Search

Hashing learning is efficient for large-scale image retrieval by using the nearest neighbor search with binary codes instead of continuous representations. With the success of deep neural networks in related tasks such as data representation, recent hashing methods based on deep learning can further improve image retrieval quality and classification accuracy. However, most existing methods are primarily designed to maximize the performance of retrieval based on linear scan of hash codes which is still time-consuming on large-scale datasets. Fortunately, Hamming space retrieval is an alternative as it is less time-consuming by retrieving data points that are within a Hamming ball with a given Hamming radius, but few works focus on that. In this paper, we propose a concentrated hashing method with neighborhood embedding (CHNE) for efficient and effective image retrieval and classification. By integrating Cauchy cross-entropy and pair-wise weighted similarity loss, CHNE can enable similar data pairs with smaller Hamming distance and dissimilar data pairs with larger Hamming distance. In addition, existing hashing methods are usually designed for retrieval, thus the performance of classification using the binary codes is not guaranteed. To tackle this problem, we jointly minimize the regression quantization and neighborhood structure reconstruction errors in the loss function to improve the classification accuracy. The proposed end-to-end deep hashing method can be optimized by back-propagation in a standard manner. Experimental results on several datasets demonstrate that the proposed method can improve the performance of retrieval and classification. Due to its generality, the proposed method is expected to be useful for image retrieval and classification in broader areas.

Developing low-cost, well-performing, and resource-efficient photocatalysts with enhanced solar light utilization can contribute to the practicability of photocatalytic techniques in organic wastewater treatment. This study fabricated and characterized a novel sunlight-driven BiOBr- and acetylene black (AB)-loaded monolithic photocatalyst. The fly ash-based geopolymer acts as photocatalyst support that can also provide adsorption sites and semiconductor metal oxide (Fe

The oxygen level is key benthic ecosystem health. In this study, a new kind of slow-release oxygen material (SROM) was developed and evaluated in a simulation experiment. The effects of SROM dose and dosing method on the pH and DO, the release of nitrogen and phosphorus, and greenhouse gas emissions were studied. The restoration of typical benthic species (Ceratophyllum represented submerged plants and Cipangopaludina cahayensis represented benthic animals) was also evaluated based on the analysis of catalase and peroxidase activities, survival rate, and biomass. The result shows that dosing SROM on mud surfaces had a better effect than dosing in mud. When dosing SROM on the surface of mud at a suitable dose, the DO of water increased from 0.5 mg/L to higher than 4 mg/L, and the pH was below 9, which would be suitable for the survival of benthos. Dosing SROM could also cause the concentrations of nutrient elements (NH

IMPORTANCEChildren exposed to a high level of prenatal alcohol exposure (PAE) are more likely to develop fetal alcohol spectrum disorder with adverse phenotypes on their faces. However, it is still poorly understood, which level of PAE is associated with facial manifestation and if such associations persist during childhood.OBJECTIVETo examine the association between PAE and children’s facial phenotype in a prospective multi-ethnic population-based study.DESIGN, SETTING, AND PARTICIPANTSThis study was based on the Generation R Study, a prospective cohort from fetal life onwards with maternal and offspring data. Children had a 3D facial image taken at ages 9 (n=3160) and 13 years (n=2492).EXPOSURESWe defined 6 levels of PAE based on the frequency and dose of alcohol consumption, and defined three tiers based on the timing of alcohol exposure of the unborn child.MAIN OUTCOMES AND MEASURESFor image analysis, we used 3D graph convolutional networks for non-linear dimensionality reduction, which compressed the high-dimensional images into 200 endophenotypes representing facial morphology. These facial endophenotypes were used as dependent variables in a linear regression analysis to search for associations with PAE. Finally, to detect specific facial components associated with PAE, we mapped statistically significant endophenotypes back to the 3D facial shape. We generated heatmaps to display the facial changes associated with PAE. PAE prediction based on facial shape was also performed; the prediction accuracy was estimated by area under the receiver operating characteristic curve (AUC).RESULTSA significant association between PAE and facial shape was found in the 9-year-old children, at all levels of alcohol exposure: the higher the level of exposure, the stronger the association. Moreover, PAE before and during pregnancy was associated with facial shape. The most common detected facial phenotypes included turned-up nose tip, shortened nose, turned-out jaw and turned-in lower-eyelid-related regions. For the 13-year-old children, the associations were weaker and the AUCs lower than those of the 9-years-old children.CONCLUSIONS AND RELEVANCEPAE before and during pregnancy, even at low level, is associated with the facial shape of children, and these associations become weaker as children grow older.Key pointsQuestionsWhich level of prenatal alcohol exposure (PAE) is associated with children’s facial shape at different lifetimes?FindingsA 3D facial phenotype analysis of 9- (N=3160) and 13- (N=2492) year-old children was conducted. PAE before and during pregnancy, even at low level, associates with the children’s facial shape; the association becomes weaker as children grow older.MeaningThe results suggest avoiding drinking alcohol at least three months before and during pregnancy is the safest option for the child. More research is needed to investigate the consequences of PAE on children; the discovered facial phenotypes could be used as a biomarker for further investigations.

