It is a difficult task to classify images with multiple class labels using only a small number of labeled examples, especially when the label (class) distribution is imbalanced. Emotion classification is such an example of imbalanced label distribution, because some classes of emotions like disgusted are relatively rare comparing to other labels like happy or sad. In this paper, we propose a data augmentation method using generative adversarial networks (GAN). It can complement and complete the data manifold and find better margins between neighboring classes. Specifically, we design a framework using a CNN model as the classifier and a cycle-consistent adversarial networks (CycleGAN) as the generator. In order to avoid gradient vanishing problem, we employ the least-squared loss as adversarial loss. We also propose several evaluation methods on three benchmark datasets to validate GAN's performance. Empirical results show that we can obtain 5%∼10% increase in the classification accuracy after employing the GAN-based data augmentation techniques.
A dual physical crosslinking (DPC) strategy is used to construct moldable hydrogels with ultrahigh strength and toughness. First, polyelectrolyte complex (PEC) hydrogels are prepared through the in situ polymerization of acrylic acid monomers in the concentrated chitosan (Ch) solutions. Subsequently, Ag + ions are introduced into the PEC hydrogels to form coordination bonds between NH 2 and COOH groups. High-density electrostatic interaction and coordination bonds endow the DPC hydrogels with high strength and toughness. The mechanical properties of the DPC hydrogels strongly depend on the weight ratio of Ch to poly(acrylic acid) and the loading concentration of Ag + ions. When the loading concentration of Ag + ions is in the range of 1.0-1.5 mol L −1 , DPC 0.10-0.25 hydrogels display the maximum tensile strength (24.0 MPa) and toughness (84.7 MJ m −3 ) with a strain of more than 600%. Specially, the DPC hydrogels display an excellent moldable behavior due to the reversible properties of the electrostatic interaction and coordination bonds. The DPC strategy can also be applied in various other systems and opens a new avenue to fabricate hydrogels with outstanding mechanical properties and antibacterial activities.
A novel chiral surfactant-type catalyst is developed. Micelles formed in water by association of the catalysts themselves, and this was confirmed by TEM analyses. Asymmetric transfer hydrogenation of aliphatic ketones catalyzed by the chiral metallomicellar catalyst gave good to excellent conversions and remarkable stereoselectivities (up to 95% ee). Synergistic effects between the metal-catalyzed center and the hydrophobic microenvironment of the core in the metallomicelle led to high enantioselectivities.
Our data demonstrate that ITF(l) but not ITF(s) delays the development of T1D via modulation of gut-pancreatic immunity, barrier function, and microbiota homeostasis.
Recent evidence indicates that indigenous Clostridium species induce colonic regulatory T cells (Tregs), and gut lymphocytes are able to migrate to pancreatic islets in an inflammatory environment. Thus, we speculate that supplementation with the well-characterized probiotics Clostridium butyricum CGMCC0313.1 (CB0313.1) may induce pancreatic Tregs and consequently inhibit the diabetes incidence in non-obese diabetic (NOD) mice. CB0313.1 was administered daily to female NOD mice from 3 to 45 weeks of age. The control group received an equal volume of sterile water. Fasting glucose was measured twice a week. Pyrosequencing of the gut microbiota and flow cytometry of mesenteric lymph node (MLN), pancreatic lymph node (PLN), pancreatic and splenic immune cells were performed to investigate the effect of CB0313.1 treatment. Early oral administration of CB0313.1 mitigated insulitis, delayed the onset of diabetes, and improved energy metabolic dysfunction. Protection may involve increased Tregs, rebalanced Th1/Th2/Th17 cells and changes to a less proinflammatory immunological milieu in the gut, PLN, and pancreas. An increase of α4β7+ (the gut homing receptor) Tregs in the PLN suggests that the mechanism may involve increased migration of gut-primed Tregs to the pancreas. Furthermore, 16S rRNA gene sequencing revealed that CB0313.1 enhanced the Firmicutes/Bacteroidetes ratio, enriched Clostridium-subgroups and butyrate-producing bacteria subgroups. Our results provide the basis for future clinical investigations in preventing type 1 diabetes by oral CB0313.1 administration.
