Search Results (2,585)

Proactive beam management is essential for high-mobility millimeter-wave communication systems, where transmission latency may cause beams selected from current observations to become mismatched with the optimal beams at future transmission times. This paper investigates radar-only causal multi-horizon proactive beam prediction, aiming to directly rank future communication beams from past radar observations. To this end, we propose Shared-Beam-GCN-TCN, a multi-task framework that learns a shared predictive sensing-communication representation and performs structured decoding in the beam space. Specifically, four consecutive range-angle radar maps are encoded by a frame-wise convolutional neural network and a causal temporal convolutional network to capture short-term motion dynamics. The shared temporal representation is jointly optimized for future beam prediction and short-term radar prediction, encouraging the extraction of predictive sensing dynamics relevant to future communication states. A beam-space graph convolutional decoder then exploits the adjacency structure among 64 beam classes to model correlations between neighboring beams and improve future beam ranking. The framework is evaluated in two high-mobility mmWave communication scenarios, including a trajectory-disjoint DeepSense 6G vehicular scenario and a simulated unmanned aerial vehicle (UAV) scenario. In the vehicular scenario, Shared-Beam-GCN-TCN achieves the best overall performance across different prediction horizons, improving the average Top-3 and Top-5 accuracies by 8.2 and 6.9 percentage points over Seq-CNN-TCN, while reducing the mean beam-index error. In the UAV scenario, the proposed framework achieves an accuracy of 93.4% and a beam-index error of 1.022. These results demonstrate the effectiveness of shared predictive representation learning and structured beam-space decoding for causal multi-horizon beam prediction in high-mobility environments. Full article

Reference-library comparisons are widely used in target-focused cheminformatics to characterize the chemical space occupied by bioactive compounds. The separation obtained from such comparisons may, however, reflect differences in library provenance rather than target-related chemistry. The present study quantifies this effect using UI-Ref, a curated reference set of 1218 compounds reported in urease assays in ChEMBL, BindingDB and PubChem, for which assay- and document-level provenance was retained. Within this single target-focused collection, a random forest trained on 25 physicochemical descriptors predicted the depositing database with a receiver operating characteristic area under the curve (ROC AUC) of 0.954 (95% confidence interval, CI, 0.939–0.968) and the source publication with a median one-versus-rest AUC of 0.997, providing an empirical reference scale for provenance-associated separability. Under a conventional pairwise design, comparisons between UI-Ref and external reference libraries yielded AUCs of 0.956–0.982. After property matching and publication-grouped cross-validation, the corresponding AUCs decreased to 0.762–0.841. Because the two tasks differ in terms of class priors, sample size and validation structure, this reference scale is reported as a diagnostic covariate rather than as a decision threshold, and the residual separation was compared against null models constructed within the same design. A null using only features that describe a molecule's relative position within its source library, rather than its absolute physicochemical descriptor values, reached AUCs of 0.995–0.996 in that same task and exceeded the observed models by 0.154–0.234 (paired bootstrap; DeLong Z = −20.1 to −24.6). Models predicting an administrative provenance label within an external library achieved AUCs of 0.740–0.749, close to but below the 0.762–0.841 range obtained for collection membership. This similarity indicates that descriptor-based predictability is not specific to target-related chemistry. The residual separation exceeded the within-pair label-permutation null by 0.263–0.341, which indicates a detectable difference; however, the design cannot distinguish target-related chemistry from residual provenance structure. Following conditioning on provenance predictability, approximately 1% of compound pairs remained comparable. Matching on four physicochemical properties, moreover, left 25 to 27 of the 39 covariates visible to the models imbalanced, including thiourea and chelating motifs. Nearest-neighbor rankings changed after sample-size balancing, and the apparent overlap with approved drugs corresponded to a median nearest-neighbor Tanimoto similarity of only 0.355. Clusters in the joint chemical spaces were associated more strongly with source publication than with collection membership (adjusted mutual information 0.814 versus 0.187, respectively). No association between a structural motif and potency survived hierarchical estimation using a document-level random intercept and false discovery rate control. A revised classification cascade was therefore developed, incorporating explicit label definitions, probability calibration, a prespecified applicability domain and conformal prediction. Between 92.8% and 97.0% of the compounds in each external library fell outside the ECFP4-based applicability domain, and only 10.3% of an external screening cohort completed the cascade with an informative prediction. The measured-active-against-measured-inactive stage is reported as a narrow-scope diagnostic rather than as a general predictive stage, and every model-based stage reports empirical coverage without a conformal guarantee, because exchangeability across source publications is rejected. Provenance-associated confounding can therefore dominate conventional cross-collection comparisons and should be quantified before chemical or biological interpretations are drawn. The curated data, analysis code, trained models and domain-aware application are publicly available. Full article