This study aimed to evaluate the global distribution, associated factors, and epidemiologic trends of gallbladder cancer (GBC) by country, sex, and age groups. The Global Cancer Observatory was interrogated for the disease burden of GBC using age-standardized rates (ASR). The prevalence of different potential risk factors for each country was extracted from Global Health Observatory and their associations with GBC incidence and mortality were examined by linear regression analysis using beta coefficients (β). The Cancer Incidence in Five Continents I-XI and the WHO Mortality database were searched and Average Annual Percent Change (AAPC) was generated from joinpoint regression analysis. The incidence (ASR = 2.3) and mortality (ASR = 1.7) of GBC varied globally in 2018 and were higher in more developed countries and among females. Countries with higher incidence had higher human development index (βmale = 0.37; βfemale = 0.27), gross domestic products (βmale = 0.13) and higher prevalence of current smoking (βfemale = 0.05), overweight (βmale = 0.02), obesity (βmale = 0.03), and hypercholesterolaemia (βmale = 0.07). Similar patterns of associations were also observed for mortality with an additional association found for diabetes (βfemale = 0.07). Although there was an overall decreasing trend in mortality, an increasing trend in incidence was observed among some populations, particularly in males (AAPCs, 8.97 to 1.92) and in younger individuals aged

Background This study aimed to evaluate the updated incidence, mortality, risk factors, and epidemiologic trends of colorectal cancer (CRC) for different regions and sex using publicly available data from 186 countries. Methods The Global Cancer Observatory (GLOBOCAN) database was retrieved for the age-standardized rates (ASR) for incidence and mortality of CRC in 2018 and 2012. The prevalence of risk factors (2008–2016) was retrieved from the Global Health Observatory database. The associations between the prevalence of risk factors and incidence and mortality of CRC and were measured using beta coefficients (β) with 95% confidence intervals (CI) generated from a univariable linear regression analysis. Results In 2018, the highest incidence and mortality of CRC were in regions of very high human development index (HDI) (ASRs, 30.6 and 11.1) with men to women ratio of 1.4 and 1.2. Population in countries with higher incidence have higher alcohol consumption (β=1.30, 95% CI 1.02 to 1.57 for men; β=0.46, CI 0.35 to 0.57 for women), higher prevalence of tobacco use (β=0.11, CI 0.02 to 0.19 for women), insufficient physical activity (β=0.59, CI 0.34 to 0.84; β=0.19, CI 0.09 to 0.29), overweight (β=0.47, CI 0.38 to 0.57; β=0.16, CI 0.12 to 0.20), obesity (β=0.24, CI 0.11 to 0.38; β=0.34, CI 0.25 to 0.43), and elevated cholesterol (β=1.80, CI 1.53 to 2.07; β=1.22, CI 1.05 to 1.38). Similar associations were also found for mortality (figure 1). From 2012 to 2018, there was an overall increase in the trend of incidence and mortality, particularly in Asia (+29.7% and +17.4%) and Africa (+24.3% and +17.6%), and among men (+14.6% and +8.0%) compared with women (+14.0% and +4.3%). Conclusions The variation in disease burden of CRC was associated with HDI and the prevalence of risk factors. There was an increasing trend in the incidence and mortality of CRC, particularly in regions with low and middle incomes and among men. More intensive lifestyle modifications and population-based screening are recommended for these populations.

Reclaimed wastewater (RW) is increasingly used to irrigate agricultural land and to alleviate agricultural water shortages worldwide. This usage has resulted in concerns about soil contamination by pharmaceuticals and personal care products (PPCPs) and the human health risks associated with dietary crop intake. In this study, we systematically analysed the occurrence and accumulation of 11 PPCPs and one active metabolite in soils and various crops (cucumber, eggplant, long bean and wheat) from realistic RW irrigation fields with different irrigation histories (20, 30 and 40 years) in Beijing and evaluated the human health risks associated with the consumption of these crops. The 11 PPCPs and one active metabolite were detected at concentrations ranging from 0.67 to 22.92 ng L−1 in RW, 0.029–28.13 μg kg−1 in irrigated soil, and 0.05). The concentrations of the studied PPCPs in the RW used for irrigation followed different patterns from the concentrations detected in the irrigated soils and crops. Although the concentrations of sulfamethoxazole, sulfisoxazole, sulfamethazine and trimethoprim in RW were higher than those of many other studied PPCPs, their respective values in the irrigated soils and crops did not display a similar tendency. The uptake and accumulation of PPCPs varied among the crop species (P

