基于多池化融合与背景消除网络的跨数据集行人再识别方法

现有跨数据集行人再识别方法一般致力于减小2个数据集之间的数据分布差异,忽略了背景信息对识别性能的影响。针对上述问题,提出了一种基于多池化融合与背景消除网络的跨数据集行人再识别方法。为了兼顾全局特征和局部特征,同时实现特征的多细粒度表示,构建了多池化融合网络。为了使监督网络能提取有用的行人前景特征,构建了特征级有监督背景消除网络。采用结合行人分类损失及特征激活损失的多任务学习损失函数,在3个公开行人再识别数据集上对方法进行评估,当MSMT17作为训练集时,Market-1501上的跨数据集识别性能mAP为35.53%,相比ResNet50网络提升了9.24%;DukeMTMC-reID上的跨数据集识别性能m AP为41.45%,相比于ResNet50网络提升了10.72%。与现有方法相比,所提方法具有更优的跨数据集行人再识别性能。

融合词性语义扩展信息的事件检测模型

事件检测是事件抽取中的关键步骤,依赖于触发词进行事件类型分类。现有主流事件检测方法在稀疏标记数据上性能较差,模型过度拟合密集标注的触发词,在稀疏标记的触发词或者未见过的触发词上容易失效。改进方法通常通过扩充更多训练实例来缓解这一问题,但扩充后的数据分布不平衡,存在内置偏差,仍然表现不佳。为此,建立一种融合词性语义扩展信息的事件检测模型。对词粒度扩展信息进行分析,在不增加训练实例的条件下缩小候选触发词的范围,并对候选触发词进行语义扩展,挖掘候选触发词的上下文中蕴含的丰富语义,缓解了标记数据稀疏造成模型训练不充分的情况。通过词性筛选模块寻找候选触发词并对其进行语义扩展挖掘词粒度语义信息,融合句子粒度语义信息提升语义表征的鲁棒性,最终利用Softmax分类器进行分类完成事件检测任务。实验结果表明,该模型在ACE2005和KBP2015数据集上的事件检测任务中的F1值分别达到79.5%和67.5%,有效提升了事件检测性能,并且在稀疏标记数据实验中的F1值达到78.5%,明显改善了标记数据稀疏带来的不良影响。

High-resolution reconstruction of the ablative RT instability flowfield via convolutional neural networks

High-resolution flow field data has important applications in meteorology,aerospace engineering,high-energy physics and other fields.Experiments and numerical simulations are two main ways to obtain high-resolution flow field data,while the high experiment cost and computing resources for simulation hinder the specificanalysis of flow field evolution.With the development of deep learning technology,convolutional neural networks areused to achieve high-resolution reconstruction of the flow field.In this paper,an ordinary convolutional neuralnetwork and a multi-time-path convolutional neural network are established for the ablative Rayleigh-Taylorinstability.These two methods can reconstruct the high-resolution flow field in just a few seconds,and further greatlyenrich the application of high-resolution reconstruction technology in fluid instability.Compared with the ordinaryconvolutional neural network,the multi-time-path convolutional neural network model has smaller error and canrestore more details of the flow field.The influence of low-resolution flow field data obtained by the two poolingmethods on the convolutional neural networks model is also discussed.

Catalytically altering the redox pathway of sulfur in propylene carbonate electrolyte using dual-nitrogen/oxygen-containing carbon

Carbonate electrolytes are one of the most desirable electrolytes for high-energy lithium-sulfur batteries(LSBs)because of their successful implementation in commercial Li-ion batteries.The low-polysulfide-solubility feature of some carbonate solvents also makes them very promising for overcoming the shuttle effects of LSBs.However,regular sulfur electrodes experience undesired electrochemical mechanisms in carbonate electrolytes due to side reactions.In this study,we report a catalytic redox mechanism of sulfur in propylene carbonate(PC)electrolyte based on a compari-son study.The catalytic mechanism is characterized by the interactions between polysulfides and dual N/O functional groups on the host carbon,which largely prevents side reactions between polysulfides and the carbonate electrolyte.Such a mechanism coupled with the low-polysulfide-solubility feature leads to stable cycling of LSBs in PC electrolyte.Favorable dual N/O functional groups are identified via a density functional theory study.This work provides an alternative route for enabling LSBs in carbonate electrolytes.

