Dual-Path Vision Transformer用于急性缺血性脑卒中辅助诊断

急性缺血性脑卒中是由于脑组织血液供应障碍导致的脑功能障碍,数字减影脑血管造影(DSA)是诊断脑血管疾病的金标准。基于患者的正面和侧面DSA图像,对急性缺血性脑卒中的治疗效果进行分级评估,构建基于Vision Transformer的双路径图像分类智能模型DPVF。为了提高辅助诊断速度,基于EdgeViT的轻量化设计思想进行了模型的构建;为了使模型保持轻量化的同时具有较高的精度,提出空间-通道自注意力模块,促进Transformer模型捕获更全面的特征信息,提高模型的表达能力;此外,对于DPVF的两分支的特征融合,构建交叉注意力模块对两分支输出进行交叉融合,促使模型提取更丰富的特征,从而提高模型表现。实验结果显示DPVF在测试集上的准确率达98.5%,满足实际需求。

An aqueous BiI_(3)-Zn battery with dual mechanisms of Zn^(2+)(de)intercalation and I^(-)/I_(2)redox

The development of aqueous battery with dual mechanisms is now arousing more and more interest.The dual mechanisms of Zn^(2+)(de)intercalation and I^(-)/I_(2)redox bring unexpected effects.Herein,differing from previous studies using Zn I_(2)additive,this work designs an aqueous Bi I_(3)-Zn battery with selfsupplied I^(-).Ex situ tests reveal the conversion of Bi I_(3)into Bi(discharge)and Bi OI(charge)at the 1st cycle and the dissolved I^(-)in electrolyte.The active I^(-)species enhances the specific capacity and discharge medium voltage of electrode as well as improves the generation of Zn dendrite and by-product.Furthermore,the porous hard carbon is introduced to enhance the electronic/ionic conductivity and adsorb iodine species,proven by experimental and theoretical studies.Accordingly,the well-designed Bi I_(3)-Zn battery delivers a high reversible capacity of 182 m A h g^(-1)at 0.2 A g^(-1),an excellent rate capability with 88 m A h g^(-1)at 10 A g^(-1),and an impressive cyclability with 63%capacity retention over 20 K cycles at 10 A g^(-1).An excellent electrochemical performance is obtained even at a high mass loading of 6 mg cm^(-2).Moreover,a flexible quasi-solid-state Bi I_(3)-Zn battery exhibits satisfactory battery performances.This work provides a new idea for designing high-performance aqueous battery with dual mechanisms.

Dual-Delay-Path Ring Oscillator with Self-Biased Delay Cells for Clock Generation

This work summarizes the structure and operating features of a high-performance 3-stage dual-delay-path (DDP) voltage-controlled ring oscillator (VCRO) with self-biased delay cells for Phase-Locked Loop (PLL) structurebased clock generation and digital system driving. For a voltage supply VDD = 1.8 V, the resulting set of performance parameters include power consumption PDC = 4.68 mW and phase noise PN@1MHz = -107.8 dBc/Hz. From the trade-off involving PDC and PN, a system level high performance is obtained considering a reference figure-of-merit ( FoM = -224 dBc/Hz ). Implemented at schematic level by applying CMOS-based technology (UMC L180), the proposed VCRO was designed at Cadence environment and optimized at MunEDA WiCkeD tool.

“30-60”双碳背景下新疆电解铝产业发展的风险识别与建议

近年来,凭借新疆丰富的能源优势,电解铝产业得到迅猛发展,并逐步形成了煤电铝一体化的发展格局。电解铝是高耗能、高碳排放行业,每生产1 t电解铝的碳素材料消耗量约为0.4 t,而其所产生的二氧化碳排放量则达到1.8 t。因此,必须采取相应措施来实现碳中和的目标。本文主要阐述了新疆地区电解铝行业主要存在的能源结构单一、交通运输成本高、电解铝产业链不完善等问题,识别并分析了碳达峰碳中和背景下新疆地区电解铝产业发展面临的挑战和风险,有针对性提出了新疆地区电解铝产业绿色低碳发展的路径与建议,为新疆电解铝产业实现跨越式、高质量发展提供参考。

