Class imbalance is a formidable and ongoing challenge in image classification tasks. Existing methods address this issue by emphasizing minority classes through class redistribution in the feature space or adjusting decision boundaries. Although such approaches improve the accuracy of minority classes, they often lead to unstable training and performance degradation on majority classes. To alleviate these challenges, we propose a unified redistribution framework termed as ComReg, which explicitly enforces complementary regularization on feature learning and decision boundary optimization in long-tailed image classification. Specifically, ComReg employs a multi-expert learning framework combined with prior-knowledge-guided online distillation to construct distribution-aware decision boundaries. From the feature space learning perspective, we enhance intra-class compactness and inter-class separability through decoupled-balanced contrastive learning. To further align the distributions in both spaces, we introduce a delay-weighted prototype learning strategy, which incorporates the decision boundary constructed by the head-class expert into the decoupled-balanced contrastive learning process. Extensive experiments on widely used long-tailed benchmarks, including CIFAR10-LT and CIFAR100-LT, as well as the real-world long-tailed datasets such as subsets of MedMNIST v2, demonstrate that our method achieves state-of-the-art performance.

The anomalous Hall effect (AHE) has been understood as a transport phenomenon usually observed in ferromagnetic and non-collinear antiferromagnetic materials where broken time-reversal symmetry combined with spin-orbit coupling produces a net Berry curvature. Combined spatial inversion (P) and time-reversal (T) symmetries forbid an AHE in collinear antiferromagnets. In this study, we demonstrate a pronounced AHE in (110)-oriented FeRh thin films epitaxially grown on Al2O3 substrates, even though the films remain collinear antiferromagnetic at low temperatures. Unlike the bulk B2 phase, where PT symmetry enforces the cancellation of Berry curvature, the epitaxial (110) orientation and substrate-induced strain explicitly break the spatial inversion symmetry (P). This symmetry-lowering mechanism, which lifts the PT constraint, enables a finite Berry curvature distribution in momentum space. Consequently, this allows for robust anomalous transverse transport even in the collinear antiferromagnetic regime, providing a new degree of freedom to engineer topological properties in antiferromagnets. First-principles density-functional calculations reproduce the strain-induced Berry curvature and quantitatively account for the measured AHE in the 5-100 K range. Our results show that intrinsic strain can be harnessed to tailor Berry curvature in collinear antiferromagnets, opening a pathway toward antiferromagnetic spintronic applications.

The quality of three-dimensional macromolecular image reconstruction by cryo electron microscopy (cryo-EM) strongly depends on the number and the quality of the respective two-dimensional projections and on their angular distribution in space. Distributions with one or a few strongly preferred particle orientations may result in maps that are deformed in certain directions. A simple removal of overrepresented views may improve the quality of the reconstructed maps when the level of noise in the two-dimensional (2D) projections is low and the data-set size can afford this removal, but is counterproductive otherwise. Complementarily, giving an increased weight to underrepresented views, or taking multiple copies of them during the reconstruction, may improve the results, naturally, depending on how non-uniform the view distribution is. This work describes the results of three-dimensional (3D) reconstructions using an explicit correction of the number of overrepresented and underrepresented projections for non-uniformly distributed sets. Such correction can be considered as a potential preprocessing, fast and simple, during 3D reconstruction in the image-processing and cryo-EM structure-determination workflow.

