Search

Reconstructing the Holocene megaflood history is a key component of understanding the mechanism of past climate change and assessing the potential impact of future catastrophic events. The Pearl River is the longest watercourse in southern China, and its lower reach has been identified as one of the world's most vulnerable regions for flood exposure. However, there is a complete lack of millennial-scale geological records of paleomegafloods for the future prediction of once-in-a-hundred (even once-in-a-thousand) year floods in southern China. Here, we identified a series of paleomegaflood deposits interbedded with wood-rich peat layers in the lower West Pearl River area. All paleoflood layers have been well dated using AMS 14C dating method. According to the regional correlation of the flood sequence, sediment characteristics and provenance analysis, there have been at least 7 megafloods corresponding to once-in-a-thousand-year events in the lower reaches of the West Pearl River during the past 6000 years, with an average return period of approximately 855 years. The identified paleomegafloods were coeval with periods of strong El Niño-Southern Oscillation (ENSO), indicating that weakening of the Asian summer monsoon, associated with enhanced ENSO variability, may have triggered abnormally high precipitation leading to flooding of exceptional magnitude in southern China. In addition, the most prominent paleomegafloods identified in the lower Pearl River coincided with intervals of lower precipitation and fewer storms in central-eastern China, indicating the intensification of the meridional "tripole" pattern of precipitation across eastern China during the latter half of the Holocene. Increased land use and deforestation over the last 2000 years have resulted in soil loss and rapid degradation of local primeval forest ecosystems, leading to more catastrophic flooding. Large amounts of rice pollen in the uppermost flood layer during the Song Dynasty indicate that this megaflood may have inundated a large area of cultivated land. The periodic occurrence of Holocene megafloods not only caused damage to human existence, but also affected the evolution of local civilization. This study reveals for the first time a series of Holocene millennial-scale megafloods and sheds new light on the importance of atmosphere-ocean interactions in the tropical Pacific and monsoon subtropical climate dynamics for precipitation anomalies in East Asia. Our data yield valuable information for future research into climate extremes and hazard prevention. [ABSTRACT FROM AUTHOR]

Nowadays, more and more federated learning algorithms have been implemented in edge computing, to provide various customized services for mobile users, which has strongly supported the rapid development of edge intelligence. However, most of them are designed relying on the reliable device-to-device communications, which is not a realistic assumption in the wireless environment. This paper considers a realistic aggregation problem for federated learning in a single-hop wireless network, in which the parameters of machine learning models are aggregated from the learning agents to a parameter server via a wireless channel with physical interference constraint. Assuming that all the learning agents and the parameter server are within a distance Γ from each other, we show that it is possible to construct a spanning tree to connect all the learning agents to the parameter server for federated learning within O (log Γ) time steps. After the spanning tree is constructed, it only takes O (log Γ) time steps to aggregate all the training parameters from the learning agents to the parameter server. Thus, the server can update its machine learning model once according to the aggregated results. Theoretical analyses and numerical simulations are conducted to show the performance of our algorithm. [ABSTRACT FROM AUTHOR]

Person reidentification (ReID) is a challenging computer vision task for identifying or verifying one or more persons when the faces are not available. In ReID, the indistinguishable background usually affects the model's perception of the foreground, which reduces the performance of ReID. Generally, the background of the same camera is similar, whereas that of different cameras is quite different. Based on this finding, we propose a template-aware transformer (TAT) method which can learn intersample indistinguishable features by introducing a learnable template for the transformer structure to cut down the model's attention to regions of the image with low discrimination, including backgrounds and occlusions. In the multiheaded attention module of the encoder, this template directs template-aware attention to indistinguishable features of the image and gradually increases the attention to distinguishable features as the encoder block deepens. We also increase the number of templates using side information considering the characteristics of ReID tasks to adapt the model to backgrounds that vary significantly with different camera IDs. Finally, we demonstrate the validity of our theories using various public data sets and achieve competitive results via a quantitative evaluation. [ABSTRACT FROM AUTHOR]

Volume calculation from 3D point cloud is widely used in engineering and applications. The existing methods either have large errors or are time-consuming. This paper focuses on the coal measurement. Based on the triangular mesh generated from the point cloud, each triangle is projected downward to the base plane to form a voxel. We derive the calculation formula of voxel by an integral method, which is more efficient than the method of decomposing voxel into tetrahedrons and more accurate than slicing methods. Furthermore, this paper proposes a Delaunay triangulation-driven volume calculation (DTVC) method. DTVC does not preserve the Delaunay triangles but directly calculates the volume in the process of triangulation. It saves memory and running time. Experimental results show that DTVC has achieved a good balance between error and efficiency. [ABSTRACT FROM AUTHOR]

