Diabetic Retinopathy Classification Based on Bilinear Cross Attention Network

Computer-aided diagnostic systems can assist doctors in diagnosing and treating DR cases more effectively, thereby improving work efficiency, reducing the burden on doctors during examinations, and alleviating problems related to uneven distribution of medical resources and shortage of doctors. In this article, we propose a classification method for diabetic retinopathy based on a bilinear multi-attention network. This method uses two backbone networks to extract features, and cross-shares the features using two attention modules to further deepen feature extraction. The non-local attention module is added to address the limitations of traditional convolutional neural networks in capturing global information. By paying attention to highly correlated pathological areas globally, performance improvement can be achieved. We achieved an accuracy of 91.7% on the Messidor dataset.

Path Computing Scheme with Low-Latency and Low-Power in Hybrid Cloud-Fog Network for IIoT

With the rapid development of the Industrial Internet of Things(IIoT),the traditional centralized cloud processing model has encountered the challenges of high communication latency and high energy consumption in handling industrial big data tasks.This paper aims to propose a low-latency and lowenergy path computing scheme for the above problems.This scheme is based on the cloud-fog network architecture.The computing resources of fog network devices in the fog computing layer are used to complete task processing step by step during the data interaction from industrial field devices to the cloud center.A collaborative scheduling strategy based on the particle diversity discrete binary particle swarm optimization(PDBPSO)algorithm is proposed to deploy manufacturing tasks to the fog computing layer reasonably.The task in the form of a directed acyclic graph(DAG)is mapped to a factory fog network in the form of an undirected graph(UG)to find the appropriate computing path for the task,significantly reducing the task processing latency under energy consumption constraints.Simulation experiments show that this scheme’s latency performance outperforms the strategy that tasks are wholly offloaded to the cloud and the strategy that tasks are entirely offloaded to the edge equipment.

A Novel Load Balancing Strategy of Software-Defined Cloud/Fog Networking in the Internet of Vehicles

The Internet of Vehicles(IoV)has been widely researched in recent years,and cloud computing has been one of the key technologies in the IoV.Although cloud computing provides high performance compute,storage and networking services,the IoV still suffers with high processing latency,less mobility support and location awareness.In this paper,we integrate fog computing and software defined networking(SDN) to address those problems.Fog computing extends computing and storing to the edge of the network,which could decrease latency remarkably in addition to enable mobility support and location awareness.Meanwhile,SDN provides flexible centralized control and global knowledge to the network.In order to apply the software defined cloud/fog networking(SDCFN) architecture in the IoV effectively,we propose a novel SDN-based modified constrained optimization particle swarm optimization(MPSO-CO) algorithm which uses the reverse of the flight of mutation particles and linear decrease inertia weight to enhance the performance of constrained optimization particle swarm optimization(PSO-CO).The simulation results indicate that the SDN-based MPSO-CO algorithm could effectively decrease the latency and improve the quality of service(QoS) in the SDCFN architecture.

Observations and modelling of the travel time delay and leading negative phase of the 16 September 2015 Illapel,Chile tsunami