To study the crack resistance of UHPC precast composite slabs, this paper conducts flexural performance tests on one UHPC monolithic slab and four UHPC precast composite slabs, investigating the influence of structural form, loading method, and shear reinforcement on the failure mode and crack resistance of UHPC precast composite slabs. The test results showed that UHPC precast composite slabs do not experience shear failure along the composite interface. They exhibit extensive microcracks and do not fail due to the immediate appearance of a single wide crack, demonstrating good plasticity and toughness. The cracking load of the monolithic slab is 6.6% to 12.5% higher than that of the composite slabs. However, the yield load and ultimate load of composite slabs equipped with shear reinforcement are 19.5% to 26.5% and 24.5% to 29.5% higher than those of the monolithic slab, respectively. These composite slabs are also characterized by extensive, dense microcracks with high quantity, small width, small spacing, short length, and dense distribution. Shear reinforcement can effectively improve the bearing capacity and crack resistance of UHPC precast composite slabs, with truss reinforcement showing a better effect in enhancing bearing capacity and inhibiting cracks. The comparison between positive and reverse loading methods better explains the “strain lag” of concrete and “stress advance” of reinforcement in composite slabs. Based on the section internal force equilibrium and the bond stress transfer principle between reinforcement and concrete, considering the enhancement effect of UHPC on bond stress, the calculation formulas for average crack spacing and maximum crack width in existing codes are modified. The calculated values are in good agreement with the test results.

The rapid expansion of urban areas has intensified the Urban Heat Island (UHI) effect, particularly in high-density cities characterized by extensive impervious surfaces and limited green spaces. Rising surface temperatures increase energy consumption, reduce outdoor thermal comfort, and pose serious public health risks during extreme heat events. Urban greening is widely recognized as an effective heat mitigation strategy; however, in densely built environments, indiscriminate or uniform distribution of green spaces often fails to achieve optimal cooling benefits. This study proposes an artificial intelligence–based framework for strategically optimizing urban green space placement to maximize heat reduction while accounting for land-use constraints. The proposed approach integrates multisource remote sensing data, vegetation indices, land surface temperature measurements, and urban morphological indicators with machine learning–based thermal modeling. A Random Forest Regression model is employed to capture the nonlinear relationships between vegetation cover, built-up density, and surface temperature, followed by a spatial optimization process to identify priority locations for greening interventions. Experimental results demonstrate a strong negative relationship between vegetation density and land surface temperature, with optimized greening scenarios achieving temperature reductions of up to 2.6°C, significantly outperforming uniform greening strategies with equivalent green area allocation. The findings highlight that the spatial configuration and targeted placement of green spaces are more influential than total green cover alone. By incorporating explainable AI techniques, the framework also provides interpretable insights into the dominant drivers of urban heat, enhancing transparency for planning applications. Overall, this study offers a data-driven and decision-oriented methodology that can support urban planners and policymakers in designing effective, climate-resilient strategies for mitigating heat stress in high-density urban environments.

C4 plants operate a highly efficient photosynthetic CO2 concentrating mechanism. However, C4 photosynthesis represented by maize is based on the typical Kranz-type leaf anatomy, which involves complex regulation of vascular development coupling with metabolic distribution. To explore the possibility of using alternative C4 leaf anatomy as reference for engineering C3 crops, we sequenced, assembled and annotated the genome of Arundinella anomala, a C4 grass with variant Kranz anatomy and interveinal distinctive cells (DC). Following single-cell level transcriptomes for comparative analyses between C4 bundle sheath and DC cells, genetic and metabolic support for the intensified C4 function of DC cells were observed, including increased cyclic photosynthetic electron transport, carbon fixation and starch synthesis. Further, the mechanism involving SHORT-ROOT (SHR) and auxin to trigger independent development or proliferation of DC cells was explored. Notably, spaced distribution of DC-like cells can be achieved in rice leaves by inducing the expression of ZmSHR1. This work laid a foundation for introducing functional DC-like cells among the intervascular mesophyll cells of C3 grass leaves, and provided resources and strategies for engineering C4 traits into C3 crops.