The protection of intangible cultural heritage (ICH) has now evolved from a conventional mode based on project application and inventory management to a new stage of spatial, networked, and regionally integrated governance. China is moving from list-based management toward spatial governance. This study examines ICH spatial patterns, potential corridor configurations, and regional resource characteristics within an integrated analytical framework. This study examined 3610 sub-items of national ICH across the country. We employed the nearest-neighbor index (NNI) and kernel density estimation (KDE) to characterize the spatial distribution of ICH, the minimum cumulative resistance (MCR) model to identify potential corridor networks, and Ward’s hierarchical clustering to classify provincial ICH resource structures. The findings indicate that: (1) National ICH items exhibit significant spatial clustering (NNI = 0.317), with differences in spatial distribution across the five batches and a greater representation of central, western, and border regions in later batches. (2) Potential corridors form an interconnected network in eastern China and exhibit a linear, bead-like pattern in western China, broadly corresponding to the spatial gradient of lower resistance in the east and higher resistance in the west. (3) Ward’s hierarchical clustering identifies six groups of provincial-level regions with distinct ICH category-frequency profiles. Their spatial distributions are not necessarily contiguous, indicating that similar resource profiles may occur in geographically separated regions. (4) The potential corridor network and provincial ICH resource groups reveal regional differences in the resource characteristics associated with corridor configurations. Eastern corridors connect regions with diverse ICH resources, whereas western corridors predominantly link more dispersed source regions across provinces with different resource profiles. These findings provide a spatial reference for the coordinated protection of ICH across administrative boundaries and the spatial organization of cultural heritage resources, offering insights for the planning of national cultural parks and the integrated development of ICH and tourism. Full article

The digital industrial cluster policy is a key policy instrument for fostering the digital economy and facilitating urban low-carbon transition. Given that no official carbon quota allocation system has yet been established at the city level in China, this paper constructs a multidimensional urban carbon quota index that integrates social, economic, and ecological dimensions, and uses it to measure the carbon quota level of 297 prefecture-level-and-above cities in China over the period 2006–2024. Treating China’s digital industrial cluster policy as a quasi-natural experiment, we employ a staggered difference-in-differences (DID) design to systematically evaluate the policy’s effect on urban carbon quota and the underlying mechanisms. The results show that ① the digital industrial cluster policy significantly raises the urban carbon quota level; ② mechanism tests indicate that this effect is channeled through three paths—stimulating green technological innovation, promoting industrial structure advancement, and improving electricity energy efficiency; ③ heterogeneity analysis reveals that the effect is more pronounced in western regions, in central cities, and in areas with more advanced digital infrastructure and higher carbon quota endowment; ④ spatial econometric results further show that the policy not only enhances local carbon quota but also generates a significant positive spatial spillover effect, raising the carbon quota level of neighboring cities. These findings offer insights for optimizing urban carbon quota allocation through spatial coordination, path coupling, and targeted policy design. Full article