基于语言模型及循环卷积神经网络的事件检测

目前,事件检测的难点在于一词多义和多事件句的检测.为了解决这些问题,提出了一个新的基于语言模型的带注意力机制的循环卷积神经网络模型(recurrent and convolutional neural network with attention based on language models,LM-ARCNN).该模型利用语言模型计算输入句子的词向量,将句子的词向量输入长短期记忆网络获取句子级别的特征,并使用注意力机制捕获句子级别特征中与触发词相关性高的特征,最后将这两部分的特征输入到包含多个最大值池化层的卷积神经网络,提取更多上下文有效组块.在ACE2005英文语料库上进行实验,结果表明,该模型的 F 1 值为74.4%,比现有最优的文本嵌入增强模型(DEEB)高0.4%.

基于轻量级神经网络的雷暴云团识别系统研究与实现

现阶段关于雷暴天气等气象信息都由气象工作人员进行确认识别,这种人工识别的方式无法满足对天气预警的实时性要求,给人们的人身财产安全保护带来了极大的挑战。为了解决上述问题,文中提出了一种将深度可分离卷积和残差块相结合,并利用多池化结构对全局信息进行特征提取的网络结构,通过降低网络的参数数量,提升雷暴云团识别的响应时间,且不会对识别的准确率造成影响。实验表明,利用这种网络结构对比常用的网络分割网络有优势,对雷暴云团的识别准确率提升了0.5%,并且对于卫星云图的响应时间缩短了10%;同时为了便于一些非相关的人员使用,对数据解析和网络结构进行了实用性和系统性的封装。

Pool boiling performance of porous surface tubes under vacuum conditions

Two types of tube bundles are designed,which are,respectively,composed of six tubes arranged in the boiling chamber.The nucleate pool boiling performance of smooth tube bundles and sintered porous surface tube bundles with deionized water as a medium are experimentally studied at atmospheric and sub-atmospheric pressures,respectively.The experimental results indicate that the boiling heat transfer coefficients of the two types of tube bundles increase with the increase in pressure under vacuum conditions as they behave under ordinary pressure.As the pressure varies from 10 to 100 kPa,it also can be seen that the heat transfer coefficient of the sintered porous surface tube is increased by 0.2 to 4 times compared with the smooth one under the same operating parameters.In addition,the experimental data show that a definite bundle effect exists in both sintered porous surface tubes and smooth tubes under vacuum conditions.

Synthesis of a novel hexagonal porous TT-Nb2O5 via solid state reaction for high-performance lithium ion battery anodes

Hexagonal porous Nb2O5 was synthesized for the first time via a facile solid-state reaction.The structure and electrochemical properties have been optimized through tuning heating temperature.X-ray diffraction results indicate that pseudo hexagonal Nb2O5(TT-Nb2O5)and orthorhombic Nb2O5 have been synthesized at different temperatures.Hexagonal sheet and porous structure of Nb2O5 were characterized by scanning electron microscopy and N2-adsorption-desorption isotherms.The as-prepared TT-Nb2O5(heated at 600℃)shows the best performance with a remarkable charge capacity of 178 mA∙h/g at 0.2C,which is higher than that of T-Nb2O5.Even at 20℃,TT-Nb2O5 offers unprecedented rate capability up to 86 mA∙h/g.The high rate capacity is due to pseudocapacitive Li+intercalation mechanism of TT-Nb2O5.The reported results demonstrate that Nb2O5 with good crystal structure and high specific surface area is a powerful composite design for high-rate and safe anode materials.

Preparation of a novel supported electrode comprising a nickel(Ⅱ)hydroxide-modified carbon paste electrode(Ni(OH)_2-X/CPE)for the electrocatalytic oxidation of formaldehyde

We prepared a novel nickel（II）hydroxide-modified carbon paste electrode（Ni（OH）2-X/CPE）for the electrocatalytic oxidation of formaldehyde.The electrode was prepared by a simple method without the use of linking chemicals.The prepared Ni（OH）2-X/CPE material was characterized by scanning electron microscopy and energy dispersive X-ray spectrometry.The electrochemical performance of the proposed electrode was investigated using cyclic voltammetry,electrochemical impedance spectroscopy,and chronoamperometry.The results indicate that Ni（OH）2-X/CPE exhibits good electrocatalytic activity with regards to formaldehyde oxidation owing to its nanoporous structure and the large surface area of zeolite X.The values of the electron transfer coefficient and the catalytic rate constant were 0.7 and 6.1 × 104 cm3/（mol·s）,respectively.Therefore,the proposed electrode,which showed remarkable electroactivity with regards to formaldehyde oxidation with long-term stability and good reproducibility,could be useful in fuel cells.

