Search Results (2,743)

Large imaging and spectroscopic surveys now produce heterogeneous data at a scale that challenges supervised approaches which depend on scarce labels and task-specific retraining. In this paper, we conduct a systematic evaluation and analysis of AstroCLIP, a cross-modal self-supervised model that aligns galaxy images and optical spectra within a shared embedding space. Our overarching aim is to extend the released benchmark with a more fine-grained assessment of robustness and embedding behaviour. Using the released DESI and DESI Legacy Imaging Survey evaluation suite, we first reproduce the main downstream galaxy-property regression results and then extend the evaluation in two novel ways: (i) by stratifying predictive performance across a neighbour-count proxy for local environment density, and (ii) by comparing the suite’s observational categories labelled Low-z (Bright) and High-z (Faint). We further inspect the embedding space using UMAP and unsupervised clustering, and quantify cluster–property agreement using the adjusted mutual information (AMI). Across tasks, spectral embeddings consistently outperform image embeddings; for example, zero-shot prediction reaches $R^2=0.87$ for $log\left(M_{*}\right)$ and $R^2=0.63$ for $log\left(\mathrm{sSFR}\right)$. Under our environment proxy, moderate-density bins often yield the strongest predictive performance, while very sparse or crowded bins tend to underperform. Image-based predictions benefit substantially from the Low-z (Bright) subset, whereas spectral embeddings are more stable across the observational split. At the same time, UMAP and clustering reveal only weak discrete separation by individual physical properties, so the results are most consistent with useful information being encoded in a largely continuous rather than sharply clustered form. Full article

To address the complex challenge of simultaneously optimizing the operation time, motion impact, and energy consumption in industrial robotic arm trajectory planning, this study proposes a novel multi-objective optimization framework based on an improved multi-objective golden jackal optimization (IMGJO) algorithm. Firstly, the original single-objective Golden Jackal Optimization is extended into a multi-objective formulation by integrating an external Pareto archive and a crowding distance sorting mechanism. This extension effectively generates a well-distributed and highly convergent Pareto-optimal solution set. Secondly, to enhance global exploration capabilities and improve convergence stability, the escape energy model is refined. This is achieved through the synergistic integration of three key strategies: tent chaotic mapping for enhancing the initial population diversity, opposition-based learning to accelerate the early-stage search process, and an elitism preservation strategy to prevent premature convergence and mitigate the risk of entrapment in local optima. Thirdly, the IMGJO algorithm is integrated with a 3-5-3 polynomial interpolation scheme to establish a kinematically constrained trajectory planning model, ensuring a generation of smooth, continuous, and dynamically feasible joint space trajectories. Finally, comprehensive comparative experiments against several state-of-the-art benchmark algorithms demonstrate that the proposed IMGJO framework significantly outperforms its counterparts in terms of both convergence speed and the quality of the Pareto solution set. Furthermore, experimental validation on the Yaskawa HP-20D robotic arm platform demonstrates that the proposed method can effectively achieve a comprehensive optimization of execution time, impact, and energy consumption. Compared with the pre-optimization trajectory, the total operation time is reduced by 2.42%; the impacts of Joint 1 and Joint 2 are reduced by 74.65% and 75.82%, respectively; and the energy consumption of Joint 1 and Joint 2 are reduced by 27.11% and 26.83%, respectively. Moreover, the generated trajectory is smooth and continuous, thereby significantly improving the operational efficiency and stability of the robotic arm. Full article

