Search

Top-K and selection operations are critical in data processing and analysis, and their efficient implementation on GPUs is increasingly important due to the growing demands of data analysis. Existing methods, primarily relying on the bucket partition execution model, encounter challenges such as uneven bucket distribution and latency in merging processes. To address these issues, we introduce a novel Split-Bucket Partition (SBP) execution model that specifically addresses these challenges. Additionally, we propose task and control flow optimizations targeted at top-K and selection algorithms, which further contribute to performance improvements. Our optimized algorithms significantly outperform existing approaches, delivering performance gains of up to 2.3 times and 1.6 times for different bucket partitioning rules. Our algorithms show robust performance improvements in non-uniform data scenarios, with gains ranging from 1.9 times to 15.5 times. However, it should be noted that the SBP model has limitations related to shared memory and register utilization, potentially impacting performance. Tests on TU102 and A100 GPU architectures validate the effectiveness of our approach, achieving a maximum speedup of 2.9 times. The study suggests that while the SBP model is effective for top-K and selection algorithms, it also holds promise for other computational tasks, setting the stage for future research. [ABSTRACT FROM AUTHOR]

Sparse matrix-vector multiplication (SpMV) plays a critical role in a wide range of linear algebra computations, particularly in scientific and engineering disciplines. However, the irregular memory access patterns, extensive memory usage, high bandwidth requirements, and underutilization of parallelism hinder the computational efficiency of SpMV on GPUs. In this paper, we propose a novel approach called block-wise dynamic mixed-precision (BDMP) to address these challenges. Our methodology involves partitioning the original matrix into uniformly sized blocks, with each block's size determined by considering architectural characteristics and accuracy requirements. Additionally, we dynamically assign precision to each block using a precision selection method that takes into account the value distribution of the original sparse matrix. We develop two distinct SpMV computation algorithms for BDMP: BDMP-PBP (Precision-based partitioning) and BDMP-TCKI (Tailored compression and kernel implementation). BDMP-PBP partitions the matrix into two independent matrices for separate computations based on block precision, offering flexibility for integration with other optimization techniques. Meanwhile, BDMP-TCKI focuses on achieving significant thread-level parallelism and memory utilization by tailoring an appropriate compressed storage format and kernel implementation for each block. We compare BDMP with NVIDIA's cuSPARSE library and three state-of-the-art SpMV methods, including SELLP, MergeBase, and BalanceCSR, using matrices from the University of Florida's SuiteSparse dataset collection. BDMP-PBP and BDMP-TCKI show average speedups up to 2.64 × and 2.91 × on Turing RTX 2080Ti, and up to 2.99 × and 3.22 × on Ampere A100. The results demonstrate that BDMP enables the optimization of computation speed without compromising the precision necessary for reliable results. [ABSTRACT FROM AUTHOR]

Deep neural networks have made great achievements in remote sensing image analyses; however, previous studies have shown that deep neural networks exhibit incredible vulnerability to adversarial examples, which raises concerns about regional safety and production safety. In this paper, we propose an adversarial denoising method based on latent representation guidance for remote sensing image scene classification. In the training phase, we train a variational autoencoder to reconstruct the data using only the clean dataset. At test time, we first calculate the normalized mutual information between the reconstructed image using the variational autoencoder and the reference image as denoised by a discrete cosine transform. The reconstructed image is selectively utilized according to the result of the image quality assessment. Then, the latent representation of the current image is iteratively updated according to the reconstruction loss so as to gradually eliminate the influence of adversarial noise. Because the training of the denoiser only involves clean data, the proposed method is more robust against unknown adversarial noise. Experimental results on the scene classification dataset show the effectiveness of the proposed method. Furthermore, the method achieves better robust accuracy compared with state-of-the-art adversarial defense methods in image classification tasks. [ABSTRACT FROM AUTHOR]

