37495 research outputs found
Sort by
Spectral–Spatial Transformer With Multiscale Convolutional Attention for Hyperspectral Image Classification
Get PDFHyperspectral image (HSI) classification plays a vital role in remote sensing by leveraging rich spectral and spatial information for accurate material recognition. However, existing methods, particularly Transformer-based approaches, still face challenges in effectively modeling multiscale spatial–spectral features, preserving local details, and maintaining robustness to noise. To mitigate these limitations, we propose TMCANet, a spectral–spatial Transformer with multiscale convolutional attention, designed to effectively leverage both local and global contextual dependencies for HSI classification. Our design is guided by three core strategies: first, a convolutional feature extraction module, consisting of four convolutional layers, to learn hierarchical spectral multiscale representations and enhance local feature capture; second, a hybrid spatial–spectral cross-fusion attention block, combining local spatial attention, and spectral tokenization attention, to dynamically integrate spatial textures and spectral dependencies, and third a spectral–spatial tokenization transformer equipped with an adaptive cross-layer fusion mechanism to aggregate multilevel features, reduce redundancy, and strengthen global contextual modeling. Furthermore, an enhanced focal loss is adopted to alleviate class imbalance and improve multiclass classification robustness. We validate TMCANet on three benchmark datasets—Indian Pines, Pavia University, and Houston—achieving overall accuracies of 94.91%, 95.49%, and 89.89%, respectively, with performance comparable to or exceeding state-of-the-art baselines. These results demonstrate that the proposed multiscale convolutional attention and hybrid spatial–spectral fusion strategies effectively preserve both local detail and global context, thereby enhancing classification performance and supporting practical applicability in remote sensing
Coordination in Time
Get PDFWe study how well people are able to solve pure coordination problems in continuous time. Subjects decide whether and when to pay a cost to go to market with their goods and earn money only if another person shows up at the same time. We show that coordination failure is common in a baseline, and we introduce treatments that feature public coordination devices (meant to mimic clocks) and assess the extent to which coordination improves when such devices are provided via different institutions. A publicly provided device outperforms a variety of privately provided alternatives. Our evidence suggests this is because reliable public provision eliminates uncertainty about whether (and how many) other people expect to observe the coordinating signal
A Feature Engineering Technique for Enhancing the Generalization of Machine Learning Models in Estimating Crop Evapotranspiration
Get PDFAccurate and precise estimation of evapotranspiration (ET) is crucial for understanding the terrestrial carbon, water, and energy cycles. While process-based models of ET, such as the Penman–Monteith model offer robust generalization capabilities, they are limited by the need for detailed parameters (e.g., stomatal conductance,) that are challenging to measure continuously. On the other hand, machine learning models can estimate ET by capturing relationships between ET and environmental variables without experimentally measuring model parameters. However, machine learning models face the challenge of limited generalizability. This issue is particularly significant given the uncertainty introduced by changing climatic conditions, which can restrict the model\u27s predictive performance when it is applied to different environmental contexts. Therefore, we propose a hybrid modeling approach that combines feature engineering using process-based models with machine learning to improve generalizability while maintaining practicality. Our model first converts environmental variables into leaf-scale ET using mechanistic process-based models and then uses these features along with the leaf area index to estimate the canopy-scale ET using an artificial neural network (ANN). We evaluated the generalization of the hybrid model against a pure ANN model using FLUXNET2015 data. Results show that the hybrid model significantly outperformed the pure ANN model, especially when tested on data beyond the range of the training dataset. Furthermore, the estimation accuracy of the hybrid model was stable even when the values of the model parameters in the process-based models used for feature engineering were varied by ±50 %. This indicates that incorporating a mechanistic understanding of plant environmental responses enhances the generalizability and robustness of ET predictions. These findings underscore the potential of hybrid models to combine the strengths of process-based and machine learning approaches
Neonatal Striatal Volume is Associated with Infant Anhedonia