基于双路径递归网络与Conv-TasNet的多头注意力机制视听语音分离

目前的视听语音分离模型基本是将视频特征和音频特征进行简单拼接,没有充分考虑各个模态的相互关系,导致视觉信息未被充分利用,分离效果不理想。该文充分考虑视觉特征、音频特征之间的相互联系,采用多头注意力机制,结合卷积时域分离模型(Conv-TasNet)和双路径递归神经网络(DPRNN),提出多头注意力机制时域视听语音分离(MHATD-AVSS)模型。通过音频编码器与视觉编码器获得音频特征与视频的唇部特征,并采用多头注意力机制将音频特征与视觉特征进行跨模态融合,得到融合视听特征,将其经DPRNN分离网络,获得不同说话者的分离语音。利用客观语音质量评估(PESQ)、短时客观可懂度(STOI)及信噪比(SNR)评价指标,在VoxCeleb2数据集进行实验测试。研究表明,当分离两位、3位或4位说话者的混合语音时,该文方法与传统分离网络相比,SDR提高量均在1.87 dB以上,最高可达2.29 dB。由此可见,该文方法能考虑音频信号的相位信息,更好地利用视觉信息与音频信息的相关性,提取更为准确的音视频特性,获得更好的分离效果。

基于特征调节器和双路径引导的RGB-D室内语义分割

针对室内场景图像语义分割结果不精确、显著图粗糙的问题,提出一种基于多模态特征优化提取和双路径引导解码的网络架构(feature regulator and dual-path guidance,FG-Net)。具体来说,设计的特征调节器对每个阶段的多模态特征依次进行噪声过滤、重加权表示、差异性互补和交互融合,通过强化RGB和深度特征聚合,优化特征提取过程中的多模态特征表示。然后,在解码阶段引入特征交互融合后丰富的跨模态线索,进一步发挥多模态特征的优势。结合双路径协同引导结构,在解码阶段融合多尺度、多层次的特征信息,从而输出更细致的显著图。实验在公开数据集NYUD-v2和SUN RGB-D上进行,在主要评价指标mIoU上达到48.5%,优于其他先进算法。结果表明,该算法实现了更精细的室内场景图像语义分割,表现出了较好的泛化性和鲁棒性。

Development of Dual-light Path Monitoring System of Optical Thin-film Thickness

The accurate monitoring of optical thin-film thickness is a key technique for depositing optical thin-film. For existing coating equipments, which are low precision and automation level on monitoring thin-film thickness, a new photoelectric control and analysis system has been developed. In the new system, main techniques include a photoelectric system with dual-light path, a dual-lock-phase circuit system and a comprehensive digital processing-control-analysis system.The test results of new system show that the static and dynamic stabilities and the control precision of thin-film thickness are extremely increased. The standard deviation of thin-film thickness, which indicates the duplication of thin-film thickness monitoring, is equal to or less than 0.72%. The display resolution limit on reflectivity is 0.02 %. In the system, the linearity of drift is very high, and the static drift ratio approaches zero.

融合双路CNN-LSTM与注意力机制的语音情感识别模型

针对现有以卷积神经网络为基础的语音情感识别方法存在特征提取不足、模型识别效果不佳等问题,提出融合双路CNN-LSTM与注意力机制的语音情感识别模型。模型采用双路多维多尺度特征提取方法,结合残差块、多尺度卷积提取MFCC、Chroma和语谱图深层特征,增加特征多样性;采用注意力机制,分别计算双路特征的自注意力与交叉注意力参数,分配不同权重系数并进行加权融合,综合互补信息,减少特征冗余影响;采用LSTM网络提取时序特征,获取上下文语义信息,采用Softmax函数在数据集RAVDESS与SEWA上的分类准确率分别为90.19%和89.23%。

Fuzzy Q learning algorithm for dual-aircraft path planning to cooperatively detect targets by passive radars

The problem of passive detection discussed in this paper involves searching and locating an aerial emitter by dualaircraft using passive radars. In order to improve the detection probability and accuracy, a fuzzy Q learning algorithrn for dual-aircraft flight path planning is proposed. The passive detection task model of the dual-aircraft is set up based on the partition of the target active radar＇s radiation area. The problem is formulated as a Markov decision process （MDP） by using the fuzzy theory to make a generalization of the state space and defining the transition functions, action space and reward function properly. Details of the path planning algorithm are presented. Simulation results indicate that the algorithm can provide adaptive strategies for dual-aircraft to control their flight paths to detect a non-maneuvering or maneu- vering target.