Background: Azelaic acid, an established dermatologic dicarboxylic acid, was previously nominated as an unusual anticancer repurposing candidate. However, it remains unclear whether physicochemical properties, predicted ADME, or structural proximity to established endocrine anticancer drugs support this hypothesis. This study positioned azelaic acid relative to a clinically anchored endocrine anticancer reference panel using reference-defined space based on physicochemical and predicted ADME descriptors and two-dimensional structural similarity analyses. Methods: Sixteen clinically used endocrine anticancer drugs defined the reference set, with finasteride and dutasteride included as mechanistic comparators. SwissADME descriptors were screened for redundancy, standardized, and analyzed by principal component analysis and Euclidean nearest neighbor distances. Robustness was assessed by leave-one-reference-out, bootstrap, and WLOGP sensitivity analyses. Structural similarity was evaluated independently using Morgan fingerprints and Tanimoto coefficients. Results: The first two principal components explained 75.31% of reference variance. Flutamide was the nearest Clinical Reference to azelaic acid in physicochemical and predicted ADME space (d = 3.0063), but this distance exceeded the reference $q_{90}$ threshold (2.6521), classifying azelaic acid as extrapolative. This classification persisted in all 16 leave-one-reference-out analyses and in 73.69% of 992 valid bootstrap replicates. Structurally, the highest reference similarity was observed with toremifene (T = 0.08696), below the minimum internal Clinical Reference nearest-neighbor similarity (0.13761); radius-3 sensitivity analysis preserved this result. Conclusions: Azelaic acid is neither a close structural analog nor a physicochemical and predicted ADME neighbor of established endocrine anticancer drugs. Its network-derived nomination therefore represents a non-canonical repurposing hypothesis requiring further experimental validation. Full article

Molecular dynamics (MD) simulations were used to study the deposition and growth of Fe-Co-Ni-M-Al multi-component alloy films on a Si(100) substrate. By comparing the surface topography at incident energies of 0.1~0.5 eV, it was observed that an increase in incident energy enhances local impact perturbations and induces surface roughening, with the film’s root-mean-square (RMS) roughness increasing from 0.3393 nm to 0.4870 nm. The selection of input data was consistent with the preparation conditions of the electron beam evaporation deposition process. The 0.1 eV condition, which yielded the best surface quality, was selected to study the deposition and growth process. The results indicate that deposited atoms first form a continuous covering layer on the substrate surface, followed by localized accumulation and island-like thickening, exhibiting characteristics of a mixed layer–island growth pattern. The radial distribution function (RDF) exhibits a stable first-neighbor peak at approximately 2.5 Å, with higher-order peaks and shoulders gradually appearing during the deposition process, indicating that short-range order formed within the film and gradually evolved into medium-range order. Polyhedral template matching (PTM) results show that the film exhibits a non-equilibrium structure characterized by a predominance of FCC structure, accompanied by the coexistence of HCP, BCC, and amorphous structures. Based on the solute formation rules derived from the atomic size difference δ and Ω parameter, as well as the empirical criteria of VEC, this composition exhibits the potential to form a solid solution and a structure with a tendency toward FCC. Meanwhile, both Debye-simulated XRD results and experimental data show a broadened main peak near 44°, cross-sectional SEM characterization confirms that the film surface is not completely flat but exhibits some local topographical variations, and TEM characterization reveals the presence of nanocrystalline/polycrystalline structures within the film; together, these findings demonstrate the ability of the MD model to predict the average atomic spacing and primary local order characteristics of the experimental film. Full article

Stereoscopic vision relies on the ability of the visual system to estimate binocular disparity across the visual field. In computational models, stereo matching can be represented through multiple candidate disparity hypotheses defined over retinotopic position. In this work, we reinterpret the Stereo Cellular Neural Network, here referred to as Stereo-CeNN, within a continuous variational framework and introduce a neural-field representation defined over the joint domain of retinotopic position and disparity. In this interpretation, the discrete activities of the Stereo-CeNN are regarded as samples of a continuous activity field r(x,d,t), where different disparity layers correspond to populations tuned to different disparity preferences. The binocular input drives these populations according to local matching evidence, while spatial interactions within each disparity channel promote coherent activity across neighboring retinotopic locations. The final disparity estimate is obtained, consistently with the original Stereo-CeNN architecture, by selecting the disparity channel with the largest steady-state response at each retinotopic position. We show that the layer-wise organization and local spatial dynamics of the Stereo-CeNN admit a continuous neural-field counterpart, establishing an explicit structural and variational correspondence between the discrete retinotopy–disparity lattice and a population-level representation of stereopsis. This formulation provides a conceptual bridge between the independently developed Stereo-CeNN architecture and the mesoscale organization of disparity-selective populations observed in visual cortex, without implying a one-to-one anatomical or physiological correspondence. Because Cellular Neural Networks are locally connected dynamical systems originally conceived for analog and mixed-signal VLSI implementation, the same organization is also compatible with general neuromorphic principles of local, parallel, and distributed computation. Full article