Thermoelectric technology plays a crucial role in harnessing waste heat from automobile exhaust, and developing low-resistance and high-efficiency thermoelectric generator is currently a focal point of research. The intermediate fluid thermoelectric generator (IFTEG) introduces a structural modification to conventional designs by utilizing boiling-condensation heat transfer of the intermediate fluid to enhance power generation performance. In this study, we constructed an experimental system to investigate the thermoelectric characteristics of IFTEG. The results demonstrate that the IFTEG outperforms traditional thermoelectric generators, exhibiting a remarkable increase in total output power ranging from 171% to 283%, a reduction in pressure drop ranging from 23% to 42%, and a significant improvement in voltage distribution uniformity across the modules. Increasing the exhaust temperature and flow rate significantly enhances the net output power while minimally affecting the voltage distribution uniformity. The net output power of 10.17 W and the net thermoelectric conversion efficiency of 1.39% are achieved at 200 °C and 30 m3/h. Overall, this novel generator offers advantages such as low pressure loss, high efficiency, and high reliability. The findings of this study provide valuable guidance for the development of exhaust thermoelectric generators. • The thermoelectric performance of IFTEG is experimentally investigated. • The power output of IFTEG can be enhanced by 171%–283%. • The pressure drop of IFTEG be reduced by 23%–42%. • The distribution of module voltage becomes more uniform. [ABSTRACT FROM AUTHOR]

• New method for determining waveguide axial stress using guided waves. • Analytical insight into how axial stress affects guided wave dispersion properties. • In-depth investigation on selecting optimum modes for guided wave testing. • Detailed parametric study for predicting longitudinal rail stress. The determination of the longitudinal stress in continuous welded rails is essential for the safe operation of high-speed railways. Guided wave inspections have many advantages, and can provide an effective tool for non-destructive measurements of the longitudinal stress in rails. This paper provides a new method for determining longitudinal rail stress using ultrasonic guided wave dispersion properties. First, the characteristic equations of guided waves propagating in a waveguide with existing longitudinal stress are constructed, generating the dispersion properties of the waveguide such as the wavenumber, phase velocity and group velocity. The sensitivity of the dispersion properties of propagative modes to the longitudinal rail stress is investigated, which is then used for selecting appropriate modes for guided wave inspections. On the basis of the perturbation method, the relationship between the change in longitudinal rail stress and the change in dispersion properties is developed, which can be used for determining the longitudinal rail stress from dispersion properties. Furthermore, a mode selection index considering both the stress sensitivity and mode shape characteristics is employed for selecting appropriate propagative modes for guide wave inspections. Finally, a numerical example for CHN60 rail is adopted to demonstrate the effectiveness of the proposed method. From the results, the proposed method can accurately determine the longitudinal stress in the rail by using the dispersion properties available. [ABSTRACT FROM AUTHOR]

Social networks are ubiquitous in contemporary society, where one's opinions and behaviors are easily influenced by those around them. The effectiveness of social influence analysis has become significant for numerous application fields such as online recommendation, advertising, and viral marketing. Graph-based deep learning has achieved significant success in influence prediction. However, the existing methods ignore the privacy issue and are incapable of cross-organizational collaboration. Inspired by the recent success of federated learning in privacy protection and breaking the data barrier. In this paper, a novel federated learning framework, FedInf, is proposed to tackle the problem of social influence prediction. Specially, to further protect user privacy and achieve secure aggregation for multiple local models, we introduce differential privacy into the local parameters, adding artificial noise before model aggregation. In order to trade-off utility and privacy, we first freeze the embedding layers to reduce the number of upload parameters, and then project model parameters into low-dimensional space. Therefore, less noise is required to provide the same level of privacy protection. Extensive experiments on four real-world datasets demonstrate that the proposed FedInf is effective while providing privacy protection. [ABSTRACT FROM AUTHOR]