Dual-doping for enhancing chemical stability of functional anionic units in sulfide for high-performance all-solid-state lithium batteries

The sulfide-based solid-state electrolytes(SEs)reactivity toward moisture and Li-metal are huge barriers that impede their large-scale manufactu ring and applications in all-solid-state lithium batteries(ASSLBs).Herein,we proposed an Al and O dual-doped strategy for Li_(3)PS_(4)SE to regulate the chemical/electrochemical stability of anionic PS_(4)^(3-)tetrahedra to mitigate structural hydrolysis and parasitic reactions at the SE/Li interface.The optimized Li_(3.08)A_(10.04)P_(0.96)S_(3.92)O_(0.08)SE presents the highestσLi+of 3.27 mS cm^(-1),which is~6.8 times higher than the pristine Li_(3)PS_(4)and excellently inhibits the structural hydrolysis for~25 min@25%humidity at RT.DFT calculations confirmed that the enhanced chemical stability was revealed to the intrinsically stable entities,e.g.,POS33-units.Moreover,Li_(3.08)Al_(0.04)P_(0.96)S_(3.92)O_(0.08)SE cycled stably in Li//Li symmetric cell over 1000 h@0.1 mA cm^(-2)/0.1 mA h cm^(-2),could be revealed to Li-Al alloy and Li_(2)Oat SE/Li interface impeding the growth of Li-dendrites during cycling.Resultantly,LNO@LCO/Li_(3.08)Al_(0.04)P_(0.96)S_(3.92)O_(0.08)/Li-In cell delivered initial discharge capacities of 129.8 mA h g^(-1)and 83.74%capacity retention over 300 cycles@0.2 C at RT.Moreover,the Li_(3.08)Al_(0.04)P_(0.96)S_(3.92)O_(0.08)SE presented>90%capacity retention over 200 and 300 cycles when the cell was tested with LiNi_(0.8)Co_(0.15)Al_(0.05)O_(2)(NCA)cathode material vs.5 and 10 mg cm^(-2)@RT.

转变助产服务模式联合情绪-认知双路径干预对自然分娩初产妇的影响

目的:观察转变助产服务模式联合情绪-认知双路径干预对自然分娩初产产妇母乳喂养率、爱丁堡产后抑郁量表(EPDS)评分及母亲角色转换情况的影响。方法:回顾性分析两院2021年1-12月300例(对照组)与2022年1-12月280例自然分娩初产产妇(研究组),均行常规护理干预,对照组在其基础上延续传统助产服务模式,研究组则采取转变助产服务模式联合情绪-认知双路径干预,比较两组干预前后认知水平、产后EPDS评分、初产妇角色转变量表评分、母亲角色适应能力评分、母乳喂养情况(初始泌乳时间、产后3d母乳喂养率)、产妇满意度。结果:研究组干预后认知水平评分显著高于对照组(t=14.704,P<0.001);研究组干预1个月EPDS评分(t=19.686,P<0.001)、2个月EPDS评分(t=12.425,P<0.001)显著低于对照组,研究组干预1个月初产妇角色转变评分(t=11.537,P<0.001)、2个月初产妇角色转变评分(t=9.978,P<0.001)、干预1个月母亲角色适应能力评分(t=6.622,P<0.001)、2个月母亲角色适应能力评分(t=7.747,P<0.001)显著高于对照组;研究组初始泌乳时间显著短于对照组(t=16.877,P<0.001),产后3d母乳喂养率显著高于对照组(χ^(2)=28.738,P<0.001);研究组产妇满意度显著优于对照组(Z=42.385,P<0.001)。结论:转变助产服务模式联合情绪-认知双路径干预可有效改善自然分娩初产产妇产后抑郁情绪,提高母亲角色转换能力及母乳喂养率,获得产妇满意。