Urban greening programs in Chinese megacities are premised on the assumption that vegetation cools the surface and improves air quality simultaneously, yet greenness is usually quantified in two dimensions and the thermal and air-quality services are evaluated separately, so the places where they reinforce or conflict remain unmapped. We measured effective leaf area index (LAIe) in 570 quadrats across 30 vegetated plots in Guangzhou, China, and found that, under the spatial and temporal conditions tested, planar spectral data represented it poorly: Moderate Resolution Imaging Spectroradiometer (MODIS) LAI was only weakly correlated with field values (r = 0.15, n = 395), 87.9% of LAIe variance occurred within plots, and all 16 locally calibrated vegetation-index models showed no skill under plot-grouped cross-validation (CV-R² ≤ 0). We therefore represented three-dimensional greenness with a structural proxy, layered vegetation volume (LVV, m³ m⁻²), defined as the product of cell-mean 10 m canopy height and tree-cover fraction (a cover-weighted canopy-height index), and coupled it with warm-season Landsat land surface temperature (LST) and 1 km estimates of PM_2.5 (2016–2022), warm-season daily maximum 8 h average (MDA8) O₃ (2016–2022) and NO₂ (2019–2022) on 1, 2, 3 and 5 km grids, analyzed with ordinary least squares (OLS), geographically weighted regression (GWR), multiscale geographically weighted regression (MGWR) and extreme gradient boosting with Shapley additive explanations (XGBoost–SHAP). The results show the following: (1) MGWR achieved lower corrected Akaike information criterion (AICc) than OLS and GWR at every scale where it was fitted (2, 3 and 5 km), and the LVV term operated at a bandwidth of 43 neighbors (median radius ≈ 12 km) versus ≥863 neighbors (≈108 km, effectively citywide) for meteorological covariates; (2) one standard deviation of LVV (7.4 m³ m⁻²) was associated with 1.3 °C lower warm-season LST (significantly negative in 75% of cells at |t| > 1.96 and 37% under MGWR-adjusted critical values), 0.30 μg m⁻³ lower PM_2.5 (69% and 50%) and 1.3 μg m⁻³ lower NO₂; (3) the citywide O₃ effect was null, but MGWR revealed offsetting local responses (31% of cells significantly negative, 17% positive), and within-cell greening was associated with rising O₃ (+0.19 SD, p < 0.001), and annual coefficients turned non-negative after 2020 as nitrogen oxides (NO_x) declined; (4) the PM_2.5 benefit leveled off near 13.6 m³ m⁻² (bootstrap 95% CI 8.4–15.2), while the cooling response flattened gradually with a poorly constrained breakpoint, and cells classified as full synergy, partial synergy, O₃ trade-off and insensitive covered 44.6%, 36.3%, 15.0% and 4.1% of the municipality, with trade-off cells concentrated in the hottest, most built-up southern lowland belt; under stricter significance criteria the classified shares contracted, while the trade-off cells remained in this belt. These results suggest that structure-based, zone-specific greening targets may help secure co-benefits while anticipating a possible ozone penalty in NO_x-rich areas. Full article

A previously reported crystalline assembly of the γ-form of trans-1,2-bis(4-pyridyl)ethylene (BPE) with citric acid (CA) was investigated under UV irradiation to examine its structural evolution and solid-state [2 + 2] photodimerization. Single-crystal X-ray diffraction was used to compare the BPE-CA (1) crystal before irradiation and after 30 min and a total of 2 h of irradiation. Structure 1 is a proton-transfer salt in which charge-assisted hydrogen bonding organizes the BPE components into parallel two-dimensional assemblies containing several offset olefin pairs. Selected neighboring C=C bonds satisfy the geometric requirements for topochemical photodimerization. Following irradiation, single-crystal diffraction data could still be collected with the same space group and closely similar unit-cell parameters. Complementary irradiation of ground crystals for 24 h, followed by dissolution and ¹H NMR analysis in DMSO-d₆, confirmed the quantitative formation of the two cyclobutane photodimers with rctt-tpcb as the major product. A separate photodimeric proton-transfer salt, Structure 3, was also identified crystallographically in crystals obtained under ambient laboratory illumination. Its photodimer topology is related to that of the previously reported α-form photoproduct, although its proton-transfer state distinguishes it from the reported neutral cocrystal. These findings show that proton transfer and hydrogen-bond-directed organization by CA govern the relative arrangement of the BPE olefins and promote solid-state [2 + 2] photodimerization. The crystallographic results further demonstrate that the lattice retains its single-crystalline character during UV-induced structural reorganization. Full article

