooperation, such as norms with sanctions. We examine video recordings of street fights between groups of young men. Their violence sometimes breaks out in a burst, wherein a majority of participants starts fighting almost simultaneously. At other times, only few group members participate and it takes them more time to do so. This difference in commencing collective violence can be understood by perceiving it as a collective action dilemma. We adapt an Ising model to show that the proportion of group members who cannot or do not want to contribute to the public good—victory over opponents—predicts whether violence takes the form of a burst or not.
Antenna geometry strongly influences electromagnetic-field distribution, mode transition, and plasma-density formation in low-pressure inductively coupled plasma (ICP) sources. This study compares three external antenna configurations in the same 13.56 MHz, 3.0 mTorr argon plasma system: a helical-type antenna, a planar spiral antenna, and a combined spiral–helical antenna. The comparison is based on spatially resolved measurements of plasma density, RF plasma-potential oscillation, and RF magnetic-field components. The helical-type antenna produced a gradual density increase and a source-localized plasma, whereas the planar spiral antenna showed a clearer abrupt transition from a capacitively influenced low-density state to a higher-density inductive state. The combined spiral–helical antenna gave the strongest overall experimental response, producing the highest measured plasma-density values, the best density uniformity among the three configurations, the broadest useful high-density region, and strong suppression of RF plasma-potential oscillation in the high-density regime. The experimental magnetic-field measurements showed that the combined antenna produced a hybrid electromagnetic structure, with a hill-shaped axial RF magnetic-field profile and an off-axis radial-field structure. To support the interpretation, a verified three-dimensional vacuum-field Biot–Savart benchmark was developed under equal-current conditions. The benchmark is used only as a geometry-controlled field comparison and does not include plasma loading, dielectric boundaries, chamber currents, antenna-current variation, capacitive coupling, or self-consistent particle balance. The combined experimental and numerical results show that RF magnetic-field structure and plasma-density distribution are related but not identical, because density formation also depends on induced electric field, plasma-current density, ionization, transport, and wall losses. The work provides guidance for optimizing antenna geometry in controllable, energy-efficient, low-pressure cold plasma sources. This study should be read as a geometry-comparison case study: energy-efficiency gains are inferred from higher plasma density at equal applied RF generator power, not from direct absorbed-power measurements, and the combined-antenna geometry and 3.0 mTorr operating pressure investigated here represent a single tested case rather than a generally optimized design.
Recent advances in high-power lasers for secondary particle generation highlight the need for reliable and affordable particle sources. In laser-driven ion acceleration, the most routinely obtainable acceleration mechanism is the so-called Target Normal Sheath Acceleration (TNSA) mechanism. Predicting the characteristics of particle sources produced through this mechanism is essential for modeling experiments at existing and future laser facilities. In this work, we present a versatile and fast predictive model, Hermione, capable of reproducing proton spectra across several laser facilities, including the laser facilities PETAL, Apollon, LFEX, and ALLS. Our results show very good agreement with experimental data for laser pulses longer than 100 fs, while a slight overestimation of the proton yield is observed for shorter pulses (<100 fs). We also demonstrate the model’s ability to reproduce wavelength-dependent effects for lasers operating between 0.8 μm and 2 μm. Hermione delivers results on very short timescales (less than 1 minute on a standard laptop), making it an efficient complementary tool to more complex and computationally demanding Particle-In-Cell (PIC) simulations. This enables rapid optimization of the TNSA acceleration regime by varying laser energy, pulse duration, and focal spot size. The code is open access, allowing the community to refine and adapt it to specific applications.