基于语义与全局双重注意力机制的长链非编码RNA-疾病关联预测模型

针对现有长链非编码RNA(lncRNA)-疾病关联预测模型在综合利用异构生物网络的交互、语义信息上存在局限性的问题,提出一种基于语义与全局双重注意力机制的lncRNA-疾病关联预测模型(SGALDA)。首先,基于相似性和已知关联构建一个lncRNA-疾病-微小RNA(miRNA)异构网络,并基于消息传递类型设计特征提取模块来提取和融合异构网络上同质、异质节点的邻域特征,以捕捉异构网络上的多层面交互关系。其次,基于元路径将异构网络分解为多个语义子网络,并分别在各个子网络上应用图卷积网络(GCN)来提取节点的语义特征,以捕捉异构网络上的高阶交互关系。然后,基于语义与全局双重注意力机制融合节点的语义和邻域特征,以获得更具代表性的节点特征。最后,利用lncRNA节点特征和疾病节点特征的内积运算重建lncRNA-疾病关联。5折交叉验证结果显示,SGALDA的受试者工作特征曲线下面积(AUROC)为0.9945±0.0002,PR曲线下面积(AUPR)为0.9167±0.0011,在所有对比模型中均为最高,验证了SGALDA良好的预测性能。对乳腺癌、胃癌的案例研究进一步证实了SGALDA识别潜在lncRNA-疾病关联的能力,说明SGALDA有潜力成为一种可靠的lncRNA-疾病关联预测模型。

Development Path of Urban-rural Integration

The urban and rural areas are regarded as two major components of the regional economic system. Only through joint balanced development of the two can we achieve overall economic optimization and social welfare maximization. But the great social division of labor has separated urban areas from rural areas,which casts the shadow of city-oriented theory on cooperative relations between urban and rural areas. Mutual separation between urban and rural settlements and independent development trigger off a range of social problems. We must undertake guidance through rational development path of urban-rural integration,to eliminate the phenomenon of urban-rural dual structure,and promote the sustainable development of population,resources and environment in urban and rural areas as soon as possible.

Novel dual-band antenna for multi-mode GNSS applications

A novel dual-band antenna is proposed for mitigating the multi-path interference in the global navigation satellite system（GNSS） applications. The radiation patches consist of a shortedannular-ring reduced-surface-wave（SAR-RSW） element and an inverted-shorted-annular-ring reduced-surface-wave（ISAR-RSW）element. One key feature of the design is the proximity-coupled probe feeds to increase impedance bandwidth. The other is the defected ground structure band rejection filters to suppress the interaction effect between the SAR-RSW and the ISAR-RSW elements. In addition, trans-directional couplers are used to obtain tight coupling. Measurement results indicate that the antenna has a larger than 10 d B return loss bandwidth and a less than 3 d B axial-ratio（AR） bandwidth in the range of（1.164 – 1.255） GHz and（1.552 – 1.610） GHz. The gain of the passive antenna in the whole operating band is more than 7 d Bi.

基于DP-DBNet和MHA-CRNN的船牌号检测与识别

船牌号的检测和识别对于港口的智能化管理和解决传统人工方式监管渔船中存在的耗时耗力的问题具有重要意义.针对船牌悬挂位置,背景颜色和字符个数不统一等特点,本文提出两阶段双模型的检测和识别方法.首先,提出将双路径网络(dual path networks, DPN)与可微二值化网络(differentiable binarization network, DBNet)相结合的DP-DBNet船牌号位置检测模型.其次,提出将多头注意力机制(multi-head-attention mechanism, MHA)与改进的卷积循环神经网络(convolutional recurrent neural network, CRNN)相结合的MHA-CRNN船牌号文字识别模型.最后,以烟台芝罘区新型现代化智慧渔港项目为数据来源,并进行算法对比实验分析;实验结果表明,两种模型结合的两阶段识别方法可以使船牌号的识别准确率达到76.39%,充分证明了该模型的有效性和在海洋港口管理方面的应用价值.

Preliminary research on the transmission path of nssSO_4^(2-) and NO_3^- in Antarctic ice sheet

The main sources of nssSO 2- 4 and NO - 3 were summarized in this paper. By analyzing the spatial distribution features of major ions in Antarctic ice sheet and studying on the different time of the same volcanic event recorded by different ice cores from different regions in Antarctica, this paper intends to study the transmission path of nssSO 2- 4 and NO - 3. Results show that nssSO 2- 4 and NO - 3 are transmitted to the ice sheet through long distance and high altitude. The procedure of the transmission is that nssSO 2- 4 and NO - 3 are transmitted to the level between the top of troposphere and the bottom of stratosphere, then subsided to the ice sheet surface and spread to other regions.