Educational virtual worlds and metaverse environments are increasingly utilized to foster immersive and experiential learning in higher education. However, their instructional efficacy depends not only on visual fidelity and presence, but fundamentally on how effectively learners’ cognitive resources are managed. This study investigates intrinsic (ICL), extraneous (ECL), and germane cognitive load (GCL) alongside NASA Task Load Index (NASA-TLX) workload dimensions in a desktop-based educational virtual world developed in Spatial.io for an undergraduate Computer Networks course. A cohort of 120 second-year Computer Science students (75 males, 45 females) participated in a five-week intervention designed to model the relationships among cognitive load, perceived workload, and meaningful engagement. A multi-method analytical framework was implemented, combining psychometric validation and Partial Least Squares Structural Equation Modeling (PLS-SEM) with machine learning techniques (K-Means clustering, K-Nearest Neighbors (KNN), Random Forest, Decision Trees) and Psychometric Network Analysis. The findings revealed strong internal consistency across the cognitive load scales, moderate overall workload, and significant positive associations between all three cognitive load dimensions and perceived workload, with extraneous cognitive load emerging as the variable most strongly and consistently linked to high-overload cluster membership, frustration, and reduced engagement across the machine learning and network analyses. These results suggest that desktop-based educational virtual worlds may support self-reported germane cognitive engagement, provided that extraneous interface demands are systematically minimized and navigation is scaffolded. The study contributes an empirical and computational perspective to educational metaverse research, offering practical guidelines for designing cognitively sustainable immersive learning environments. Full article

Textual-edge graphs represent interactions through both connectivity and relation-specific text. Learning community structure from these graphs requires an encoder that preserves the relevance of individual relations when combining semantic and structural information. We propose Dual-Stage Attention Embedding for Community Detection on Textual-Edge Graphs (DAETG). The first attention stage summarizes incident-edge text, and the second transfers its coefficients into neighboring-node aggregation as a learnable prior. Following topology-reconstruction pretraining, Kullback–Leibler self-training and differentiable modularity jointly refine the encoder and a centroid-based community assignment. Training uses fixed node and edge text embeddings and a supplied community count, without category-membership supervision. Experiments on four benchmark graphs with base and large Bidirectional Encoder Representations from Transformers (BERT) embeddings show competitive community-detection performance, with modest mean improvements that vary across datasets and evaluation metrics. Ablation studies on two graphs support the contributions of correctly associated edge text, the dual-stage encoder, and joint community refinement. Together, these results support using relation-specific text alongside topology to learn community representations. Full article

Hyperspectral image classification requires joint modeling of spectral signatures and spatial structure. In patch-based classification, each input patch is assigned the label of its center pixel, although neighboring pixels may belong to different land-cover classes. This setting motivates coordinating shallow spatial features, contextual refinement within the patch, and supervision of the resulting representation. We propose DSFormer++, an enhanced framework based on the original Dual Selective Fusion Transformer Network (DSFormer). Shallow Feature Compensation (SFC) combines shallow and patch-embedded features through gated residual fusion before the backbone. The Collaborative Global–Local (CGL) module applies an additional refinement operation with shared parameters before token selection at two backbone stages. Pixel Contrastive Loss (PCL) uses center-pixel labels to supervise projected patch embeddings during training. The proposed method achieves overall accuracies of 96.50%, 97.60%, 95.35%, and 94.83% on the Pavia University, Houston13, Indian Pines, and WHU-HH datasets, respectively. Component comparisons and approximately parameter-matched controls characterize configuration-dependent effects. Numerical checks assess consistency between the stated operations and their implementation, and an additional Botswana experiment extends the evaluation to a satellite wetland and woodland scene. Full article