Cognitive models of depression posit a core role for attentional biases, though empirical evidence remains inconsistent, likely due to variations in task demands. This study utilized eye-tracking to assess attentional patterns in individuals with depressive symptoms during a goal-directed visual search task, specifically dissociating early orienting and late disengagement. Seventy-seven participants, classified into high (HD) and low (LD) depressive-symptom groups based on PHQ-9 scores, completed a “face-in-the-crowd” (FITC) task. The set size (4, 8, or 12 faces) was varied to examine the role of perceptual load. The task involved searching for a single emotional target among neutral distractors (assessing early orienting) and searching for a single neutral target among emotional distractors (assessing late disengagement). Contrary to the negativity-bias hypothesis, the HD group demonstrated what might be interpreted as adaptive attentional regulation. During early orienting (8-face condition), the HD group showed reduced total dwell time on happy targets, suggesting accelerated identification. An attentional bias index (sad minus happy dwell time) correlated positively with depression severity. During late disengagement (8-face condition), the HD group exhibited shorter target fixation latency specifically with sad distractors, indicating facilitated disengagement from negative information. The corresponding bias index correlated negatively with depression levels. Under explicit goal-directed demands, individuals with high depressive symptoms displayed facilitated processing of happy faces and accelerated disengagement from sad faces, rather than an enhanced negativity bias. This pattern tentatively suggests a possible adaptive attentional regulatory mechanism in early depression, although the findings were limited to the 8-face condition and no significant group differences emerged at set sizes 4 or 12. Replication is required before firm conclusions can be drawn. The result underscores the critical influence of task demands and highlights the value of early identification and targeted intervention. Full article

This study systematically examines the impacts of extreme temperatures on the digital economy development index and the underlying mechanisms based on panel data from 281 prefecture-level cities in China from 2012 to 2023. This study explicitly distinguishes the distinctive adaptive capacity of the digital economy in responding to climate risks. Through global and local spatial autocorrelation analysis, the study finds that both extreme temperatures and the digital economy exhibit significant spatial clustering. This study employs the spatial Durbin model (SDM) and effect decomposition and further incorporates the GS2SLS estimator alongside dual instrumental variables constructed from historical geographic characteristics to address endogeneity, thereby identifying the asymmetrical impacts of extreme heat and extreme cold on the digital economy with great rigor. Specifically, extreme heat fosters short-term local digital demand that is subsequently translated into long-term growth in IT human capital and infrastructure, thereby increasing the DEDI. However, its net spatial effect is inhibitory due to energy crowding out. Extreme cold, by contrast, primarily disrupts supply chains and intensifies energy consumption, with its impact largely confined to the local scope. Green technological innovation mitigates the impact of extreme heat on the digital economy through demand substitution, while, under extreme cold, it manifests as the physical protection of infrastructure. Meanwhile, an optimized industrial structure substantially reduces the economy’s dependence on supply chains, amplifying the promotional effect of extreme temperatures on the digital economy and reflecting the transformation capacity of regions under complex environmental conditions. Heterogeneity analysis demonstrates that the effects of extreme temperatures vary significantly across different urban agglomerations, economic zones, geographic regions and city types. This study not only extends the theoretical framework for the economic assessment of climate risks and spatial econometric analysis to the climate sensitivity of the digital economy but also provides empirical evidence for understanding the complex relationship between climate change and digital economy development and offers references for differentiated policies in a coordinated regional digital economy. Full article

COVID-19 was a prolonged public-health shock that disrupted mobility, access to services, and household spending. Although the official U.S. poverty rate declined to 11.1%, the Supplemental Poverty Measure rose to 12.9%, suggesting that material hardship persisted unevenly across places. This study asks whether pre-existing livelihood vulnerability and local epidemic burden translated into geographically concentrated consumption losses during 2020–2022. Because sustained consumption loss can erode households’ health-related spending, tracking where spending declines concentrate helps connect local social and environmental conditions to how communities withstand a health crisis. We analyze consumer expenditure, unlike prior research relying on aggregate retail sales, to capture fine-grained economic strains as a proxy for shock-absorption capacity. A Livelihood Vulnerability Index (LVI) was calculated for each U.S. county using 16 socio-economic variables, and counties were classified as high- or low-risk. A multilevel model then examined how socio-economic and COVID-19 factors at county and census tract levels shaped consumption changes. Higher-risk communities experienced greater consumption reductions. At the census tract level, the non-White ratio, vacancy rate, built year, per capita income, education level, and housing value were significant. At the county level, COVID-19 cases and deaths, crowding, public transportation use, and vehicle availability mattered most. These findings support place-targeted strategies that combine public-health response with socio-environmental interventions to reduce disparities rooted in pre-existing vulnerability. Full article