Microvascular networks (MVNs) formed via endothelial cell self-assembly in 3D hydrogels have emerged as a widely used platform for modeling vascularized tissues and studying vascular pathophysiology. Conventional MVN systems incorporate supporting fibroblasts and may include biochemical cues such as VEGF, FGF, or S1P, as well as mechanical stimuli like luminal flow, yet the impact of these variables on MVN morphology and function remains incompletely understood. Here, we systematically investigated the effects of fibroblast concentration, fibroblast-conditioned media, angiogenic factors, and luminal flow on the morphology, perfusability, and vessel wall integrity of MVNs cultured in microfluidic vasculature-on-a-chip. In addition to standard branch-based metrics, such as vessel coverage area and vessel diameter, we developed and applied novel void-based morphological parameters that quantify the size, shape, and spatial distribution of vessel-free spaces. These metrics enabled us to capture subtle morphological differences across MVN culture conditions and to quantify the dynamic morphogenesis events that shaped the resulting MVNs including branch formation, vessel fusion, and pruning. Our results demonstrate that high fibroblast-to-endothelial cell ratios accelerate MVN formation but promote excessive vessel fusion, while MVNs cultured without fibroblasts─using only conditioned media or soluble factors─exhibited patch-like, nonphysiological morphology with reduced branch formation. Direct inclusion of fibroblasts proved to be essential for promoting the thin, interconnected vascular structures characteristic of in vivo microvasculature and could not be substituted by soluble cues alone. Overall, our void-based analysis method enabled more sensitive discrimination of MVN morphological features than traditional branch-based metrics and offers a reduced-data, high-content approach suitable for potential integration with machine learning and AI-assisted image analysis pipelines. This platform provides a new framework for optimizing MVN culture protocols and advancing vascular tissue engineering studies, particularly for the advancement of organ-on-a-chip (OOC) and microphysiological systems.

A fundamental question in ecology is why plant communities have large trait space yet strong coordination among those traits across large scales, despite these patterns seeming contradictory. Answering this question requires quantitatively linking the geographic distribution of trait space and coordination with gross primary productivity (GPP). We leveraged an unprecedented large-scale dataset of nine leaf traits for 5718 species-site combinations with simultaneous field measurements of plant community composition in 64 naturally assembled communities to investigate trait spaces (hypervolume, quantity dimension) and trait compactness (coordination, efficiency dimension) and their influence on GPP. Trait space and compactness combined explained 72% of the variation of GPP. Interestingly, a larger trait space (more diverse trait combinations) drove higher GPP in resource-poor communities, while higher trait compactness (greater coordination of traits) determined higher GPP in resource-rich communities. Our findings provide a new perspective that natural plant communities increase both trait space and compactness to improve GPP, shedding light on the development of multidimensional functional ecology.

Pre-ionization assisted synthesis of tailored magnesium‑aluminum bimetallic metal-organic framework-derived carbon with controlled layer spacing and porosity for performance supercapacitor and sodium-ion battery.

Magnetic particle imaging technology, a novel medical imaging technology, possesses rapid imaging, high penetration depth, and is free from ionizing radiation. However, the system point spread function causes imaging blurring, which can be further exacerbated by external environmental interferences. Although hardware improvements and system optimization can mitigate blurring, these approaches are often expensive and time-consuming, particularly for low-field imaging in large-scale systems. This article proposes a Fast Context-aware Saliency-enhanced Deblurring Network, FCS-edNET, to solve the challenging issue by deblurring the reconstructed images. The network introduces the Multi-scale Global module to enhance the multi-scale feature perception ability. The Multi-scale Denoising Prior algorithm, which employs a low-frequency filter operator to restrict image noise and offers priors for each layer of subnetworks, is designed to improve the model robustness. Finally, proposing a Multi-level Joint loss optimizes model parameters to promote model convergence speed and space distribution simulation capability. Extensive experiments on multiple public and private datasets demonstrate that FCS-edNET outperforms the state-of-the-art methods in MPI image deblurring efficiently, suggesting its potential to support future research toward clinical imaging applications. The code is available at https://github.com/ydz1118/FCS-edNET.