Respiratory motion in thoracic positron emission tomography (PET) introduces spatially heterogeneous non-rigid deformation that can blur lesions, weaken local boundary definition, and reduce structural fidelity. To address this problem, we developed TLCE-morph, a Tri-Path Lie Convolution Encoder-based learning framework for deformable respiratory motion correction in thoracic PET. The framework combines an SO(3)-based group-aware convolution module with a Tri-Path Fusion Encoder to couple orientation-aware geometric modeling with structurally guided feature encoding at local, global, and cross-scale levels. TLCE-morph was evaluated on simulated respiratory motion datasets and a two-center clinical gated PET cohort using Dice, correlation coefficient, and 95th percentile Hausdorff distance. Additional analyses included lesion-level normalized PET uptake consistency, local line-profile and full width at half maximum measurements in motion-sensitive regions, architectural ablation, group-representation comparison, and computational profiling. Across the simulated datasets, TLCE-morph remained comparatively stable as deformation increased from relatively regular displacement to more heterogeneous and coupled motion. In the clinical gated PET cohort, it achieved the most favorable overall quantitative performance among the evaluated methods and showed more consistent local structural recovery in representative motion-sensitive regions. Additional comparisons of group representations and architectural ablation indicated that the observed advantage was associated with the joint contribution of 3D orientation-aware feature modeling and complementary structural constraints rather than with any single component alone. These findings suggest that stable respiratory motion correction in thoracic PET may benefit from coupling geometric sensitivity with structurally guided feature encoding under heterogeneous deformation, rather than relying on appearance matching alone.
This paper presents a novel approach to skin, blood, and breast cancer detection using a penta-band Terahertz (THz) metamaterial absorber and deep learning. The proposed absorber, designed on a Gallium Arsenide substrate, exhibits high absorption peaks of 99.37%, 99.11%, 99.25%, 91.95%, and 99.37% at 0.541 THz, 2.904 THz, 3.291 THz, 3.423 THz, and 3.661 THz, respectively. Cancerous and non-cancerous breast, blood, and skin cells of varying thicknesses are placed on top of the absorber, and the corresponding absorption spectra are obtained under different incident and polarization angles of THz radiation. The underlying principle is that the absorption spectrum varies with the refractive index of the tissue. A dataset of 252 unique absorption spectra is generated for each cell type. Three different deep neural networks have been built, one for each of the following applications: breast cancer detection (Model I), blood cancer detection (Model II), and skin cancer detection (Model III). Model I achieved 100% training accuracy and 94.4% validation accuracy, while Models II and III both achieved perfect accuracy on both training and validation sets (100% and 100%, respectively). On the unseen test data, Model I achieves an accuracy of 100%, Model II achieves an accuracy of 93.88%, and Model III achieves an accuracy of 100%. The results confirm the potential of integrating THz absorption spectroscopy with machine learning for cancer diagnosis. This research offers significant advancements in diagnostic technology, paving the way for more efficient, cancer detection methods.
The structure of an interaction network strongly shapes opinion clustering and the emergence of echo chambers in bounded-confidence (BC) models. We ask whether a controller can steer this clustering by rewiring edges adaptively and how a learned policy compares with hand-designed heuristics. We train a dueling double deep Q-network (DQN) with a candidate-aware state to select degree-preserving edge swaps in a Deffuant model on a sparse Erdős–Rényi (ER) graph (N=500, ⟨k⟩=7.8). Augmenting the state with the signed discord change of each candidate is essential: without it, the agent fails to learn. With it, training reward saturates within ∼200 episodes. Pooling over 10 independent graph realizations and 25 opinion initializations per graph (250 trials per strategy), the learned heterophilic policy reaches a median cluster count C=7.0 at ε=0.18, significantly above the no-rewiring baseline (C=3.0, p<10−3) and the greedy heterophilic oracle (C=6.0). The RL benefit is mode-asymmetric: reinforcement learning (RL) exceeds greedy heterophilic rewiring in 9/10 graphs (Cliff’s δ=+0.26, small effect) but is exceeded by greedy homophilic rewiring in 9/10 graphs (Cliff’s δ=−0.21, small effect). A sweep over ε∈[0.12,0.30] reveals a sigmoidal nucleation barrier for reaching C≥6, with directed strategies giving up to 2.5× speedup over random rewiring. Interpretation: at intermediate ε, the natural attractor is bipolar, so homophilic and heterophilic rewiring amplifies fragmentation; what they control is the depth, not the sign. This finding delimits what rewiring-based interventions can and cannot achieve within the abstract BC framework and identifies the regimes where multistep planning outperforms myopic heuristics.