No full textAnhedonia is increasingly recognized as a transdiagnostic risk factor for psychopathology. New evidence demonstrates that anhedonia is present in infancy and early childhood. Structural variability in striatal regions involved in reward processing and pleasure seeking is concurrently linked to anhedonia, yet few studies have examined whether striatal differences presage anhedonia, and none have examined prospective associations before middle childhood. The present study examined whether neuroanatomical markers that confer risk for anhedonia can be detected as early as the neonatal period. Specifically, we tested whether the volume of striatal regions in neonates predicted the emergence of infant anhedonic behaviors at six months of age. Our sample included 89 neonates (47 females, 42 males) from the Care Project - a longitudinal cohort study of pregnant individuals and their children. Neonatal striatal volume (nucleus accumbens, putamen, and caudate) was assessed using structural magnetic resonance imaging (MRI; Mage: 43 postconceptional weeks). Anhedonic behaviors were measured using the Infant Hedonic/Anhedonic Processing Index (HAPI-Infant) at six months. Larger neonatal nucleus accumbens (right: β = 0.32, p = .003; left: β = 0.22, p = .04) and putamen (right: β = 0.29, p = .03; left: β = 0.28, p = .04) volume predicted more anhedonic behaviors at six months, covarying for intracranial volume, postconceptional age at scan, sex at birth, and birth weight percentile. This study demonstrates that neonatal striatal volume is associated with anhedonia in infancy and provides the first evidence that markers of vulnerability to anhedonia can be detected in neonates
Is Academia Becoming Démodé — or Can It Reinvent Itself?
No full textAnd yet, in the wider cultural imagination, academia has become, in many ways, démodé. Out of fashion. Out of step. No longer the epicenter of intellectual authority or social innovation that it once was. The challenge before us is whether we allow academia to drift further toward irrelevance, or whether we reinvent it for the generations now walking through our doors
Comprehensive Descriptive Analysis of Large Alzheimer\u27s Disease Patient Cohorts
No full textBackground Precise estimates of the prevalence of Alzheimer\u27s disease (AD), the distribution of demographic characteristics, comorbidities, treatment plans, insurance types, cost of treatment and survival probabilities at various time points are crucially important to advancing our understanding and for improving future AD research studies. Objective We analyzed two of the largest and high-quality medical databases, Oracle EHR Real-World Data and IQVIA. The results provide the most complete description of the AD patients in the US. Methods We present high-accuracy summary statistics of many important variables related to AD patients. Proportions, means and 95% confidence intervals were provided for all levels of the categorical and quantitative variables. Results We report high accuracy estimates of the overall survival probabilities for the first five years after initial diagnosis, drug treatments and patterns of use, demographics, insurance types, hospitalization duration, number of hospital visits, and a detailed list of comorbidities. We also report estimates of the annual total average cost of treatment per patient as well as itemized allocations for drugs, hospitalizations, surgery, and management costs. Conclusions We present the most complete, detailed and high-accuracy descriptive analysis of AD patients to date
Palpation Characteristics of an Instrumented Virtual Cricothyroidotomy Simulator
No full textCricothyroidotomy (CCT) is a critical, life-saving procedure requiring the identification of key neck landmarks through palpation. Interactive virtual simulation offers a promising, cost-effective approach to CCT training with high visual realism. However, developing the palpation skills necessary for CCT requires a haptic interface with tactile sensitivity comparable to human fingers. Such interfaces are often represented by plastic partial mannequins, which require further adaptation to integrate into virtual environments. This study introduces an instrumented physical palpation interface for CCT, integrated into a virtual surgical simulator, and tested on 10 surgeons who practiced the procedure over a training period. Data on haptic interactions collected during the training was analyzed to evaluate participants’ palpation skills and explore their force modulation strategies about landmark identification scores. Our findings suggest that trainees become more precise in their exploration over time, apply greater normal forces around target areas. Initial landmark identification performance influences adjustments in the overall applied pressure
Social Media Sponsorship: Metrics and Strategies for Finding the Right Content Creator-Sponsor Matches