The systematic discrepancies in both tsunami arrival time and leading negative phase(LNP)were identified for the recent transoceanic tsunami on 16 September 2015 in Illapel,Chile by examining the wave characteristics from the tsunami records at 21 Deep-ocean Assessment and Reporting of Tsunami(DART)sites and 29 coastal tide gauge stations.The results revealed systematic travel time delay of as much as 22 min(approximately 1.7%of the total travel time)relative to the simulated long waves from the 2015 Chilean tsunami.The delay discrepancy was found to increase with travel time.It was difficult to identify the LNP from the near-shore observation system due to the strong background noise,but the initial negative phase feature became more obvious as the tsunami propagated away from the source area in the deep ocean.We determined that the LNP for the Chilean tsunami had an average duration of 33 min,which was close to the dominant period of the tsunami source.Most of the amplitude ratios to the first elevation phase were approximately 40%,with the largest equivalent to the first positive phase amplitude.We performed numerical analyses by applying the corrected long wave model,which accounted for the effects of seawater density stratification due to compressibility,self-attraction and loading(SAL)of the earth,and wave dispersion compared with observed tsunami waveforms.We attempted to accurately calculate the arrival time and LNP,and to understand how much of a role the physical mechanism played in the discrepancies for the moderate transoceanic tsunami event.The mainly focus of the study is to quantitatively evaluate the contribution of each secondary physical effect to the systematic discrepancies using the corrected shallow water model.Taking all of these effects into consideration,our results demonstrated good agreement between the observed and simulated waveforms.We can conclude that the corrected shallow water model can reduce the tsunami propagation speed and reproduce the LNP,which is observed for tsunamis that have propagated over long distances frequently.The travel time delay between the observed and corrected simulated waveforms is reduced to<8 min and the amplitude discrepancy between them was also markedly diminished.The incorporated effects amounted to approximately 78%of the travel time delay correction,with seawater density stratification,SAL,and Boussinesq dispersion contributing approximately 39%,21%,and 18%,respectively.The simulated results showed that the elastic loading and Boussinesq dispersion not only affected travel time but also changed the simulated waveforms for this event.In contrast,the seawater stratification only reduced the tsunami speed,whereas the earth’s elasticity loading was responsible for LNP due to the depression of the seafloor surrounding additional tsunami loading at far-field stations.This study revealed that the traditional shallow water model has inherent defects in estimating tsunami arrival,and the leading negative phase of a tsunami is a typical recognizable feature of a moderately strong transoceanic tsunami.These results also support previous theory and can help to explain the observed discrepancies.

Hazard analysis of tsunami disaster on the Maritime Silk Road

The Maritime Silk Road is not only a passageway for business and trade, but also the road of friendship between eastern and western civilizations. The Indian Ocean tsunami in 2004 caused major damage to several coastal countries. Tsunami occurrence regularity and hazard analysis are needed to ensure economic and cultural exchange on the Maritime Silk Road. To explore and identify tsunami hazard on the Maritime Silk Road, the spatial and temporal characteristics of historical tsunami events were given out. Some useful information hidden in historical tsunamis was searched from source parameters, such as seismic magnitude, focal depth and water depth. The tsunami possibility in the case of earthquake occurrence was also studied, exploring the probability of tsunami caused by different magnitudes. The analysis result shows that tsunamis on the Maritime Silk Road mainly occurred in 8 major tectonic faults, each of which has different tsunami occurrence regularity. On the basis of statistical analysis, a numerical model was used to simulate the potential tsunamis and show the tsunami hazard levels along the coast of Maritime Silk Road. The research results of this paper can help the tsunami early warning and ensure the safety of economic and cultural exchanges on the Maritime Silk Road.

Crustal structure beneath the central and western North China from receiver function analysis

The North China Craton(NCC)is one of the oldest cratons on earth.Several important tectonic transformations of MesozoicCenozoic tectonic regime led to the destruction of the North China craton.The knowledge of crustal structure can provide important constraints for the formation and evolution of cratons.New maps of sediment thickness,crustal thickness(H)and v_(P)/v_(S)(κ)in the central and western NCC were obtained using sequential H-κstacking.P-wave receiver functions are calculated using teleseismic waveform data recorded by 405 stations from Chin Array project.Benefiting from the densely distribution of temporary seismic stations,our results reveal details of the crustal structure in the study area.The thickness of sedimentary layer in North China ranges from 0–6.4 km,and the thickest sedimentary layer is in Ordos block and its surroundings(about 2.8–6 km);The thickness of sedimentary layer in the Mongolia fold belt and Yinshan orogenic belt is relatively thin(less than 1 km).The crustal thickness of the study area varies between 27–48 km,of which the crust of the North China Plain is about 30–33 km,the central NCC is about 33–40 km,and the Ordos block is 40–48 km thick.The average v_(P)/v_(S)ratios in the study area is mostly between 1.66 and 1.90,and that in the Yanshan-Taihang mountain fold belt is between 1.70 and 1.85,and that in the Ordos block is between 1.65 and 1.90,with an average value of 1.77,indicating the absence of a thick basaltic lower crust.The obvious negative correlation between crustal thickness and average v_(P)/v_(S)ratio within Ordos and Central Asia orogenic belt may be related to magmatic underplating during the crustal formation.There is no significant correlation between the crustal thickness and the v_(P)/v_(S)ratio in the Lüliang-Taihang mountain fold belt,which may be related to the multiple geological processes such as underplating and crustal extension and thinning in this area.The lack of correlation between crust thickness and topography in the central orogenic belt and the North China Basin indicates the topography of these areas are controlled not only by crustal isostatic adjustment but also by the lithospheric mantle processes.