Underwater images intuitively reflect the underwater environment information. However, underwater images have defects such as colour distortion and low contrast, which seriously affect the processing of complex underwater visual tasks. Here, a novel mixed model called mixed underwater image generative adversarial network (MUGAN) is presented, consisting of a generator and corresponding discriminator. The generator is built following the U‐shaped architecture, where a mixed block of convolution and self‐attention is developed. It effectively exploits the complementarity between the two paradigms. In addition, a dual discriminator is employed to induce the generator to produce realistic images at both the global semantic and local detail levels, which is only discriminates based on the patch‐level information. Meanwhile, a multi‐term loss function is formulated to supervise adversarial training by evaluating the perceptual quality of an image based on its global content, local texture and illumination smoothness. To validate the proposed approach, extensive experiments are conducted on the public underwater datasets. MUGAN achieves promising performance in terms of colour, contrast and naturalness, showing a significant improvement over other competitive models in visual quality and quantitative metrics. [ABSTRACT FROM AUTHOR]

The Tabei Uplift in the Tarim Basin of NW China represents an area with significant hydrocarbon exploration potential. This paper describes the effects of multi-phase faults on structural and hydrocarbon systems in the Tabei Uplift utilizing high-quality 3-D seismic, geological outcrops and drilling and production data. The interpretation and analysis of 10 fault systems from the Cambrian through the Cretaceous, shows that the deep, middle and shallow faults developed in four stages: the Early Caledonian, Mid-Late Caledonian, Hercynian and Himalayan. Specifically, the Early Caledonian fault system is favorable for the hydrocarbon expulsion of Cambrian source rocks, while Mid-Late Caledonian strike-slip faults facilitate hydrocarbon migration and accumulation. Hercynian and Himalayan strike-slip faults were formed as a result of the reactivation of the Mid-Late Caledonian strike-slip faults and have a destructive effect on the older oil reservoirs. Similarly, Permian volcanic activity also caused varying degrees of damage to Ordovician reservoirs. Oil production along fault zones more than 100 km show that the complex fracture-vuggy (FV) structure is more likely to trap oil and gas when it is damaged than the simple FV structure. Fault-controlled, fracture-vuggy carbonate reservoirs (FVCRs) were classified into six grades, which correlate to the productivity of oil wells in the region. Using this classification system, several drilling locations were selected and resulted in 80% success rate. This classification provides a novel approach for assessing the interplay between fault activity and reservoir formation, enhancing the prediction of hydrocarbon potential in varying strike-slip fault zones. This analysis not only refines our understanding of the region's geodynamic history but also aids in optimizing exploration and development strategies. • Geological and geophysical data document four Paleozoic and Mesozoic fault systems. • Deep faults are beneficial to hydrocarbon expulsion and shallow fault destroy cap rock. • A classification of faults that create traps closely aligns with the productivity of oil wells. • Correlation identified between fault activities and the deformation in multi-type fracture-vuggy reservoirs. [ABSTRACT FROM AUTHOR]

Highly accuracy classification and mapping of seafloor hydrothermal fields provides an important addition to research into the geological background and exploration of seafloor massive sulfides (SMS). Currently, acoustic remote sensing using multibeam sounding systems (MBES) is the primary means of achieving large-scale seafloor classification. However, the characteristics of complex topography, complex seafloor distribution, and deep water depth of the hydrothermal fields make it difficult to obtain accurate high-resolution seafloor classification maps using shipboard MBES surveys. Here, a seafloor classification strategy combining shipboard MBES and autonomous underwater vehicle (AUV) sidescan sonar data is proposed. First, a downscaling model is established, which downscales the MBES backscatter mosaic (12-kHz and 10-m resolution) to a resolution of 2 m. The second step performs feature extraction of the downscaled MBES backscatter mosaic (12-kHz and 2-m resolution), the AUV sidescan sonar backscatter mosaic (150-kHz and 2-m resolution), and a seafloor digital elevation model (2-m resolution). Finally, a deep neural network model was built for training and classification. To evaluate the classification performance of the model, the method was applied to the survey of the Duanqiao hydrothermal field. Results were verified using field data (deep-tow video). The overall root mean square error and coefficient of determination ($R^{2}$) for the classification were 0.032 and 0.840, respectively. The experimental results show that the method can effectively meet the challenges to seafloor classification presented by complex seafloor distribution, and can obtain an accurate high-resolution seafloor classification map. [ABSTRACT FROM AUTHOR]