No full textSocial media video sponsorship, where sponsors partner with content creators to promote their brands, has become increasingly popular. Leveraging rich data on sponsored and nonsponsored videos on Facebook, we introduce two metrics, Content Similarity and Audience Closeness, to help sponsors find effective creator-sponsor matches. Content Similarity captures the thematic alignment between a creator’s content and a sponsor’s brand, whereas Audience Closeness measures how closely aligned a creator’s and sponsor’s audiences are within a social media network. We find that both metrics significantly increase video viewership, but their effects vary over time. Audience Closeness generates more immediate, short-term engagement, whereas Content Similarity drives sustained, long-term engagement. We additionally examine how congruency metrics can complement other sponsorship strategies, including partnering with more established creators, forming new partnerships, and managing the sponsor’s visibility within content. Our findings suggest that Content Similarity can mitigate the negative effects of frequent sponsor appearances and enhance viewer engagement, even in repetitive partnerships. Moreover, we find that Audience Closeness is particularly beneficial for larger, more established creators, whereas Content Similarity is less sensitive to the size of a creator’s following. Finally, we demonstrate how intermediary platforms can adopt these metrics to reduce information asymmetry and improve creator-sponsor matching, especially for long-tail creators and smaller brands
Sex-specific Associations Between Early Life Unpredictability and Trajectories of Body Mass Index from Infancy to Adolescence
No full textObjective:
Unpredictability in early life is an understudied aspect of early life adversity that provides important signals to the child about their environment. Recent work shows the predictability of patterns of parental signals on a moment-to-moment timescale sculpts the developing brain during sensitive windows, offering a potential pathway through which early life experiences shape child development. Both experimental non-human animal models and observational human research show sex-specific links between early life exposure to unpredictable patterns of parental sensory signals and aberrant offspring development. The impact of predictability of sensory signals on physical health outcomes, such as BMI, remains unknown. Here we examined the sex-specific association between unpredictable maternal sensory signals during infancy and child BMI trajectories from infancy through adolescence. Methods:
In a prospective cohort (N=190 mother/child dyads), we quantified the unpredictability of maternal sensory signals (unpredictability of transitions between visual, auditory, and tactile signals mothers provide their infant) from free-play interactions at 6- and 12-months of age. Child BMI was measured longitudinally 7 times from infancy through adolescence. General linear mixed models assessed relations of unpredictability with BMI trajectories. Results:
Greater unpredictability of maternal sensory signals in infancy was associated with higher BMI over time for females (b=0.07, SE=0.03, P=0.047 for overall trajectory difference), but not males (b=−0.03, SE=0.04, P=0.450 for overall trajectory difference). Conclusion:
Findings suggest early life unpredictability may be an important signal shaping biobehavioral pathways to later child weight status among females
Emotional Balance, Health, and Resilience at the Start of COVID-19 Pandemic
No full textSelf-organizing systems can shift between stability and flexibility in response to perturbation, a potential adaptive mechanism for understanding biopsychosocial resilience. Inverse Power-Law (IPL) structure, a frequency distribution that describes fractal patterns commonly produced by self-organization, produce measurements of stability and flexibility. This study applies these measures to emotional resilience at the start of the Covid-19 pandemic. Ratings of frequency over the past week (1-5 Likert scale) across 12 emotions (six positive and six negative) gathered in mid-April 2020 as part of a survey of adults’ (N = 4,094) pandemic experiences and health in the USA. The distributions of everyone’s emotion ratings were tested for IPL fit, resulting in a mean R2 = .75. A steeper IPL shape parameter, reflecting greater emotional stability, was associated with better mental (anxiety, depression, and stress) and physical (fatigue, headache, and diarrhea) health overall. However, when total scores for positive and negative emotion were controlled, the reverse effect was found. Finally, a significant interaction effect was found between a measure of COVID-19 impact and IPL shape on each of the six health outcomes, suggesting that greater emotional flexibility may provide buffering against large-scale and unexpected challenges. Altogether, these results suggest that emotional stability may be most beneficial against illness when life is relatively stable, while emotional flexibility may be more adaptive when life is unstable