Electrochemical behavior of Ti-6.5Al-2Zr-1Mo-1V alloy under rotating condition with periodically fluctuating current density

Titanium alloy plays a crucial role in the electrochemical field due to its excellent corrosion resistance.The passivation and dissolution behaviors of Ti-6.5Al-2Zr-1Mo-1V(TA15)alloy in Na Cl solution were studied by simulating the electrochemical machining process in a rotating condition,which made the anode in a state with alternating high and low current density.Electron probe micro analysis,ultra-depth microscope,scanning electron microscope,and X-ray photoelectron spectrometer were used to reveal the evolution of TA15 under fluctuating current density.The existence state of the passivation film on TA15 surface was closely related to the pulse frequency of the periodically fluctuating current density.At higher pulse frequency of 0.20 Hz,the material was hardly dissolved because passivation dominated the electrolysis behavior,while at lower pulse frequency of 0.01 Hz,the passivation and dissolution behaviors occurred alternately with the variation of the current density.Herein,the thickness of the passivation film was inversely proportional to the applied current density.Due to the different electrochemical characteristics of a phase andβphase,the surface of the TA15 changed from being smooth to porous after a period.In addition,the change of microstructure affected the content of O^(2–)and exposed the suboxides of titanium.In a word,the change of pulse frequency and current density affected the electrochemical behavior of TA15,which was different from the conventional steady condition.

J-comb: An image fusion algorithm to combine observations covering different spatial frequency ranges

Ground-based, high-resolution bolometric(sub)millimeter continuum mapping observations on spatially extended target sources are often subject to significant missing fluxes. This hampers accurate quantitative analyses. Missing flux can be recovered by fusing high-resolution images with observations that preserve extended structures. However, the commonly adopted image fusion approaches do not maintain the simplicity of the beam response function and do not try to elaborate the details of the yielded beam response functions. These make the comparison of the observations at multiple wavelengths not straightforward. We present a new algorithm, J-comb, which combines the high and low-resolution images linearly. By applying a taper function to the low-pass filtered image and combining it with a high-pass filtered image using proper weights, the beam response functions of our combined images are guaranteed to have near-Gaussian shapes. This makes it easy to convolve the observations at multiple wavelengths to share the same beam response functions. Moreover, we introduce a strategy to tackle the specific problem that the imaging at 850 μm from the present-date ground-based bolometric instrument and that taken with the Planck satellite do not overlap in the Fourier domain. We benchmarked our method against two other widely-used image combination algorithms, CASA-feather and MIRIAD-immerge, with mock observations of star-forming molecular clouds. We demonstrate that the performance of the J-comb algorithm is superior to those of the other two algorithms. We applied the J-comb algorithm to real observational data of the Orion A star-forming region. We successfully produced dust temperature and column density maps with ~10′′angular resolution, unveiling much greater details than the previous results. A py Thon code release of J-comb and implementation of the algorithm are available at https://github.com/Sihan Jiao/J-comb.