A Method for Identifying Channeling Paths in Low-Permeability Fractured Reservoirs

Often oilfield fractured horizontal wells produce water flowing in multiple directions.In this study,a method to identify such channeling paths is developed.The dual-medium model is based on the principle of inter-well connectivity and considers the flow characteristics and related channeling terms.The Lorentz curve is drawn to qualitatively discern the geological type of the low-permeability fractured reservoir and determine the channeling direction and size.The practical application of such an approach to a sample oilfield shows that it can accurately identify the channeling paths of the considered low-permeability fractured reservoir and predict production performances according to the inter-well connectivity model.As a result,early detection of water channeling becomes possible,paving the way to real-time production system optimization in low-permeability fractured reservoirs.

Bis(1-(4-(dimethylamino)benzylidene)-4-phenylthiose micarbaznato)palladium(Ⅱ):Synthesis,Crystal Structure and Its Dual Fluorescence Property

The title compound bis（1-（4-（dimethylamino）benzylidene）-4-phenylthiosemicar-bazato）-palladium（Ⅱ）（PdL2） was obtained by reacting 1-（4-（dimethylamino）benzylidene）-4-phenyl-thiosemicarbazide with dichlorobis（benzonitrile）palladium（Ⅱ） in methanol,and its structure was characterized by single-crystal X-ray diffraction.The crystal of PdL2 was obtained in dimethyl-formamide（DMF） solvent with solvent molecules involved in the cell and crystallizes in the monoclinic system,space group C2 with a = 18.485（15）,b = 7.090（5）,c = 17.595（11） ,β = 121.21（3）o,V = 1972（2） 3,Z = 2,Mr = 847.40,Dc = 1.427 g/cm3,μ = 0.624 mm-1,F（000） = 880,R = 0.0607 and wR = 0.1358.The Pd atom adopts a distorted square planar coordination geometry with two Pd-N and two Pd-S bonds.The ligand loses a proton from its tautomeric thiol form and coordinates to the Pd atom via mercapto sulfur and the imine nitrogen atom,which binds to palladium as bidentate N,S-donors forming five-membered chelate rings.The complex formed hydrogen bonding interaction with solvent DMF molecules from the hydrogen of phenylamine to the oxygen of DMF and several intramolecular hydrogen bonds.Pd（Ⅱ） perturbed ligand π-π＊ transition and metal-to-ligand charge transfer（MLCT） transition are observed in its electronic absorption spectra.The complex exhibits intraligand 1π-π＊（IL） state and MLCT state dual fluorescent emissions in organic solvent at room temperature.

2D Coupled 3D:A New Numerical Model for Dual - Structured - Aquifer System

This paper points out that a successful numerical simulation is to construct a correct conceptional model which is very dose to the natural condition. A new model, two dimensional coupled three dimensional model (2D -3D ) is presented in the Present paper,which is the most suitable one for the dual - structured - aquifer system. An example of Wenyinghu area is shown.By using the 2D-3D model, a satisfied result of the simulated area is achieved.

双路径合作的原型矫正小样本分类模型

基于度量的元学习在学习过程中存在由于稀缺数据分布导致习得的先验知识不足、从样本中提取到的单一视图特征易受弱相关或无关特征的干扰以及因分类造成的代表性特征偏差的问题。针对这些问题,提出了一种双路径合作的原型矫正小样本分类模型。首先,通过双路径合作模块从多视图角度自适应地突出关键特征和弱化弱相关特征,充分利用特征信息获得先验知识来提升特征的表达能力;其次,通过基于查询集样本特征信息的原型矫正分类策略来解决类内原型的偏差问题;最后,通过损失函数反向更新模型参数,模型分类准确率得以提升。在五个公开的数据集上进行了5-way 1-shot和5-way 5-shot对比实验,较基准模型而言,在miniImageNet数据集上,准确率提升了5.57个百分点和3.90个百分点;在tieredImageNet数据集上,准确率提升了5.68个百分点和3.93个百分点;在CUB数据集上,准确率提升了6.93个百分点和3.13个百分点;在CIFAR-FS数据集上,准确率提升了8.03个百分点和1.65个百分点;在FC-100数据集上,准确率提升了4.25个百分点和4.89个百分点。实验结果表明,提出的双路径合作的原型矫正小样本分类模型能在小样本学习领域有良好的性能,且模型中的模块可迁移到其他模型中使用。