To improve the removal of pharmaceutical and personal care products (PPCPs) from contaminated water, a novel electrolysis-integrated tidal flow constructed wetland system (E-TFCW) was developed. E-TFCWs have the highest removal efficiency for the four investigated PPCPs when the electrodes are arranged at the bottom of the E-TFCWs. The removal efficiency of four PPCPs increases from 4.41% to 11.71% to 28.83%-68.32% when using electrolysis combined with tidal flow constructed wetlands. The synergy effect for the degradation of PPCPs between electrolysis and constructed wetlands on the degradation of PPCPs in E-TFCWs were significant (enhancement factor > 1), while the removal efficiency of PPCPs after the electrolysis (10 h, accounting for 14.98–44.41%) is higher than that before electrolysis (10 h, accounting for 9.62–27.14%) and electrolysis process (4 h) accounting for 42.81%-68.47%. Electrolysis can enhance the removal of PPCPs through direct oxidation. Moreover, the increase in intermediate oxidation compounds in the E-TFCWs may imply the enhanced biodegradability of PPCPs during the electrolysis process since the by-products of PPCPs were identified. The contents of Thermomonas, Microbacterium, Cloacibacterium and Acidovorax at the bottom of the E-TFCWs were significantly higher than those of other bacteria, which contributed to the removal of PPCPs. Electrolysis can enhance the biodegradability of PPCPs and the formation of anoxic/aerobic conditions for the growth of predominant bacteria, which contributes to the degradation of PPCPs in the E-TFCWs. The results illustrate that electrolysis-integrated tidal flow constructed wetlands are an effective technology for the removal of PPCPs.

The lack of high-quality labeled training data has been one of the critical challenges facing many industrial machine learning tasks. To tackle this challenge, in this paper, we propose a semi-supervised learning method to utilize unlabeled data and user feedback signals to improve the performance of ML models. The method employs a primary model Main and an auxiliary evaluation model Eval, where Main and Eval models are trained iteratively by automatically generating labeled data from unlabeled data and/or users’ feedback signals. The proposed approach is applied to different text classification tasks. We report results on both the publicly available Yahoo! Answers dataset and our e-commerce product classification dataset. The experimental results show that the proposed method reduces the classification error rate by 4% and up to 15% across various experimental setups and datasets. A detailed comparison with other semi-supervised learning approaches is also presented later in the paper. The results from various text classification tasks demonstrate that our method outperforms those developed in previous related studies.

The temporal and spatial variations of the thermospheric mass density during a series of idealized substorms were investigated using the Global Ionosphere Thermosphere Model (GITM). The maximum mass density perturbation of an idealized substorm with a peak variation of hemispheric power (HP) of 50 GigaWatts (GW) and interplanetary magnetic field (IMF) Bz of -2 nT was ~14% about 50 min after the substorm onset in the nightside sector of the auroral zone. The mass density response to different types of energy input has a strong local time dependence, with the mass density perturbation due to only an IMF Bz variation peaking in the dusk sector and the density perturbation due to only HP variations peaks in the nightside sector. Simulations with IMF Bz changes only, and HP changes only showed that the system behaves slightly nonlinearly when both IMF and HP variations are included (a maximum of 6% of the non-linearity), and that the non-linearity grows with energy input. The neutral gas heating rate due to Joule heating was of same magnitude as the heating rate due to precipitation, but the majority of the temperature enhancement due to the heating due to precipitation occurs at lower altitude as compared to the auroral heating. About 110 min after onset, a negative mass density perturbation (~-5%) occurred in the night sector, which was consistent with the mass density measurement of the CHAMP satellite.

We have found that consideration of neutral helium as a major species leads to a more complete physics-based modeling description of the Earth's upper thermosphere. An augmented version of the composition equation employed by the Thermosphere-Ionosphere-Electrodynamic General Circulation Model (TIE-GCM) is presented, enabling the inclusion of helium as the fourth major neutral constituent. Exospheric transport acting above the upper boundary of the model is considered, further improving the local time and latitudinal distributions of helium. The new model successfully simulates a previously observed phenomenon known as the “winter helium bulge,” yielding behavior very similar to that of an empirical model based on mass spectrometer observations. This inclusion has direct consequence on the study of atmospheric drag for low-Earth-orbiting satellites, as well as potential implications on exospheric and topside ionospheric research.

Accelerometer data from coplanar orbits of Challenging Minisatellite Payload (CHAMP) and Gravity Recovery and Climate Experiment (GRACE) satellites were used to study the complex altitude and latitude variations of the thermosphere mass density response to geomagnetic activity during 1–10 December 2008 near 09 LT. Helium number densities near 500 km altitude were extracted from the CHAMP and GRACE measurements and clearly show the presence of a winter hemisphere helium bulge. This recent extreme solar minimum indicates that wintertime helium concentrations exceed NRLMSISE-00 model estimates by 30%–70% during quiet geomagnetic activity after adjusting F10.7 input into MSIS. The perturbation in mass density from quiet to active conditions is found to be less enhanced in the winter hemisphere at the higher GRACE altitudes (25%) than at the lower CHAMP altitudes (60%) and is attributed to dynamic behavior in the helium/oxygen transition. The investigation revealed the maximum storm time density perturbation to occur near the He/O transition region with a much weaker maximum near the O/N2 transition region. The altitude of maximum density perturbation occurs where the perturbation in the weighted pressure scale height is equal and opposite to the perturbation in the weighted mean molecular weight scale height. The altitude structure of density scale height perturbation is significantly influenced by the changes in the molecular weight scale height and can account for 50% of the change in mass density scale height in a region correspondingly close to the He/O transition during the 2008 solar minimum period.