Search for: All records

Many deep generative models, such as variational autoencoders (VAEs) and generative adversarial networks (GANs), learn an immersion mapping from a standard normal distribution in a low-dimensional latent space into a higher-dimensional data space. As such, these mappings are only capable of producing simple data topologies, i.e., those equivalent to an immersion of Euclidean space. In this work, we demonstrate the limitations of such latent space generative models when trained on data distributions with non-trivial topologies. We do this by training these models on synthetic image datasets with known topologies (spheres, torii, etc.). We then show how this results in failures of both data generation as well as data interpolation. Next, we compare this behavior to two classes of deep generative models that in principle allow for more complex data topologies. First, we look at chart autoencoders (CAEs), which construct a smooth data manifold from multiple latent space chart mappings. Second, we explore score-based models, e.g., denoising diffusion probabilistic models, which estimate gradients of the data distribution without resorting to an explicit mapping to a latent space. Our results show that these models do demonstrate improved ability over latent space models in modeling data distributions with complex topologies, however, challenges still remain. 

Abstract A new measurement protocol, labeled Acoustic Mapping Velocimetry (AMV), has been successfully tested for in‐situ estimation of bedload transport features in sandy beds. The AMV has proven efficient in using the dune‐tracking method (DTM) for characterizing the bedform geometry and dynamics as well as for estimation of the rates of bedload transport. Given the novelty of the AMV protocol and its extensive reliance on multiple site‐specific assumptions and user‐defined parameters, a comparison of this emerging technique with other three non‐intrusive DTM‐based methods and analytical predictors is attempted in this paper. The comparison highlights that the AMV estimates are within 22% of the estimates with the other non‐intrusive protocols and up to 98% different from analytical predictions. The observed differences are related to the possible sources of uncertainty in the AMV workflows and to the means to reduce their impact on the targeted estimations. 

Abstract Previous studies have considered floral humidity to be an inadvertent consequence of nectar evaporation, which could be exploited as a cue by nectar-seeking pollinators. By contrast, our interdisciplinary study of a night-blooming flower, Datura wrightii , and its hawkmoth pollinator, Manduca sexta , reveals that floral relative humidity acts as a mutually beneficial signal in this system. The distinction between cue- and signal-based functions is illustrated by three experimental findings. First, floral humidity gradients in Datura are nearly ten-fold greater than those reported for other species, and result from active (stomatal conductance) rather than passive (nectar evaporation) processes. These humidity gradients are sustained in the face of wind and are reconstituted within seconds of moth visitation, implying substantial physiological costs to these desert plants. Second, the water balance costs in Datura are compensated through increased visitation by Manduca moths, with concomitant increases in pollen export. We show that moths are innately attracted to humid flowers, even when floral humidity and nectar rewards are experimentally decoupled. Moreover, moths can track minute changes in humidity via antennal hygrosensory sensilla but fail to do so when these sensilla are experimentally occluded. Third, their preference for humid flowers benefits hawkmoths by reducing the energetic costs of flower handling during nectar foraging. Taken together, these findings suggest that floral humidity may function as a signal mediating the final stages of floral choice by hawkmoths, complementing the attractive functions of visual and olfactory signals beyond the floral threshold in this nocturnal plant-pollinator system. 

Abstract BackgroundRecent reports of extreme levels of undersaturation in internal leaf air spaces have called into question one of the foundational assumptions of leaf gas exchange analysis, that leaf air spaces are effectively saturated with water vapour at leaf surface temperature. Historically, inferring the biophysical states controlling assimilation and transpiration from the fluxes directly measured by gas exchange systems has presented a number of challenges, including: (1) a mismatch in scales between the area of flux measurement, the biochemical cellular scale and the meso-scale introduced by the localization of the fluxes to stomatal pores; (2) the inaccessibility of the internal states of CO2 and water vapour required to define conductances; and (3) uncertainties about the pathways these internal fluxes travel. In response, plant physiologists have adopted a set of simplifying assumptions that define phenomenological concepts such as stomatal and mesophyll conductances. ScopeInvestigators have long been concerned that a failure of basic assumptions could be distorting our understanding of these phenomenological conductances, and the biophysical states inside leaves. Here we review these assumptions and historical efforts to test them. We then explore whether artefacts in analysis arising from the averaging of fluxes over macroscopic leaf areas could provide alternative explanations for some part, if not all, of reported extreme states of undersaturation. ConclusionsSpatial heterogeneities can, in some cases, create the appearance of undersaturation in the internal air spaces of leaves. Further refinement of experimental approaches will be required to separate undersaturation from the effects of spatial variations in fluxes or conductances. Novel combinations of current and emerging technologies hold promise for meeting this challenge. 

