Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

    155 research outputs found

    Pairwise Confusion for Fine-Grained Visual Classification

    Full text link
    Fine-Grained Visual Classification (FGVC) datasets contain small sample sizes, along with significant intra-class variation and inter-class similarity. While prior work has addressed intra-class variation using localization and segmentation techniques, inter-class similarity may also affect feature learning and reduce classification performance. In this work, we address this problem using a novel optimization procedure for the end-to-end neural network training on FGVC tasks. Our procedure, called Pairwise Confusion (PC) reduces overfitting by intentionally {introducing confusion} in the activations. With PC regularization, we obtain state-of-the-art performance on six of the most widely-used FGVC datasets and demonstrate improved localization ability. {PC} is easy to implement, does not need excessive hyperparameter tuning during training, and does not add significant overhead during test time.Comment: Camera-Ready version for ECCV 201

    Validation of OMPS Suomi NPP and OMPS NOAA‐20 Formaldehyde Total Columns With NDACC FTIR Observations

    Get PDF
    We validate formaldehyde (HCHO) vertical column densities (VCDs) from Ozone Mapping and Profiler Suite Nadir Mapper (OMPS-NM) instruments onboard the Suomi National Polar-orbiting Partnership (Suomi NPP) satellite for 2012–2020 and National Oceanic and Atmospheric Administration-20 (NOAA-20) satellite for 2018–2020, hereafter referred to as OMPS-NPP and OMPS-N20, with ground-based Fourier-Transform Infrared (FTIR) observations of the Network for the Detection of Atmospheric Composition Change (NDACC). OMPS-NPP/N20 HCHO products reproduce seasonal variability at 24 FTIR sites. Monthly variability of OMPS-NPP/N20 has a very good agreement with FTIR, showing correlation coefficients of 0.83 and 0.88, respectively. OMPS-NPP (N20) biases averaged over all sites are −0.9 (4) ± 3 (6)%. However, at clean sites (with VCDs 4.0 × 1015^{15} molecules cm2^{−2}, negative biases of −15% ± 4% appear for OMPS-NPP, but OMPS-N20 shows smaller bias of 0.5% ± 6% due to its smaller ground pixel footprints. Therefore, smaller satellite footprint sizes are important in distinguishing small-scale plumes. In addition, we discuss a bias correction and provide lower limit for the monthly uncertainty of OMPS-NPP/N20 HCHO products. The total uncertainty for OMPS-NPP (N20) at clean sites is 0.7 (0.8) × 1015^{15} molecules cm2^{−2}, corresponding to a relative uncertainty of 32 (30)%. In the case of HCHO VCDs > 4.0 × 1015^{15} molecules cm2^{−2}, however, the relative uncertainty in HCHO VCDs for OMPS-NPP (N20) decreases to 31 (18)%

    Revisiting the mechanism of coagulation factor XIII activation and regulation from a structure/functional perspective

    Get PDF
    The activation and regulation of coagulation Factor XIII (FXIII) protein has been the subject of active research for the past three decades. Although discrete evidence exists on various aspects of FXIII activation and regulation a combinatorial structure/functional view in this regard is lacking. In this study, we present results of a structure/function study of the functional chain of events for FXIII. Our study shows how subtle chronological submolecular changes within calcium binding sites can bring about the detailed transformation of the zymogenic FXIII to its activated form especially in the context of FXIIIA and FXIIIB subunit interactions. We demonstrate what aspects of FXIII are important for the stabilization (first calcium binding site) of its zymogenic form and the possible modes of deactivation (thrombin mediated secondary cleavage) of the activated form. Our study for the first time provides a structural outlook of the FXIIIA 2 B 2 heterotetramer assembly, its association and dissociation. The FXIIIB subunits regulatory role in the overall process has also been elaborated upon. In summary, this study provides detailed structural insight into the mechanisms of FXIII activation and regulation that can be used as a template for the development of future highly specific therapeutic inhibitors targeting FXIII in pathological conditions like thrombosis

    The role of acyl-coenzyme A carboxylase complex in lipstatin biosynthesis of Streptomyces toxytricini