Multi-tracer positron emission tomography (PET), which assesses key neurological biomarkers such as tau pathology, neuroinflammatory, -amyloid deposition, and glucose metabolism, plays a vital role in diagnosing neurological disorders by providing complementary insights into the brain's molecular and functional state. Acquiring multi-tracer PET scans remains challenging due to high costs, radiation exposure, and limited tracer availability. Recent studies have attempted to synthesize multi-tracer PET images from structural MRI. However, these approaches typically either rely on direct mappings to individual tracers or lack distributional constraints, leading to inconsistencies in image quality across tracers. To this end, we propose a normalized diffusion framework (NDF) to generate high-quality multi-tracer PET images from a single MRI through a distribution-guided class-conditioned weighted diffusion model. Specifically, a diffusion model conditioned on MRI and tracer-specific class labels is trained to synthesize PET images of multiple tracers, and a pre-trained normalizing flow model refines these outputs by mapping them into a shared distribution space. This mapping ensures that the subject-specific high-level features across different PET tracers are preserved, resulting in more consistent and accurate synthesis. Experiments on a total of 425 subjects with multi-tracer PET scans demonstrate that our NDF outperforms current state-of-the-art methods, indicating its potential for advancing multi-tracer PET synthesis.

ABSTRACT Urban bays in coastal cities like Qingdao play a pivotal role in fostering sustainable human-nature symbiosis amid rapid urbanization. However, economic-driven development has led to fragmented spatial hierarchies and heterogeneous open space distributions, limiting their capacity to meet diverse user demands. Despite their importance for localized human-environment interactions, small-scale bays have received limited scholarly attention. This study addresses this gap by employing fractal theory to analyze the hierarchical structure, scale transitions, and spatial organization of open spaces in two contrasting bays: Qingdao Bay (historically evolved) and Fushan Bay (modern development). Using grid fractal dimension analysis, box-counting methods, and GIS tools, we quantify fractal characteristics, identify scale discontinuities, and evaluate spatial balance. Results reveal that Qingdao Bay exhibits a fractal dimension of 1.784 (R2 = 0.998), aligning closely with ideal urban fractal patterns, while Fushan Bay’s higher fractal dimension (1.839, R2 = 0.999) indicates excessive spatial fragmentation. Key issues include oversized coastal parks disrupting scale continuity, insufficient human-scale interaction spaces, and homogeneous functional zoning. To address these, we propose a fractal-optimized hierarchical framework emphasizing multi-scale integration, transitional spaces, and mixed-use planning. This research advances the application of fractal theory to small-scale coastal systems in Qingdao, offering actionable insights for enhancing spatial quality within similar urban bay contexts.

For each integer n ≥ 1 n\geq 1 , we construct a G L n ( Q ) GL_n(\mathbb {Q}) -invariant modular symbol ξ n \boldsymbol {\xi }_n with coefficients in a space of distributions that takes values in the Milnor K n K_n -group of the modular function field. The Siegel distribution μ \boldsymbol {\mu } on Q 2 \mathbb {Q}^2 , with values in the modular function field, serves as the building block for ξ n \boldsymbol {\xi }_n ; we define ξ n \boldsymbol {\xi }_n essentially by taking the n n -Steinberg product of μ \boldsymbol {\mu } . The most non-trivial part of this construction is the cocycle property of ξ n \boldsymbol {\xi }_n ; we prove it by using an induction on n n based on the first two cases ξ 1 \boldsymbol {\xi }_1 and ξ 2 \boldsymbol {\xi }_2 ; the first case is trivial, and the second case essentially follows from the fact that Beilinson-Kato elements in the Milnor K 2 K_2 -group modulo torsion satisfy the Manin relations.