3D multi-object tracking (MOT) for autonomous driving remains challenging due to frequent identity switches in crowded scenes, trajectory fragmentation during occlusions, and the difficulty of adapting association strategies to varying scene complexities. While existing methods rely on fixed geometric or appearance-based associations, they struggle to handle ambiguous cases and detection failures. We present an adaptive multi-level 3D MOT framework that achieves robust tracking through three key innovations: (1) multi-granularity temporal modeling that captures both fine-grained short-term motion and coarse long-term trends via dual-scale spatio-temporal attention, enabling accurate motion prediction across different object dynamics; (2) Transformer-based Appearance Association that employs cross-attention to model global inter-object relationships, resolving ambiguous associations in crowded scenarios where geometric cues alone fail; and (3) scene-adaptive learned thresholds that automatically adjust association strictness based on object density, motion complexity, and occlusion levels, avoiding the one-size-fits-all limitations of fixed thresholds. Our hierarchical four-level tracking strategy progressively handles cases from easy geometric matching (Level 1) to complex interval-frame recovery (Level 4), with SOT-based virtual detection generation bridging detector failures. Extensive experiments on the nuScenes benchmark demonstrate state-of-the-art performance. Full article

With the rapid advancement of autonomous driving in modern society, dense pedestrian detection technology has encountered performance bottlenecks. To address this, we propose a robust and lightweight pedestrian detection algorithm, RMP-YOLO, designed to efficiently detect small, occluded, and low-light objects. Firstly, RFAConv is utilized as the core component of the backbone network, combining standard convolution with attention mechanisms and using group convolution to extract features from the spatial receptive field. Secondly, MobileViTv3 is introduced into the backbone to combine CNNs with Transformers. The model is further enhanced by adjusting feature fusion, introducing residual connections, and optimizing local representation with deep convolutional layers. Finally, the PIoUv2 loss function is employed for bounding-box regression, significantly reducing detection errors for small-scale pedestrians in crowded environments. Experimental results demonstrate that RMP-YOLO improves mAP@0.5 by 1.3% on a custom dataset and 0.91% on the WiderPerson dataset. Crucially, it maintains high efficiency with only 3.71 million parameters and 6.29 GFLOPs, meeting the deployment requirements for low computational power and high precision. Full article

Bidirectional power converters generate significant heat losses during high-frequency operation, posing a severe challenge to the performance of heat dissipation systems. Traditional thermal design methods often struggle to balance multiple objectives, such as cooling efficiency, cost, weight, and size, thereby limiting the reliability and safety of the system. To address these challenges, this paper proposes a novel Multi-Objective Exponential Distribution Optimizer algorithm based on the Exponential Distribution Optimizer. Subsequently, key design variables of the heat dissipation system are selected. Next, the Optimal Latin Hypercube Sampling method is employed to generate sample points, and a second-order response surface surrogate model for the heat pipe radiator’s volume and temperature is developed. Lastly, by integrating elite non-dominated sorting, crowding distance mechanisms, and an information feedback mechanism, the multi-objective challenge is decomposed into subproblems, thereby enhancing optimization efficiency. Through comparative simulation experiments on benchmark functions, the Wilcoxon signed-rank test results for the MOEDO algorithm on the majority of the three metrics are denoted as ‘+’, indicating statistically significant advantages over the compared algorithms, thereby demonstrating its superior performance in addressing multi-objective optimization problems. The study further conducts simulation verification of the heat pipe heat dissipation system before and after optimization using ANSYS Icepak. The simulation results demonstrate that, compared with the conventional design, the maximum Insulated Gate Bipolar Transistor (IGBT) temperature is reduced by 17.12% and the heat sink volume is reduced by 14.61%. Full article

Pedestrian trajectory prediction is a fundamental task in autonomous driving and mobile robotics, where accurate forecasting requires modeling of both social interactions and scene-related constraints. However, existing methods typically rely on a fixed interaction modeling strategy, which may be insufficient under heterogeneous crowd densities. To address this limitation, we propose DAGH-Net, a density-adaptive gated hybrid network for pedestrian trajectory prediction. Built upon an SR-LSTM (State Refinement for LSTM) backbone, the proposed framework integrates two complementary reasoning pathways: a data-driven social interaction branch and a hybrid knowledge graph branch that encodes structured relational priors among pedestrians, obstacles, and walkable regions. A local-density-conditioned gating mechanism is further introduced to adaptively fuse these features according to the surrounding crowd condition of each pedestrian. This design helps suppress redundant interaction cues in sparse settings while strengthening socially compliant and scene-consistent reasoning in dense or conflict-prone environments. Experimental results on the ETH (Eidgenössische Technische Hochschule Zürich) and UCY (University of Cyprus) benchmarks, evaluated using Mean Average Displacement (MAD) and Final Average Displacement (FAD), show that DAGH-Net improves the average MAD and FAD by 1.6% and 4.2%, respectively, compared with SR-LSTM. Ablation studies further support the complementary contributions of the hybrid knowledge graph and the density-adaptive gating mechanism. We also discuss the limitations of the current density formulation and benchmark scale, which suggest several directions for future improvement. Full article