• Phase change chambers for heat flux enhancement were experimentally investigated. • Effects and mechanisms of heat flux, filling ratio, and area ratio on heat transfer in the chamber were obtained. • The lowest thermal resistance of the chamber was achieved at the smallest area ratio of 1/5 when the boiling heat flux is lower than 7.5 kW/m2. • Visualization of bubble and droplet behaviors was performed. One promising way to increase the output power and the thermoelectric conversion efficiency of an automotive exhaust thermoelectric generator (TEG) is to add a phase change chamber between the exhaust channel and the thermoelectric (TE) modules with the chamber surface contacting with the exhaust channel where boiling occurs being larger than that with the module surfaces where the condensation occurs. The power is increased because a phase change chamber significantly increases the heat flux through the modules; thus, increasing the temperature difference between the module hot and cold sides. Although there is a significant body of literature on phase change chambers, most prior research has focused on chambers with larger condensation areas than boiling areas since they act as heat spreaders, while little is on chambers with condensation areas smaller than boiling areas for heat flux enhancement. These chambers have great potential to improve TEG power generation by enhancing the heat flux, therefore, more studies are needed to understand the chamber heat transfer characteristics. This research experimentally investigated the heat transfer characteristics of three chambers with condensation to boiling area ratios of 1/5, 2/5, and 3/5 for boiling heat fluxes from 4.50 to 22.70 kW/m2 and deionized water filling ratios from 33 to 50%. The results show that the boiling heat transfer coefficient increases with increasing heat flux or decreasing area ratio and is weakly influenced by the filling ratio. The condensation heat transfer characteristics depend on the distribution of droplets which is strongly affected by both the heat flux and the area ratio. The influence of the filling ratio on the condensation heat transfer coefficient increases with increasing area ratio. The thermal resistance of the chamber decreases with increasing heat flux and the lowest thermal resistance is obtained at 1/5 area ratio at lower heat fluxes and at 2/5 or 3/5 area ratio at higher heat fluxes. [ABSTRACT FROM AUTHOR]

The distribution of hydrothermal vents and the biogeography of associated faunal communities in the Indian Ocean are still not well studied, particularly the ultraslow-spreading Southwest Indian Ridge (SWIR). Herein, we present the geological and morphological data for the first reported active hydrothermal field on the ultraslow-spreading center. In this context, we created a detailed seafloor map based on visual data obtained from human-occupied vehicle (HOV) dives. Longqi-1 vent field (LVF) is an off-axis, mafic-hosted but detachment fault-controlled, high-temperature hydrothermal area with at least 28 hydrothermal structures, and their contacting host rocks were identified and mapped. The morphologies of hydrothermal structures in the LVF exhibit unique characteristics, with Zone M characterized by chimneys grown on the relatively flat-lying sulfide mound and Zone S characterized by large sulfide mounds and steep-sided structures hosting large flanges, and isolated beehive structures occurred more often at the boundary between hydrothermal deposits and basalts outcrops in both M and S zone. Microscopic examination showed that anhydrite is the dominant sulfate phase and chalcopyrite, pyrite, pyrrhotite, and sphalerite are the dominant sulfide minerals. Subseafloor hydrothermal circulation controlled by the detachment fault and local fracture zone may have directly determined the morphology of the sulfide structure in the LVF. Our results provide a picture of the LVF as well as complement and expand on previous studies on metal resource evaluation along the SWIR and Indian Ocean ridges, a detailed scenario for studying the surface distribution characteristics of hydrothermal circulation and its relationship with the distribution of hydrothermal fauna, and further enhance our understanding of the formation of modern seafloor hydrothermal systems. • This study report geological and morphological data for the Longqi-1vent field along the ultraslow-spreading SWIR. • In this context, we created a detailed seafloor map based on visual data obtained from human-occupied vehicle dives. • Zones M and S are characterized by relatively flat-lying sulfide mound and large sulfide edifices with steep-sided flanges, respectively. • The Longqi-1 hydrothermal field is the first documented high-temperature venting field along the ultraslow-spreading SWIR. [ABSTRACT FROM AUTHOR]