Cyber-Physical-Social Systems (CPSS) face escalating side-channel threats that undermine secure data transmission and authentication. As China’s national cryptographic hash standard, SM3 is widely deployed in CPSS-integrated social network ecosystems for identity authentication, API signing, and cross-platform data integrity verification—yet its key-dependent input vulnerabilities against side-channel attacks remain inadequately addressed. This study tackles two critical limitations of traditional side-channel attacks for HMAC-SM3 key recovery: non-profiling methods fail due to absent plaintext correlations, while profiling-based approaches suffer from error accumulation and near-zero success rates in single-trace scenarios. We propose a self-calibrating side-channel attack (SC-SCA) that enables high-accuracy HMAC-SM3 key recovery using only a single power trace during the attack phase. The method constructs a Bayesian network to integrate power trace statistics with prior knowledge of input dependencies, then performs joint probabilistic inference via belief propagation. Experimental results demonstrate 100% key recovery success under simulated noiseless conditions, 91.45% success on a real smart card system, and 73% effectiveness at a 10 dB signal-to-noise ratio. Crucially, this work exposes a previously overlooked attack surface in CPSS-based social networks: a single compromised HMAC-SM3 key can enable forged device control commands, large-scale privacy breaches, and cascading identity theft across linked social platforms. Our findings provide both a practical security benchmark for CPSS edge devices and theoretical foundations for designing side-channel-resistant cryptographic implementations.
Analytical tools derived from nonlinear dynamics and dynamical systems theory, such as phase-space reconstruction and Recurrence Quantification Analysis (RQA), provide a powerful framework for investigating complex systems across different scientific domains. These methods allow the identification of dynamical structures, including recurrence, nonlinearity, and transitions between states, in time series data originating from diverse contexts. Scientific research is often shaped by two opposing forces that resemble the dynamics of physics: a centrifugal force, associated with increasing specialization, and a centripetal force, associated with interdisciplinarity. The rapid development of technologies and analytical methods has led to highly specialized languages and frameworks, which, while enabling scientific progress, can also generate fragmentation and communication barriers between disciplines. In contrast, interdisciplinarity emerges as a centripetal force that promotes the identification of shared analytical frameworks across domains. In this context, the transfer of methods is not merely a consequence of mathematical convenience but reflects the presence of common dynamical properties governed by similar physical principles. Artificial intelligence, integrated within physics-informed computational frameworks, provides a powerful tool for analyzing complex, high-dimensional, and heterogeneous datasets while preserving the dynamical structure of the underlying system. This convergence is not merely technical: the same nonlinear dynamical principles that govern physiological and cognitive systems appear to operate within artificial ones, suggesting that AI is not external to the phenomena this manuscript addresses but continuous with them. This inherent interdisciplinarity positions AI as a centripetal force, drawing together methods, languages, and findings from otherwise distant disciplines around a shared dynamical core.
IntroductionThe terahertz (THz) band enables non-destructive molecular sensing of controlled substances including amphetamine, cocaine, and ketamine, which possess characteristic THz absorption fingerprints. Existing THz photonic crystal fiber (PCF) biosensors frequently achieve relative sensitivity below 80% with high confinement losses, limiting their practical utility.MethodsA seven-hexagon void-core photonic crystal fiber biosensor (VCPCFB) was designed in a Zeonex substrate (n = 1.53, α = 0.2 cm−1) with a 10 μm pitch and 700 μm core radius. Full-vector finite element method (FEM) simulations were performed in COMSOL Multiphysics over 1.0–3.0 THz with perfectly matched layer boundaries. Amphetamine (n = 1.518), cocaine (n = 1.5022), and ketamine (n = 1.562) were represented as frequency-independent homogeneous bulk analytes fully occupying the hollow core.ResultsAt 3 THz, relative sensitivity reaches 99.95% for ketamine, 99.84% for amphetamine, and 99.78% for cocaine. Power confinement exceeds 99% for all analytes. Effective material loss reduces to 5.98 × 10−4 cm−1 for ketamine. Confinement loss approaches 0 cm⁻¹ for ketamine at the operating point. Chromatic dispersion ranges between +0.097 and +0.149 ps/THz/cm at 3 THz, with group velocity dispersion below 0.05 ps²/cm. The nonlinear coefficient ranges from 1.87–1.91 × 10−3 W−1cm−1 and self-phase modulation length exceeds 5 × 105 cm at 1 mW. Differential group delay enables ketamine–cocaine separation over ~16.1 cm fiber lengths. Beat lengths exceed 2,353 cm, confirming negligible polarization evolution.DiscussionThe proposed seven-hexagon void-core architecture simultaneously achieves near-unity relative sensitivity, ultra-low propagation loss, flat dispersion, negligible nonlinear effects, and excellent fabrication tolerance. Pitch is identified as the most sensitive structural parameter. The design outperforms all previously reported THz PCF biosensors included in the benchmark comparison, with the proposed maximum sensitivity of 99.95% exceeding the nearest competitor by 7.75 percentage points.