Human error in evacuation guidance systems can significantly affect evacuation performance, particularly in complex underground environments where large numbers of occupants are concentrated. While previous studies have focused on optimizing evacuation routes and modeling crowd dynamics, the direct quantitative impact of human error in evacuation guidance has not been sufficiently addressed. This study aims to evaluate the effects of human error on evacuation efficiency in underground stations using a node–link-based evacuation model. A virtual three-level underground station was modeled, and evacuation simulations were conducted using two representative pathfinding algorithms, Dijkstra and A*, to compare classical and heuristic routing approaches under both normal and error conditions. Three scenarios were considered: a normal condition with accurate guidance, a misguidance scenario with incorrect information on exit availability, and a delayed evacuation scenario in which a subset of evacuees started evacuation later than others. In addition, congestion effects were incorporated by adjusting walking speeds based on crowd density. The results show that human error significantly increases evacuation time and alters congestion patterns. Compared to the normal condition, the misguidance scenario increased evacuation time by approximately 17.6%, while the delayed evacuation scenario resulted in an increase of up to 37.9%, indicating that delayed response has the most critical impact due to the interaction between late-starting evacuees and existing congestion. Although the A* algorithm demonstrated higher computational efficiency, its advantage did not consistently translate into improved evacuation performance under dynamic conditions. These findings highlight that evacuation performance is highly sensitive to the accuracy and timing of evacuation guidance, rather than being determined solely by optimal pathfinding. Therefore, improving the reliability and timeliness of evacuation guidance systems is essential for enhancing safety in underground environments. Full article

The European Union’s Carbon Border Adjustment Mechanism (CBAM), the world’s first system to impose tariffs on the carbon emissions of imported products, commenced its transition period in October 2023 and is scheduled for full implementation in January 2026. This mechanism exerts a profound impact on the global trade landscape and corporate environmental management practices. Taking the CSI All Share Index constituent companies as a research sample, this paper empirically evaluates the impact of the CBAM transition period on the environmental scores of Chinese export enterprises utilizing the Propensity Score Matching Difference-in-Differences (PSM-DID) method. The results indicate that the CBAM transition period significantly inhibits the short-term environmental performance of regulated enterprises. Mechanism analysis reveals that increased financing constraints serve as a core mediating channel, wherein escalated compliance costs and compressed cash flows crowd out resources for low-carbon investments. Furthermore, heterogeneity analysis demonstrates that the negative impact is more pronounced among state-owned enterprises, firms with lower audit quality, and firms with a higher proportion of female executives. Accordingly, the study recommends establishing targeted green transition financing mechanisms, accelerating domestic carbon market reforms, and strengthening international technical harmonization to build corporate resilience against global climate governance shocks and promote sustainable growth. Full article

Against the backdrop of China’s “dual carbon” goals, investigating whether market-based environmental regulations can effectively induce technological upgrading is critical for achieving a sustainable low-carbon transition. This study adopts a staggered difference-in-differences (DID) approach within a two-way fixed-effects framework, supplemented by propensity score matching (PSM-DID), to identify the causal impact of the carbon emission trading (CET) pilot policy. The research utilizes a comprehensive panel dataset of A-share listed companies in heavy-polluting industries from 2010 to 2024, incorporating IPC-matched green patent application data to provide a granular assessment of corporate innovation performance. The empirical findings reveal a structural divergence: while the CET policy promotes green innovation in state-owned enterprises (SOEs), it exhibits a potential “crowding-out” effect on private enterprises, a relationship further explained by the mechanisms of carbon information disclosure and financing constraints. These results suggest that the “Porter Effect” in emerging markets is highly conditional on institutional resource endowments, implying that policymakers must complement market incentives with differentiated financial support and enhanced transparency standards to foster a more equitable innovation ecosystem. Full article