An extremely magmatic accretion could exist along an ultraslow-spreading ridge that has been approved and surveyed, but less attention has been paid to its associated hydrothermal system, as it is supposed to be similar to those of fast/medium-spreading ridges. Here using an autonomous underwater vehicle and a human-occupied vehicle, a systematic investigation was conducted on the hydrothermal system of an extremely magmatic accretion segment along the Southwest Indian Ridge (50°28′ E). A collapsed chimney, microbial mat, and diffused fluid were observed in the old Duanqiao hydrothermal field (DQF), which was earlier a weakly active hydrothermal area. Notably, evident hydrothermal anomalies, including oxidation–reduction potential (ORP), turbidity, and methane, were detected above the new DQF, which lies southwest of the old DQF. Based on this investigation, we concluded that ORP is one of the most effective parameters available to identify vent location by analyzing each type of detected hydrothermal anomaly and the near-bottom magnetic inversion can image the suspected hydrothermal fluid circulation pathways. In general, we believe that the new DQF is controlled by a magmatic chamber different from those found on fast/medium-spreading ridges, and could be a large hydrothermal system. • Obvious hydrothermal anomalies observed were earlier described as inactive by AUV. • Indicators of hydrothermal activities have their own recognition features. • Duanqiao is a large magmatic-hosted hydrothermal system compared with Lucky Strike. [ABSTRACT FROM AUTHOR]

• Fresh Southwestern Indian Ridge basalt glasses are found to have high water contents and H 2 O/Ce ratios (>600). • High water contents and H 2 O/Ce ratios are not caused by assimilation or diffusion. • The water enrichment is associated with the recycling of arc-mantle residue. Trace amounts of water in the sub-oceanic mantle play crucial role in the vigor of mantle convection and the production of oceanic crust, and other many geodynamic processes. Consequently, the cycling of H 2 O between the mantle and the exosphere in the mantle is one of the critical processes governing Earth's geodynamical and geochemical evolution. While the deep cycling of altered oceanic lithosphere was considered as the main way to replenish the water in oceanic upper mantle, the significance of the arc mantle wedge after the genesis of arc magmatism dragged down by the subducting slab concomitantly has been not well constrained. Here, we report that fresh depleted basaltic glasses from the ultraslow-spreading Southwestern Indian Ridge (SWIR), located far from any recent subduction zones, show unusually high H 2 O/Ce ratios (>600), water contents and heavy hydrogen isotopic compositions. These results could be best explained by recycling of water through melting of a residual hydrous mantle wedge after early melt extraction. Considering that such mantle wedges residue dragged down to the deeper mantle could occupy a volume one order of magnitude larger than that of the subducted lithosphere in the earth history, we suggest that the potential role of such shallow recycling should be considered in studies of global water recycling and the origin of water in the upper asthenosphere. [ABSTRACT FROM AUTHOR]

Multi-period tectonic fractures are common in deep carbonate reservoirs. It's difficult to identify and has the characteristics of stochastic spatial distribution. The quantitative prediction of fractures is critical for unravelling their spatial distribution and for understanding tectonic evolution and stress field distribution. Focusing on the Halahatang oilfield in northern Tarim Basin, outcrop surveys, drilling cores, thin sections, well logs, and seismic data were comprehensively used to better determine the development characteristics, formation time, and spatial distribution of multi-period tectonic fractures based on finite element tectonic stress field simulation. The strike-slip fault system was characterised based on post-stack seismic structural interpretation. By simulating the tectonic stress field in three different tectonic movement periods, taking the fracture density as the evaluation parameter, the spatial distribution characteristics of multi-period fractures were quantified. The numerical simulation results indicated that the development of fractures is controlled by the strike-slip fault, and the Middle Caledonian period is the main formation period of these fractures. This method can accurately predict multi-period complex fractures and can also be applied to other deep carbonate reservoirs. • Quantitative prediction of multi-period fractures in deep carbonate reservoirs. • Multi data and geomechanics can improve the accuracy of fracture prediction. • Fractures were developed in fault conjugate, branch faults and fault interactions. • Fracture development has the characteristics of difference between fault segments. [ABSTRACT FROM AUTHOR]