Experimentally increasing atmospheric CO2 often stimulates plant growth and ecosystem carbon (C) uptake. Biogeochemical theory predicts that these initial responses will immobilize nitrogen (N) in plant biomass and soil organic matter, causing N availability to plants to decline, and reducing the long-term CO2-stimulation of C storage in N limited ecosystems. While many experiments have examined changes in N cycling in response to elevated CO2, empirical tests of this theoretical prediction are scarce. During seven years of postfire recovery in a scrub oak ecosystem, elevated CO2 initially increased plant N accumulation and plant uptake of tracer 15N, peaking after four years of CO2 enrichment. Between years four and seven, these responses to CO2 declined. Elevated CO2 also increased N and tracer 15N accumulation in the O horizon, and reduced 15N recovery in underlying mineral soil. These responses are consistent with progressive N limitation: the initial CO2 stimulation of plant growth immobilized N in plant biomass and in the O horizon, progressively reducing N availability to plants. Litterfall production (one measure of aboveground primary productivity) increased initially in response to elevated CO2, but the CO2 stimulation declined during years five through seven, concurrent with the accumulation of N in the O horizon and the apparent restriction of plant N availability. Yet, at the level of aboveground plant biomass (estimated by allometry), progressive N limitation was less apparent, initially because of increased N acquisition from soil and later because of reduced N concentration in biomass as N availability declined. Over this seven-year period, elevated CO2 caused a redistribution of N within the ecosystem, from mineral soils, to plants, to surface organic matter. In N limited ecosystems, such changes in N cycling are likely to reduce the response of plant production to elevated CO2.
Hurricane disturbances have profound impacts on ecosystem structure and function, yet their effects on ecosystem CO 2 exchange have not been reported. In September 2004, our research site on a fire-regenerated scrub-oak ecosystem in central Florida was struck by Hurricane Frances with sustained winds of 113 km h À1 and wind gusts as high as 152 km h À1 . We quantified the hurricane damage on this ecosystem resulting from defoliation: we measured net ecosystem CO 2 exchange, the damage and recovery of leaf area, and determined whether growth in elevated carbon dioxide concentration in the atmosphere (C a ) altered this disturbance. The hurricane decreased leaf area index (LAI) by 21%, which was equal to 60% of seasonal variation in canopy growth during the previous 3 years, but stem damage was negligible. The reduction in LAI led to a 22% decline in gross primary production (GPP) and a 25% decline in ecosystem respiration (R e ). The compensatory declines in GPP and R e resulted in no significant change in net ecosystem production (NEP). Refoliation began within a month after the hurricane, although this period was out of phase with the regular foliation period, and recovered 20% of the defoliation loss within 2.5 months. Full recovery of LAI, ecosystem CO 2 assimilation, and ecosystem respiration did not occur until the next growing season. Plants exposed to elevated C a did not sustain greater damage, nor did they recover faster than plants grown under ambient C a . Thus, our results indicate that hurricanes capable of causing significant defoliation with negligible damage to stems have negligible effects on NEP under current or future CO 2 -enriched environment.
This study reports the aboveground biomass response of a fire‐regenerated Florida scrub‐oak ecosystem exposed to elevated CO2 (1996–2007), from emergence after fire through canopy closure. Eleven years exposure to elevated CO2 caused a 67% increase in aboveground shoot biomass. Growth stimulation was sustained throughout the experiment; although there was significant variability between years. The absolute stimulation of aboveground biomass generally declined over time, reflecting increasing environmental limitations to long‐term growth response. Extensive defoliation caused by hurricanes in September 2004 was followed by a strong increase in shoot density in 2005 that may have resulted from reopening the canopy and relocating nitrogen from leaves to the nutrient‐poor soil. Biomass response to elevated CO2 was driven primarily by stimulation of growth of the dominant species, Quercus myrtifolia, while Quercus geminata, the other co‐dominant oak, displayed no significant CO2 response. Aboveground growth also displayed interannual variation, which was correlated with total annual rainfall. The rainfall × CO2 interaction was partially masked at the community level by species‐specific responses: elevated CO2 had an ameliorating effect on Q. myrtifolia growth under water stress. The results of this long‐term study not only show that atmospheric CO2 concentration had a consistent stimulating effect on aboveground biomass production, but also showed that available water is the primary driver of interannual variation in shoot growth and that the long‐term response to elevated CO2 may have been caused by other factors such as nutrient limitation and disturbance.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.