    Get PDF
    Streptomyces toxytricini produces lipstatin, a specific inhibitor of pancreatic lipase, which is derived from two fatty acid moieties with eight and 14 carbon atoms. The pccB gene locus in 10.6 kb fragment of S. toxytricini chromosomal DNA contains three genes for acyl-coenzyme A carboxylase (ACCase) complex accA3, pccB, and pccE that are presumed to be involved in secondary metabolism. The pccB gene encoding a β subunit of ACCase [carboxyltransferase (CT)] was identified upstream of pccE gene for a small protein of ε subunit. The accA3 encoding the α subunit of ACCase [biotin carboxylase (BC)] was also identified downstream of pccB gene. When the pccB and pccE genes were inactivated by homologous recombination, the lipstatin production was reduced as much as 80%. In contrast, the accumulation of another compound, tetradeca-5.8-dienoic acid (the major lipstatin precursor), was 4.5-fold increased in disruptant compared with wild-type. It implies that PccB of S. toxytricini is involved in the activation of octanoic acid to hexylmalonic acid for lipstatin biosynthesis

    Climate-smart agriculture practices for mitigating greenhouse gas emissions

    Get PDF
    Agricultural lands make up approximately 37% of the global land surface, and agriculture is a significant source of greenhouse gas (GHG) emissions, including carbon dioxide (CO2), methane (CH4) and nitrous oxide (N2O). Those GHGs are responsible for the majority of the anthropogenic globalwarming effect.Agricultural GHG emissions are associated with agricultural soil management (e.g. tillage), use of both synthetic and organic fertilisers, livestock management, burning of fossil fuel for agricultural operations, and burning of agricultural residues and land use change. When natural ecosystems such as grasslands are converted to agricultural production, 20-40% of the soil organic carbon (SOC) is lost over time, following cultivation.We thus need to develop management practices that can maintain or even increase SOC storage in and reduce GHG emissions from agricultural ecosystems. We need to design systematic approaches and agricultural strategies that can ensure sustainable food production under predicted climate change scenarios, approaches that are being called climate-smart agriculture (CSA). Climate-smart agricultural management practices, including conservation tillage, use of cover crops and biochar application to agricultural fields, and strategic application of synthetic and organic fertilisers have been considered a way to reduce GHG emission from agriculture. Agricultural management practices can be improved to decreasing disturbance to the soil by decreasing the frequency and extent of cultivation as a way to minimise soil C loss and/or to increase soil C storage. Fertiliser nitrogen (N) use efficiency can be improved to reduce fertilizer N application and N loss. Management measures can also be taken to minimise agricultural biomass burning. This chapter reviews the current literature on CSA practices that are available to reduce GHG emissions and increase soil C sequestration and develops a guideline on best management practices to reduce GHG emissions, increase C sequestration, and enhance crop productivity in agricultural production systems

    Global patient outcomes after elective surgery: prospective cohort study in 27 low-, middle- and high-income countries.

    Get PDF
    BACKGROUND: As global initiatives increase patient access to surgical treatments, there remains a need to understand the adverse effects of surgery and define appropriate levels of perioperative care. METHODS: We designed a prospective international 7-day cohort study of outcomes following elective adult inpatient surgery in 27 countries. The primary outcome was in-hospital complications. Secondary outcomes were death following a complication (failure to rescue) and death in hospital. Process measures were admission to critical care immediately after surgery or to treat a complication and duration of hospital stay. A single definition of critical care was used for all countries. RESULTS: A total of 474 hospitals in 19 high-, 7 middle- and 1 low-income country were included in the primary analysis. Data included 44 814 patients with a median hospital stay of 4 (range 2-7) days. A total of 7508 patients (16.8%) developed one or more postoperative complication and 207 died (0.5%). The overall mortality among patients who developed complications was 2.8%. Mortality following complications ranged from 2.4% for pulmonary embolism to 43.9% for cardiac arrest. A total of 4360 (9.7%) patients were admitted to a critical care unit as routine immediately after surgery, of whom 2198 (50.4%) developed a complication, with 105 (2.4%) deaths. A total of 1233 patients (16.4%) were admitted to a critical care unit to treat complications, with 119 (9.7%) deaths. Despite lower baseline risk, outcomes were similar in low- and middle-income compared with high-income countries. CONCLUSIONS: Poor patient outcomes are common after inpatient surgery. Global initiatives to increase access to surgical treatments should also address the need for safe perioperative care. STUDY REGISTRATION: ISRCTN5181700

    A CMOS temperature sensor with a voltage-calibrated inaccuracy of ±0.15°C (3σ) from -55 to 125°C

    No full text
    This paper describes an energy-efficient CMOS temperature sensor intended for use in RFID tags. The sensor achieves an inaccuracy of ±0.15°C (3σ) over the military temperature range (-55 to 125°C) and dissipates only 27nJ/conversion: over 20× less than a previous sensor with comparable accuracy and resolution [2]. This energy efficiency is achieved by the use of an improved charge-balancing scheme and a zoom ADC that combines a 5b coarse SAR conversion with a 10b fine 2 nd -order ΔΣ conversion
    corecore