Objective: The aim of this study was to examine the correlation between maternal body composition during the second trimester and the occurrence of dyslipidemia in the third trimester. Methods: A cohort of 1508 pregnant women who underwent antenatal testing at Shanghai Sixth People's Hospital, affiliated with Shanghai Jiao Tong University School of Medicine, were included in this study. Maternal body composition was assessed using bioimpedance analysis between 14 and 22 weeks of gestation. Results: Among the 1508 participants, a total of 1420 individuals (94.2%) were diagnosed with dyslipidemia. Notably, there were significant differences in body composition between the normal lipid group and the dyslipidemia group. Logistic regression analysis revealed that various factors including BMI, total body water (TBW), intra-cellular water (ICW), extra-cellular water (ECW), percent body fat (PBF), visceral fat area (VFA), fat-free mass (FFM) and arm circumference (AC) during the second trimester were all found to be associated with dyslipidemia in the third trimester. Conclusion: The present study found that maternal body composition was associated with dyslipidemia. The BMI, TBW, ICW, ECW, PBF, VFA, FFM and AC in second trimester were associated with dyslipidemia in third trimester. [ABSTRACT FROM AUTHOR]

This paper studies batch processing of core maintenance in hypergraph streams. We focus on updating the coreness of each vertex after the hypergraph evolves. Unlike existing works that mainly focus on exact coreness updates for the single hyperedge dynamic or approximate update, we propose the first known batch processing algorithms for exact core maintenance with insertions or deletions of multiple hyperedges. By proposing a hyperedge structure Joint Hyperedge Set, we tackle the challenges of quantifying the range of coreness change and finding potential vertices whose coreness may update. In addition, we accelerate coreness updates even further by finding structures that enable parallel execution. Extensive experiments illustrate the efficiency, scalability, and effectiveness of our batch core maintenance algorithms on real-world hypergraphs. It shows that our algorithms can be faster than the single hyperedge processing approaches by a factor of nearly half the number of hyperedges processed, and our parallel algorithms achieve linear acceleration with the increasing number of threads. [ABSTRACT FROM AUTHOR]

Person re-identification is an important part of smart surveillance systems which match people across different scenarios. The most challenging aspect of person re-identification is to design robust features and similarity models that reduce the impact of viewpoints, lightings, background clutters, camera settings, occlusions, and pedestrian poses under different scenarios. DeepDiff is a learning method proposed in this paper which uses deep neural networks to identify the different features of various human body parts, and then evaluate the similarities between those corresponding parts. Given those two corresponding parts, we propose three subnets that show deep difference using original data, feature maps, and spatial variations. We focus on a part-based method which introduces a pyramid partition architecture through different partition granularities on human images. In order to determine whether those two parts belong to the same identity, we utilize two combinations of our three subnets. Lastly, we present a part-based integration validation method so as to achieve better performance. The DeepDiff’s effectiveness is validated on three public datasets, including: CUHK03 (labeled, detected), VIPeR and CUHK01 (100 and 486 identities settings). The experimental study shows that the proposed DeepDiff model has promising potential. [ABSTRACT FROM AUTHOR]