BackgroundAge-dependent social activity and household contact patterns can substantially influence infectious disease transmission. This study aims to develop a network framework that jointly represents age stratification and household contact structures.MethodsThe population was divided into youth, middle-aged, and elderly groups, and family hyperedges were used to represent household membership and generate cross-age household contacts. An Age Family Hyperedge Multilayer Network (AFHMN) was constructed using differentiated intra-layer topologies. Eight simulation scenarios were examined under the SIR and SIRS frameworks, with BA and ER networks used as benchmarks. The models were further evaluated using four empirical infectious disease datasets and the RMSE, MAE, MAPE, R2, and DTW metrics.ResultsThe mixed topology, in which the youth and middle-aged layers use BA networks and the elderly layer uses an ER network, showed the closest agreement with the social characteristics of the three age groups. AFHMN achieved lower RMSE and MAE and higher R² values than the benchmark networks across the four empirical datasets and showed more stable overall agreement with infection peaks and temporal transmission patterns.ConclusionAFHMN provides a practical multilayer modeling framework for investigating how age-dependent social structures and household contacts jointly regulate infectious disease transmission.
Since social interactions are inherently embedded in multiple relational contexts, single-network models often fall short in explaining the evolution of cooperation. This study develop a two-layer coevolutionary model where behavioral strategies in the upper interaction layer are coupled with the lower signed emotional layer, representing friendly or hostile ties. The framework of this study allows behavioral strategies, emotional attitudes, and the network structure to coevolve dynamically. It found that the dynamics of the emotional layer influence the evolutionary outcomes. Counterintuitively, a relatively low cross-layer coupling strength proves more favorable for sustaining cooperation. It also showed that stochasticity is crucial for breaking the monostability of the defection-dominated state. It provides the necessary conditions for the system to enter another stable cooperative state or mixed strategy, effectively leading to the existence of system bistability. In the end, although the evolution of emotions changes the distribution of the final probability of cooperation, it is more like a regulator of cooperation frequency and cannot significantly improve the overall level of cooperation. This highlights how the constantly evolving relationship environment affects the trajectory of social cooperation.
Functional vitality in urban rail transit station areas reflects the concentration of services and urban functions around stations. This study examines 170 operational stations on Lines 1–8 of the Ningbo rail transit network. Using point-of-interest (POI), bus stop, street network, station attribute, and rail network data within 800 m catchments, we constructed a station-area functional vitality index and investigated its built-environment determinants using XGBoost, SHapley Additive exPlanations (SHAP), and partial dependence plots (PDPs). XGBoost outperformed the comparison models, with an R2 value of 0.7213, a root mean square error (RMSE) of 0.0801, and a mean absolute error (MAE) of 0.0589. Distance to the city center, intersection density, and the number of bus stops were the dominant predictors, together accounting for 71.55% of total importance. One-dimensional PDPs revealed marked nonlinear responses: predicted vitality increased at approximately 105–113 intersections/km2 and around 16–17 bus stops but declined rapidly as distance from the city center increased within the first 4 km. Two-dimensional PDPs further suggested that central location conditions the effects of other built-environment factors, while strong street connectivity combined with sufficient bus stop provision corresponds to higher predicted vitality. These findings provide evidence for station-area functional planning and bus–rail integration.