In air jet looms, relay nozzles are critical components in governing airflow velocity and air consumption during the weft insertion process. Although computational fluid dynamics (CFD) offers high-fidelity simulation for aerodynamic analysis, its computational burden hinders its practicality in iterative aerodynamic design of relay nozzles. To address the challenge, this study proposes a data-driven framework integrating a Chebyshev polynomial Kolmogorov–Arnold Network (Chebyshev KAN) surrogate model with an Improved Multi-objective Red-billed Blue Magpie Optimizer (IMORBMO). The accuracy of the Chebyshev KAN model was benchmarked against conventional multilayer perceptrons (MLP), convolutional neural networks (CNN), and the standard Kolmogorov–Arnold Network (KAN). Experimental results demonstrate that the Chebyshev KAN model achieves the lowest mean absolute error (MAE) of 0.103 for airflow velocity and 0.115 for air consumption. Building upon the non-dominated sorting and crowding distance strategies, IMORBMO was developed, incorporating an adaptive mutation mechanism by information entropy for improvement of convergence, diversity, and uniformity of the Pareto-optimal solutions. Comprehensive evaluations on the ZDT and WFG benchmark suites confirm that the IMORBMO consistently attains the best and highly competitive performance, yielding the lowest generation distance (GD), inverted generational distance (IGD) values and the highest hypervolume (HV). Applied to the aerodynamic optimization of a relay nozzle, the proposed framework delivers an optimal aerodynamic design that increases airflow velocity by 10.5% while reducing air consumption by 15.4%, as verified by CFD simulation. The steady-state flow field was simulated by solving the Reynolds-Average NavierStokes equations with the k–ω turbulent model, utilizing Fluent 2025.R2. No-slip wall, inlet pressure and outlet pressures are boundary conditions to the relay nozzle surfaces. This work establishes a computationally efficient and accurate optimization paradigm that holds significant promise for aerodynamic design and other complex real-world engineering applications. Full article

This study examines the street spaces of Shamian Island in Guangzhou and addresses the long-standing urban design challenge of quantifying subjective perception. Drawing on environmental psychology, it introduces “behavioral representation” as a proxy variable for perception. By synthesizing international street design guidelines, the study establishes a street-characteristic indicator system covering spatial scale, interface, facilities, and landscape. Multiple linear regression (MLR) models are then applied to analyze in depth how spatial elements influence five types of behavior, including lingering, passing through, and consumption. The results show that walkway width is the core driving factor across all behavior types, while artistic landscape installations exert the most significant effect on long-duration stays. In addition, different spatial elements exhibit distinct mechanisms in shaping various behaviors. The study constructs a “space–perception–behavior” cognitive framework, providing an evidence-based tool and a methodological reference for evaluating subjective perception in urban design. Full article

This paper examines whether government debt financing crowds out private consumption in South Africa or whether household behaviour is consistent with Ricardian equivalence. Using quarterly data from 1960Q1 to 2025Q1, the study employs a Bayesian time-varying parameter framework that accommodates non-stationarity, structural change, and evolving fiscal transmission mechanisms, and is complemented by a Markov-switching Bayesian VAR as a robustness check. All variables are expressed relative to GDP to avoid scale effects, and inference is based on posterior distributions. The results reveal pronounced state dependence in the debt–consumption relationship. In earlier decades, increases in the debt-to-GDP ratio are associated with statistically meaningful declines in the private consumption share, consistent with crowding-out or precautionary behaviour under weaker fiscal credibility. Over time, however, this negative association weakens and converges toward neutrality, with post-2010 estimates indicating no significant effect of debt on consumption. Conditioning on fiscal stance and financial conditions shows that debt does not exert an independent influence on consumption once government expenditure, tax revenue, and interest rates are taken into account. A constant-parameter Bayesian benchmark masks these dynamics, producing an average effect close to zero. Evidence from a Markov-switching Bayesian VAR similarly finds no persistent regime-specific crowding-out effects. Overall, the findings suggest that observed debt–consumption linkages in South Africa operate primarily through broader fiscal and macroeconomic conditions rather than debt accumulation itself, highlighting the importance of fiscal credibility and policy composition. Full article