Enhanced Pool Boiling Heat Transfer on Copper Foam Welded Surfaces

Enhanced pool boiling heat transfer of the porous structure is critical to the thermal management technology.In this paper,pool boiling heat transfer experiments are performed on copper foam welded surfaces in de-ionized water to investigate the effects of basic parameters of copper foam on heat transfer enhancement.Boiling phenomenon is observed to facilitate the understanding of enhancement mechanism.The results show that copper foam welded surfaces can significantly enhance the pool boiling heat transfer performance,reduce the boiling incipience temperature by 7-9℃,and reach two times heat transfer coefficient compared with smooth plain surfaces due to numerous nucleation sites,extended surface areas,and enhanced turbulent effect.Pore density and thickness of foam have two side effects on heat transfer.

In-situ formation of cobalt phosphide nanoparticles confined in three-dimensional porous carbon for high-performing zinc-air battery and water splitting

The rational design of efficient and stable carbon-based electrocatalysts for oxygen reduction and oxygen evolution reactions is crucial for improving energy density and long-term stability of rechargeable zinc-air batteries(ZABs).Herein,a general and controllable synthesis method was developed to prepare three-dimensional(3D)porous carbon composites embedded with diverse metal phosphide nanocrystallites by interfacial coordination of transition metal ions with phytic acid-doped polyaniline networks and subsequent pyrolysis.Phytic acid as the dopant of polyaniline provides favorable anchoring sites for metal ions owing to the coordination interaction.Specifically,adjusting the concentration of adsorbed cobalt ions can achieve the phase regulation of transition metal phosphides.Thus,with abundant cobalt phosphide nanoparticles and nitrogen-and phosphorus-doping sites,the obtained carbon-based electrocatalysts exhibited efficient electrocatalytic activities toward oxygen reduction and evolution reactions.Consequently,the fabricated ZABs exhibited a high energy density,high power density of 368 mW cm^(-2),and good cycling/mechanical stability,which could power water splitting for integrated device fabrication with high gas yields.

Inverted Organic Solar Cells with Improved Performance using Varied Cathode Buffer Layers

Organic solar cells with inverted planar heterojunction structure based on subphthalocya- nine and C60 were fabricated using several kinds of materials as cathode buffer layer （CBL）, including tris-8-hydroxy-quinolinato aluminum （Alq3）, bathophenanthroline （Bphen）, bathocuproine, 2,3,8,9,14,15-hexakis-dodecyl-sulfanyl-5,6,11,12,17,18-hexaazatrinaphthylene （HATNA）, and an inorganic compound of Cs2CO3. The influence of the lowest unoccupied molecular orbital level and the electron mobility of organic CBL on the solar cells perfor- mance was compared. The results showed that Alq3, Bphen, and HATNA could significantly improve the device performance. The highest efficiency was obtained from device with an- nealed HATNA as CBL and increased for more than 7 times compared with device without CBL. Furthermore~ the simulation results with space charge-limited current theory indicated that the Schottky barrier at the organic/electrode interface in inverted OSC structure was reduced for 27% by inserting HATNA CBL.

MF~2ResU-Net:a multi-feature fusion deep learning architecture for retinal blood vessel segmentation