Customizing diffusion models via Low-Rank Adaptation (LoRA) has become a standard approach for customized concept injection. However, synthesizing multiple customized concepts within a single image remains challenging due to the parameter pollution problem, where naive fusion leads to gradient conflicts and severe quality degradation. In this paper, we introduce ConWave-LoRA, a novel framework designed to achieve hierarchical disentanglement of object and style concepts in LoRAs. Supported by our empirical validation regarding frequency distribution in the latent space, we identify that object identities are predominantly encoded in high-frequency structural perturbations, while artistic styles manifest through low-frequency global layouts. Leveraging this insight, we propose a Discrete Wavelet Transform (DWT) based filtering strategy that projects these concepts into orthogonal optimization subspaces during contrastive learning, thereby isolating structural details from stylistic attributes. Extensive experiments, including expanded ablation studies on LoRA rank sensitivity and style consistency, demonstrate that ConWave-LoRA consistently outperforms strong baselines, producing high-fidelity images that successfully integrate multiple distinct concepts without interference.

Abstract. With the rapid urbanisation, Green Spaces (GS) are shrinking steeply in urban areas. Parks not only provide green spaces in an urban area, but play a pivotal role in the life of individuals. Apart from being the lungs of the city, they help to reduce the effect of urban heat islands. The current study aims at mapping all public parks sourced and integrated from Open Street Map, Google Earth and Pune Municipal Corporation (PMC) database in PMC using Geographic Information Systems (GIS). As per the integrated database, there are 216 public parks in the PMC area. We have analysed the spatial distribution and coverage of green spaces in the PMC. The spatial distribution reveals that the public parks are located in all 15 administrative wards of Pune. However, the analysis indicates a clustered pattern; most of the parks are concentrated in the central part of the city and as the city limits expand, there is a noticeable drop in the coverage of public parks. Cluster boundaries of the existing public parks collectively indicate that many residential areas of PMC are deprived of this GS. In the future, it will be useful to study the existing infrastructure and the usage of existing public parks. Also, an additional important aspect is studying the proximity and access to GS from densely packed informal settlements in the PMC area.

Objective – This study aims to observe social vulnerability in the face of the expansion of informal urban settlements in the city of São Paulo, seeking to understand this urban phenomenon. Methodology – The methodology used in this study is action research, in which the problems identified by the theoretical framework must also be taken into account for the reliability and validity of the research. The qualitative approach and exploratory and descriptive nature of the research contributed to understanding the interaction of elements related to the object of study and the understanding of the phenomenon as a whole. The bibliographic research conducted in the Web of Science, Scopus, and Google Scholar databases focused on the expansion of informal urban settlements, socio-spatial segregation, social vulnerability, and included the analysis of public documents, such as data from the Municipal Housing Secretariat, IPEA, and IBGE. Originality/Relevance – The research carried out is innovative for academic theory and practice, as there are no precedents for works that associate socio-spatial segregation and Social Vulnerability Indices with the expansion of informal urban settlements in the Municipality of São Paulo. Results – This study identified a gap in the organization and distribution of urban space, which effectively contributes to the expansion of the housing crisis, highlights infrastructure problems, and drives high rates of poverty and vulnerability. Therefore, informal urban settlements have become the predominant spatial segmentation in the peripheries, requiring the implementation of emerging public policies and the implementation of investments and services aimed at urban development as a whole. Theoretical/Methodological Contributions – The research carried out is relevant to academic theory and practice, as there are no precedents for works that associate socio-spatial segregation and Social Vulnerability Indices and the expansion of informal urban settlements, enabling a tangible view of poverty and slum formation in the Municipality of São Paulo. Social and Environmental Contributions – As a social contribution to the business sector and the public sector, the technical survey presents an overview of slum development and informal urban settlements, demonstrating the precariousness of peripheral and highly socially vulnerable regions. This report allows us to identify recurring problems in each social stratum. The data gathered contributes to the adoption of public policies and investments, and enables the recognition of the most vulnerable communities.