Objective To investigate the relationship between disposition index (DI) and the risk of adverse pregnancy outcomes in women with gestational diabetes mellitus (GDM). Methods A total of 1 912 women were prospectively recruited from January 2015 to December 2017 in Department of Gynecology and Obstetrics of the Shanghai Jiao Tong University of Medicine Affiliated Sixth People's Hospital. According to whether they had GDM, they were divided into GDM group (434 cases) and non-GDM group (1 478 cases). All participants received a 75 g oral glucose tolerance test (OGTT) during the 24-28 gestational weeks and complete basal and delivery information was collected. The insulin sensitivity index (ISI) and the area under the insulin curve/area under the blood glucose curve (AUCINS120/AUCGLU120) were calculated from the results of the 75 g OGTT. The DI was calculated as the product of the ISI and AUCINS120/AUCGLU120 and then log transformed. Adverse pregnancy outcomes were defined as large for gestational age (LGA), primary cesarean section, preeclampsia, and preterm birth. Comparison between groups was performed by t test, Wilcoxon rank sum test and χ² test. Logistic regression analysis was used to explore the relationship between DI and the risk of adverse pregnancy outcomes. Results Women with GDM had a lower ISI even DI than women without GDM. Compared with women with GDM in the lowest quartile (Q1, DI<2.90) of DI, multivariable adjusted (age, body mass index, other pregnancy complications, gestational age at the time of OGTT, fetal sex and parity) odds ratios (OR) of adverse pregnancy outcomes of those in the second (Q2, DI 2.90-3.02), third (Q3, DI 3.03-3.16), and highest (Q4, DI>3.16) quartiles of DI were reduced by 55% (95%CI 0.23-0.87), 34% (95%CI 0.35-1.25), and 59% (95%CI 0.24-0.71) (P for trend=0.01), respectively. When DI was considered as a continuous variable, the multivariable adjusted OR of adverse pregnancy outcomes in women with GDM was reduced by 60% (95%CI 0.21-0.73) for each one unit increase in DI. However, DI was not associated with adverse pregnancy outcomes in women without GDM (OR=0.66, 95%CI 0.36-1.18). Conclusion Elevated DI levels in women with GDM are closely associated with a reduced risk of adverse pregnancy outcomes. [ABSTRACT FROM AUTHOR]

Background: Little attention has been paid to whether snoring frequency is associated with body composition in menopausal women, particularly in China. This study objected to investigate the association between self-reported snoring and body composition in (peri-post) menopausal Chinese women as well as metabolic indicators.Methods: This cross-sectional study enrolled 715 participants aged 40-67 years from the Menopause Clinic in the Shanghai Sixth People's Hospital. Participants were categorized into four subgroups stratified by self-reported snoring frequency: never, rarely (< 1 night per week), occasionally (1-2 nights per week), regularly (≥3 nights per week), while body composition was measured using bioelectrical impedance analysis (BIA). Besides, blood sample were collected to test the glycolipid indicators.Results: In our sample of investigation, regular snoring (≥3 nights per week) was found to be an independent risk factor for higher fat mass (total, upper limbs, trunk), with the highest risk of 2.4 times for fat mass of trunk after adjusting for metabolic confounders(p = 0.003). Meanwhile, regular snoring was independently associated with higher fat mass (total and each segment) only in menopausal transition (p = 0.023).Conclusions: We suggested that self-reported regular snoring may be taken as a simple alternative to predict higher fat mass (≥17.11 kg, upper quartile) in menopausal women. Similarly, body composition should be attached to the great importance to those who in menopausal transition in order to help to prevent obstructive sleep apnea (OSA). [ABSTRACT FROM AUTHOR]

Aim Our aims were to provide new pollen data for establishing a sub-continental surface pollen database (East Asian Pollen Database, EAPD) and to study relationships between vegetation and climate. Location The sample sites covered most regions of East Asia, including China, Mongolia, the Russian Far East, Vietnam, Cambodia and Thailand. Methods Data quality control procedures were applied, including taxonomic standardization, removal of duplicates, and adjustment of geographical coordinates. Vegetation types and climate parameters were assigned to each sample. Modern pollen distribution maps were drawn using circle scattergrams. The plots of pollen percentages versus climate variables allowed quantitative estimates of climate values. The modern analogue technique (MAT) was used to predict modern biomes and climate parameters. Results Pollen assemblages extracted from 2858 sites were used to model the geographical distribution of selected taxa and their relationships with climate. For most taxa, the reconstructed range fitted the observed geographical distribution rather well. Arboreal pollen (AP) and Pinus dominated the transition zone between forest and steppe. Use of the MAT revealed that the predicted and observed biomes matched in 71% of the cases. The warm temperate evergreen broadleaf forest had the best agreement between predictions and observations. Climate values reconstructed using MAT were highly correlated with observed values in January temperature. The correlation coefficient of the temperature variables ranged from 0.799 to 0.930 and was as high as 0.939 for precipitation. Main conclusions This paper documents a new modern pollen database for East Asia and makes the data readily available. The reconstructed biomes and climate variables are significantly correlated with the observed values, thus demonstrating the utility of the pollen database for future multiscale palaeoenvironmental studies. [ABSTRACT FROM AUTHOR]