Digital industrial policies are increasingly used to promote economic restructuring and spatial reorganization, yet their effects on urban spatial structure remain poorly understood. This study examines whether China’s National E-commerce Demonstration City (NEDC) policy acts as a catalyst for urban development or as a constraint on excessive agglomeration. Using a national sample of 293 prefecture-level cities in China from 2006 to 2020, we decompose nighttime light intensity into average intensity (overall urban development), maximum intensity (core agglomeration), and minimum intensity (peripheral activity). We employ propensity score matching, staggered difference-in-differences, and spatial Durbin models to estimate both direct effects on treated cities and spillover effects on neighboring cities. The results indicate that the policy acts as both a catalyst and a constraint. As a catalyst, the policy significantly increases average intensity and generates positive spillovers to neighboring cities. As a constraint, it reduces maximum intensity in treated cities, but this effect does not extend to neighboring cities. The effect on minimum intensity is positive but confined to treated cities. These findings are consistent with a locally confined form of digitally enabled decentralization. By decomposing nighttime light data into three metrics and examining both direct and spillover effects, this study enriches the existing literature on digital policy evaluation and urban spatial restructuring, and provides targeted infrastructure suggestions for balanced regional development. Full article

This short study presents the results of micro-Raman investigations performed on the alloy Ti-6Al-4V, which was additively manufactured via metal fused filament fabrication (MF³). Local laser micro-annealing was applied at different power intensities to distinct regions. The applied laser power density was varied in the range of ~0.05 kW/cm² to ~5 MW/cm². Laser-induced compositional changes in structures, approx. 1 µm in diameter, arranged in an array, are laterally limited. Simultaneously, the surrounding matrix maintains its initial pristine material properties down to ~500 nm in spacing between neighboring structures, reproducibly. The presented results confirmed that the material contains residues of the organic-based binder system and its decomposition products. Besides the characteristic Raman modes related to hydrocarbon compounds, TiC, and defect carbon phases were identified. Unexpectedly, micro-Raman mapping measurements performed on regions exposed to laser power densities above 2.5 MW/cm² for up to 2 min revealed the complete suppression of hydrocarbon modes. Additionally, laser micro-annealing applied under identical operating conditions initiates “carbonization” processes in regions predominantly composed of residues from the organic-based binder used in material preparation. This short initialization study is a prerequisite and first building block for the development of suitable laser annealing technology which allows the modification and functional texturization of Ti-6Al-4V alloys down to the sub-micro scale. The correlation between modified materials’ structural and mechanical properties will be the aim of our future studies. This would allow the design and fabrication of sophisticated tools using MF³ technology for, e.g., surgery biocompatible implants with locally tuned, favorable mechanical properties. Full article

The National Digital Economy Innovation and Development Pilot Zone (NDEIDPZ) policy, as a form of digital place-based policy, can strongly promote the high-quality development of the digital economy in China, generating broad-based benefits across economic, social, and environmental dimensions. This study adopts a synergistic perspective that integrates economic growth, social inclusion, and environmental sustainability to assess the impact of the NDEIDPZ policy. Utilizing a quasi-natural experimental design and a difference-in-differences (DID) model, this study deeply dissects the influence of this digital place-based policy on inclusive green growth (IGG). The main finding indicates that the NDEIDPZ policy can significantly promote IGG. Mechanistic analysis reveals that promoting digital technological innovation, enhancing entrepreneurial activity, alleviating resource misallocation, and accelerating industrial structural upgrading constitute important mechanisms underlying the contribution of the NDEIDPZ policy to IGG. The effects of the NDEIDPZ policy are not uniform across cities, contributing more substantially to IGG in eastern cities, non-resource-based cities, and cities with relatively more developed digital economies. The spatial spillover analysis further demonstrates that the NDEIDPZ policy exerts positive spillovers to IGG in neighboring regions, while the magnitude of this spillover effect declines progressively with increasing geographical distance. These findings not only extend the existing literature on the effects of the NDEIDPZ policy by examining its implications for IGG, but also provide theoretical and empirical insights into how place-based policies can be designed and implemented to advance sustainable development goals. Full article