Objective For computer-aided Chinese medical diagnosis and aiming at the problem of insufficient segmentation,a novel multi-level method based on the multi-scale fusion residual neural network(MF2ResU-Net)model is proposed.Methods To obtain refined features of retinal blood vessels,three cascade connected UNet networks are employed.To deal with the problem of difference between the parts of encoder and decoder,in MF2ResU-Net,shortcut connections are used to combine the encoder and decoder layers in the blocks.To refine the feature of segmentation,atrous spatial pyramid pooling(ASPP)is embedded to achieve multi-scale features for the final segmentation networks.Results The MF2ResU-Net was superior to the existing methods on the criteria of sensitivity(Sen),specificity(Spe),accuracy(ACC),and area under curve(AUC),the values of which are 0.8013 and 0.8102,0.9842 and 0.9809,0.9700 and 0.9776,and 0.9797 and 0.9837,respectively for DRIVE and CHASE DB1.The results of experiments demonstrated the effectiveness and robustness of the model in the segmentation of complex curvature and small blood vessels.Conclusion Based on residual connections and multi-feature fusion,the proposed method can obtain accurate segmentation of retinal blood vessels by refining the segmentation features,which can provide another diagnosis method for computer-aided Chinese medical diagnosis.

Multi-objective steady-state optimization of two-chamber microbial fuel cells

A microbial fuel cell(MFC)is a novel promising technology for simultaneous renewable electricity generation and wastewater treatment.Three non-comparable objectives,i.e.power density,attainable current density and waste removal ratio,are often conflicting.A thorough understanding of the relationship among these three conflicting objectives can be greatly helpful to assist in optimal operation of MFC system.In this study,a multiobjective genetic algorithm is used to simultaneously maximizing power density,attainable current density and waste removal ratio based on a mathematical model for an acetate two-chamber MFC.Moreover,the level diagrams method is utilized to aid in graphical visualization of Pareto front and decision making.Three biobjective optimization problems and one three-objective optimization problem are thoroughly investigated.The obtained Pareto fronts illustrate the complex relationships among these three objectives,which is helpful for final decision support.Therefore,the integrated methodology of a multi-objective genetic algorithm and a graphical visualization technique provides a promising tool for the optimal operation of MFCs by simultaneously considering multiple conflicting objectives.

Preparation of PtNi/C Electrocatalysts for Direct Methanol Fuel Cell via Pulse Microwave Assisted Polyol Method

PtNi/C nanoparticles with different atomic ratios of Pt/Ni were produced in pulse microwave assisted polyol process. Transmission electron microscopy(TEM) images show uniform morphology. X-ray diffraction(XRD) pattern plus energy dispersive X-ray(EDX) spectroscopy suggests pure composition. Cyclic voltammogram study reveals that PtNi/C nanoparticles synthesized in pulse microwave assisted polyol process have better catalytic activity for the oxidation of methanol to carbon dioxide than those synthesized in continuous process.

Enhancing zinc storage performance of Mn_(3)O_(4)cathode through Ag-doping and-crosslinking dual-modification strategy

Octahedral Mn_(3)O_(4)nanoparticles with an Ag-doping and nanoporous Ag(NPS)framework was simply fabricated through an alloying-etching engineering.The dual-modified Mn_(3)O_(4)(denoted as Ag−Mn_(3)O_(4)/NPS)consists of Ag-doped Mn_(3)O_(4)nanoparticles crosslinked with three dimensional nanoporous Ag framework.The incorporated Ag dopant is effective in improving the intrinsic ionic and electronic conductivities of Mn_(3)O_(4),while the NPS framework is introduced to improve the electron/mass transfer across the entire electrode.Profiting from the dual-modification strategy,the Ag−Mn_(3)O_(4)/NPS exhibits admirable rate capability and cycling stability.A high reversible capacity of 88.7 mA·h/g can still be retained for over 1000 cycles at a current density of 1 A/g.Moreover,a series of ex-situ experimental techniques indicate that for Ag−Mn_(3)O_(4)/NPS electrode during the zinc ion storage,Mn_(3)O_(4)is electrochemically oxidized into various MnOx(e.g.,Mn_(2)O_(3),MnO2)species in the initial charging,and the subsequent battery reaction is actually the intercalation/deintercalation of H+and Zn2+into MnOx.