梯度折射率超材料透镜

超材料是由人工设计的、具有天然材料所不具备的超常物理性质的人工复合结构,有负等效质量密度、负等效弹性模量、负折射率等特性.声子晶体超材料是一种具有周期性结构的超材料,其布拉格带隙或局域共振带隙的存在使处于禁带频率下的声波或弹性波进入声子晶体后无法在其中传播.其能带特性可以通过设计进行调节,使通带频率下的声波或弹性波进入具有特定结构的声子晶体后,实现成像、聚焦以及定向传输等对波传播的极端控制功能.梯度折射率超材料是一种折射率随空间变化而变化的结构.梯度折射率透镜由局部非均匀结构组成,其折射率是空间坐标的函数.波在透镜中会沿着弯曲的轨迹传播,可以通过适当的设计实现对波的弯曲、偏转和聚焦等功能.梯度折射率的设计可以通过改变局部晶体单元的性质实现有效折射率的梯度分布,如改变声子晶体的晶格尺寸、散射体的填充率、散射体的材料等.此外,梯度折射率超材料透镜能够通过设计在宽频率范围内精准聚焦波,从而实现能量采集等工程应用.本文从光学理论原理、透镜设计及其应用三方面展开介绍,首先对理论基础以及目前构建梯度折射率器件的几种方法进行总结,然后从透镜的应用角度介绍几种典型案例,最后对透镜的未来研究作出展望.

Evolution of convex structure during counter-rotating electrochemical machining based on kinematic modeling

Counter-rotating electrochemical machining(CRECM)is a novel electrochemical machining(ECM)method,which can be used to machine convex structures with complex shapes on the outer surface of casings.In this study,the evolution of the convex structure during CRECM is studied.The complex motion form of CRECM is replaced by an equivalent kinematic model,in which the movement of the cathode tool is realized by matrix equations.The trajectory of the cathode tool center satisfies the Archimedes spiral equation,and the feed depth in adjacent cycles is a constant.The simulation results show that the variations of five quality indexes for the convex structure:as machining time increases,the height increases linearly,and the width reduces linearly,the fillets at the top and root fit the rational function,and the inclination angle of the convex satisfies the exponential function.The current density distributions with different rotation angles is investigated.Owing to the differential distribution of current density on workpiece surface,the convex is manufactured with the cathode window transferring into and out of the processing area.Experimental results agree very well with the simulation,which indicates that the proposed model is effective for prediction the evolution of the convex structure in CRECM.

Optimizing the cleanliness in multi-segment disk amplifiers based on vector flow schemes

The objective of maintaining the cleanliness of the multi-segment disk amplifier in Shenguang-II(SG-II) is to reduce laser-induced damage for optics. The flow field of clean gas, which is used for the transportation of contaminant particles,is a key factor affecting the cleanliness level in the multi-segment disk amplifier. We developed a gas–solid coupling and three-dimensional flow numerical simulation model. The three-dimensional and two-phase flow model is verified by the flow-field smog experiment and the particle concentration measurement experiment with the 130-disk amplifier in SG-II. By optimizing the boundary conditions with the same flow rate, the multi-inlet vector flow scheme can not only effectively reduce the purging time, but also prevent the reverse diffusion of contaminant particles in the multi-segment disk amplifier and the deposition of contaminant particles on the surface of the Nd:glass.

The TMRT K band observations towards 26 infrared dark clouds:NH_3, CCS, and HC_3N

We present one of the first Shanghai Tian Ma Radio Telescope (TMRT) K Band observations towards a sample of 26 infrared dark clouds (IRDCs).We observed the (1,1),(2,2),(3,3),and (4,4) transitions of NH_(3) together with CCS (2_(1)–>1_(0)) and HC_(3)NJ=2-1,simultaneously.The survey dramatically increases the existing CCS-detected IRDC sample from 8 to 23,enabling a better statistical study of the ratios of carbon-chain molecules (CCM) to N-bearing molecules in IRDCs.With the newly developed hyperfine group ratio (HFGR) method of fitting NH_(3) inversion lines,we found the gas temperature to be between 10 and18 K.The column density ratios of CCS to NH_(3) for most of the IRDCs are less than 10^(-2),distinguishing IRDCs from low-mass star-forming regions.We carried out chemical evolution simulations based on a three-phase chemical model NAUTILUS.Our measurements of the column density ratios between CCM and NH_(3) are consistent with chemical evolutionary ages of 10^(5) yr in the models.Comparisons of the data and chemical models suggest that CCS,HC_(3)N,and NH_(3) are sensitive to the chemical evolutionary stages of the sources.