Abstract The HAWC Observatory collected 6 yr of extensive data, providing an ideal platform for long-term monitoring of blazars in the very high energy (VHE) band, without bias toward specific flux states. HAWC continuously monitors blazar activity at TeV energies, focusing on sources with a redshift ofz≤ 0.3, based on the Third Fermi-LAT Catalog of High-Energy sources. We specifically focused our analysis on Mrk 421 and Mrk 501, as they are the brightest blazars observed by the HAWC Observatory. With a data set of 2143 days, this work significantly extends the monitoring previously published, which was based on 511 days of observation. By utilizing HAWC data for the VHEγ-ray emission in the 300 GeV–100 TeV energy range, in conjunction with Swift-XRT data for the 0.3–10 keV X-ray emission, we aim to explore potential correlations between these two bands. For Mrk 501, we found evidence of a long-term correlation. Additionally, we identified a period in the light curve where the flux was very low for more than 2 yr. On the other hand, our analysis of Mrk 421 measured a strong linear correlation for quasi-simultaneous observations collected by HAWC and Swift-XRT. This result is consistent with a linear dependence and a multiple-zone synchrotron self-Compton model to explain the X-ray andγ-ray emission. Finally, as suggested by previous findings, we confirm a harder-when-brighter behavior in the spectral evolution of the flux properties for Mrk 421. These findings contribute to the understanding of blazar emissions and their underlying mechanisms. 

Abstract Working memory, the ability to hold items in memory stores for further manipulation, is a higher order cognitive process that supports many aspects of daily life. Childhood trauma has been associated with altered cognitive development including particular deficits in verbal working memory (VWM), but the neural underpinnings remain poorly understood. Magnetoencephalography (MEG) studies of VWM have reliably shown decreased alpha activity in left-lateralized language regions during encoding, and increased alpha activity in parieto-occipital cortices during the maintenance phase. In this study, we examined whether childhood trauma affects behavioral performance and the oscillatory dynamics serving VWM using MEG in a cohort of 9- to 15-year-old youth. All participants completed a modified version of the UCLA Trauma History Profile and then performed a VWM task during MEG. Our findings indicated a sex-by-age-by-trauma three-way interaction, whereby younger females experiencing higher levels of trauma had the lowest d’ accuracy scores and the strongest positive correlations with age (i.e. older performed better). Likewise, females with higher levels of childhood trauma exhibited altered age-related alpha changes during the maintenance phase within the right temporal and parietal cortices. These findings suggest that trauma exposure may alter the developmental trajectory of neural oscillations serving VWM processing in a sex-specific way. 

Background: Healthcare workers are at the forefront of the COVID-19 pandemic and it is essential to monitor the relative incidence rate of this group, as compared to workers in other occupations. This study aimed to produce estimates of the relative incidence ratio between healthcare workers and workers in non-healthcare occupations. Methods: Analysis of cross-sectional data from a daily, web-based survey of 1,822,662 Facebook users from September 8, 2020 to October 20, 2020. Participants were Facebook users in the United States aged 18 and above who were tested for COVID-19 because of an employer or school requirement in the past 14 days. The exposure variable was a self-reported history of working in healthcare in the past four weeks and the main outcome was a self-reported positive test for COVID-19. Results: On October 20, 2020, in the United States, there was a relative COVID-19 incidence ratio of 0.73 (95% UI 0.68 to 0.80) between healthcare workers and workers in non-healthcare occupations. Conclusions: In fall of 2020, in the United States, healthcare workers likely had a lower COVID-19 incidence rate than workers in non-healthcare occupations.