Effect of cations in ionic liquids on the electrochemical performance of lithium-sulfur batteries

Lithium-sulfur(Li-S) battery is a promising choice for the next generation of high-energy rechargeable batteries, but its application is impeded by the high dissolution of the polysulfides in commonly used organic electrolyte. Room temperature ionic liquids(RTILs) have been considered as appealing candidates for the electrolytes in Li-S batteries. We investigated the effect of cations in RTILs on the electrochemical performance for Li-S batteries. Ex situ investigation of lithium anode for Li-S batteries indicates that during the discharge/charge process the RTIL with N-methyl-N-propylpyrrolidine cations(P13) can effectively suppress the dissolution of the polysulfides, whereas the RTIL with 1-methyl-3-propyl imidazolium cation(PMIM) barely alleviates the shuttling problem. With 0.5 mol L-1 LiTFSI/P13 TFSI as the electrolyte of Li-S battery, the ketjen black/ sulfur cathode material exhibits high capacity and remarkable cycling stability, which promise the application of the P13-based RTILs in Li-S batteries.

Polyaniline-modified cetyltrimethylammonium bromide- graphene oxide-sulfur nanocomposites with enhanced performance for lithium-sulfur batteries

Conductive polymer coatings can boost the power storage capacity of lithiumsulfur batteries. We report here on the design and preparation--by combining a facile and green chemical deposition method with an oxidative polymerization approach--of polyaniline （PANI）-modified cetyltrimethylammonium bromide （CTAB）-graphene oxide （GO）-sulfur （S） nanocomposites with significantly enhanced performance in lithium-sulfur batteries. Such conductive polymer modified CTAB-GO-S nanocomposites as sulfur cathode materials can deliver high specific discharge capacities and long-term cycling performance, i.e., -970 mAh-g-1 at 0.2 C and -715 mAh-g-1 after 300 cycles, -820 mAh.g-1 at 0.5 C and -670 mAh.g-1 after 500 cycles, -770 mAh.K at 1 C and -570 mAh.g-~ after 500 cycles. The capacity decay was as low as 0.036% per cycle at 0.5 C, and 0.051% per cycle at 1 C. Under the same condition, batteries using PANI-modified CTAB-GO-S as cathodes exhibited higher specific capacity and higher average coulombic efficiency compared with CTAB-decorated GO-S and GO--S nano- composites. The improved performance can be attributed to the lower charge transfer resistance and the alleviated dissolution of polysulfides in the PANI- modified CTAB-GO-S cathodes.

Porous honeycomb-like C3N4/rGO composite as host for high performance Li-S batteries

Lithium-sulfur (Li-S) batteries have attracted extensive attention along with the urgent increasing demand for energy storage owing to the high theoretical specific capacity and energy density, abundant reserves and low cost of sulfur. However, the practical application of Li-S batteries is still impeded due to the low utilization of sulfur and serious shuttle-effect of lithium polysulfides (LiPSs). Here, we fabricated the porous honeycomb-like C3N4 (PHCN) through a hard template method. As a polar material, graphitic C3N4 has abundant nitrogen content (-58%), which can provide enough active sites to mitigate shuttle-effect, and then conductive reduced graphene oxide (rGO) was introduced to combine with PHCN to form PHCN/rGO composite in order to improve the utilization efficiency of sulfur. After sulfur loading, the PHCN/rGO/S cathode exhibited an initial discharge capacity of 1,061.1 mA h g^-1 at 0.2 C and outstanding rate performance at high current density of 5 C (495.1 mA h g^-1), and also retained 519 mA h g^-1, after 400 cycles at 1 C. Even at high sulfur loading (4.3 mg cm^-2), the capacity fade rate was only 0.16% per cycle at 0.5 C for 200 cycles. The above results demonstrate that the special design of PHCN/rGO composite as sulfur host has high potential application for Li-S rechargeable batteries.

Three-dimensional porous graphene-metal oxide composite microspheres： Preparation and application in Li-ion batteries

The use of new three-dimensional （3D） porous graphene-metal oxide composite microspheres as an anode material for Li-ion batteries （LIBs） is first introduced here. 3D graphene microspheres are aggregates of individual hollow graphene nanospheres composed of graphene sheets. Metal oxide nanocrystals are uniformly distributed over the graphene surface of the microspheres. The 3D porous graphene-SnO2 microspheres are selected as the first target material for investigation because of their superior electrochemical properties. The 3D porous graphene-SnO2 and graphene microspheres and bare SnO2 powders deliver discharge capacities of 1,009, 196, and 52 mAh·g^-1, respectively, after 500 cycles at a current density of 2 A·g^-1 .The 3D porous graphene-SnO2 microspheres exhibit uniquely low charge transfer resistances and high Li-ion diffusivities before and after cycling.