The Dupire formula plays a significant role in pricing financial derivatives. This paper is devoted to deriving a generalized version of the Dupire formula for asset exchange options and its mathematically rigorous proofs. Financial derivatives of this type are not special cases of plain vanilla options. Moreover, the application of Lévy-type stochastic integrals in place of Itô models used in Nowak and Gatarek [Application of Itô processes and Schwartz distributions to local volatility for Margrabe options. Stochastics. 2022;94(6):807–832] allows taking into account jumps in price of the considered underlying assets. Therefore, the generalization is useful in practice and essential from a theoretical point of view. Our approach combines methods of mathematical finance, stochastic analysis and the theory of Schwartz distributions to prove the aforementioned generalized formula in the space of distributions. To illustrate the main theoretical result, we present an example of its application.

Innovation is the core driving force behind the high-quality development and global competitiveness of cities. The spatial distribution characteristics and influencing factors of innovation are of significant value in optimizing the allocation of innovation resources. This study focuses on Harbin, utilizing 46,057 patent application data from 2004 to 2022, and employs spatial analysis methods such as KDE, ESDA, and DBSCAN to analyze the spatiotemporal evolution patterns of innovation space. Through the MGWR model, this study investigates the factors influencing the spatial agglomeration of innovation from five dimensions: innovation subjects, infrastructure support, public service support, population and market, and spatial carrier. The findings are as follows: (1) In terms of spatiotemporal patterns, Harbin’s innovation space distribution shows a “one core, four cluster” spatial structure, with innovation activities clustering primarily around universities as the core and new districts as key support, gradually evolving toward a multi-center agglomeration development model. (2) Regarding influencing factors, the geographic proximity of universities is significantly associated with higher innovation density. Among infrastructure, metro stations show a positive association with innovation agglomeration, potentially reflecting the role of transport accessibility. Public service support displays significant spatial heterogeneity across different regions. (3) The Historical Area requires policy guidance for urban renewal; the New Town Area forms exogenous-driven innovation clusters; the Industrialized Park Area achieves endogenous development through industrial collaboration.

Abstract We conduct a systematic robustness analysis of the hybrid machine learning framework USmorph , which integrates unsupervised and supervised learning for galaxy morphological classification. Although USmorph has already been applied to nearly 100,000 I -band galaxy images in the COSMOS field (0.2 < z < 1.2, I mag < 25), the stability of its core modules has not been quantitatively assessed. Our tests show that the convolutional autoencoder achieves the best performance in preserving structural information when adopting an intermediate network depth, 5 × 5 convolutional kernels, and a 40D latent representation. The adaptive polar coordinate transform effectively enhances rotational invariance and improves the robustness of downstream tasks. In the unsupervised stage, a bagging clustering number of K = 50 provides the optimal trade-off between classification granularity and labeling efficiency. For supervised learning, we employ GoogLeNet, which exhibits stable performance without overfitting. We validate the reliability of the final classifications through two independent tests: (1) the t-distributed stochastic neighbor embedding visualization reveals clear clustering boundaries in the low-dimensional space; and (2) the morphological classifications are consistent with theoretical expectations of galaxy evolution, with both true and false positives showing unbiased distributions in the parameter space. These results demonstrate the strong robustness of the USmorph algorithm, providing guidance for its future application to the China Space Station Telescope mission.

Abstract Policymakers are increasingly recognizing the necessity for the equitable distribution of urban green spaces across sociodemographic groups. Nevertheless, the extant research consistently highlights disparities in this allocation, frequently neglecting the critical implications of extreme droughts on vegetation-related inequalities. Leveraging satellite observations in conjunction with census data, we examine the drought-related disparities in urban vegetation degradation (UVD) and the corresponding heat exposure across sociodemographic groups within major U.S. cities. Our findings reveal that marginalized communities experience more severe UVD during droughts, a trend particularly pronounced in Sunbelt U.S. cities, such as those in Southern California and Texas. Additionally, the unequal UVD during drought exacerbates existing inequalities in heat exposure. These results highlight the urgent need for the implementation of targeted policies, including effective water supply management strategies. Such measures could mitigate thermal environmental injustices and promote equitable vegetation distribution under a warming climate.