Polarization control of above-threshold ionization spectrum in elliptically polarized two-color laser fields

We study the above-threshold ionization(ATI)process of atoms exposed to fundamental and high-frequency lasers with arbitrary ellipticity by applying the frequency-domain theory.It is found that the angular-resolved ATI spectrum is sensitive to ellipticities of two lasers and emitted angles of the photoelectron.Particularly for the photon energy of the highfrequency laser more than atomic ionization potential,the width of plateau tends to a constant with increasing ellipticity of fundamental field,the dip structure disappears with increasing ellipticity of the high-frequency field.With the help of the quantum channel analysis,it is shown that the angular distribution depends mainly on the ellipticity of high-frequency field in the case that its frequency is high.Moreover,one can see that the maximal and minimal energies in quantum numerical results are in good agreement with the classical prediction.Our investigation may provide theoretical support for experimental research on polarization control of ionization in elliptically polarized two-color laser fields.

Manipulating the electron dynamics in the non-sequential double ionization process of Ar atoms by an orthogonal two-color laser field

Electron dynamics during non-sequential double ionization(NSDI) is one of the most attractive areas of research in the field of laser–atom or laser–molecule interaction. Based on the classic two-dimensional model, we study the process of NSDI of argon atoms driven by a few-cycle orthogonal two-color laser field composed of 800 nm and 400 nm laser pulses. By changing the relative phase of the two laser pulses, a localized enhancement of NSDI yield is observed at 0.5πand 1.5π, which could be attributed to a rapid and substantial increase in the number of electrons returning to the parent ion within extremely short time intervals at these specific phases. Through the analysis of the electron–electron momentum correlations within different time windows of NSDI events and the angular distributions of emitted electrons in different channels, we observe a more pronounced electron–electron correlation phenomenon in the recollision-induced ionization(RII) channel. This is attributed to the shorter delay time in the RII channel.

Steering the energy sharing of electrons in nonsequential double ionization with orthogonally polarized two-color field

Using the semiclassical ensemble model,the dependence of relative amplitude for the recollision dynamics in nonsequential double ionization(NSDI)of neon atom driven by the orthogonally polarized two-color field(OTC)laser field is theoretically studied.And the dynamics in two typical collision pathways,recollision-impact-ionization(RII)and recollisionexcitation with subsequent ionization(RESI),is systematically explored.Our results reveal that the V-shaped structure in the correlated momentum distribution is mainly caused by the RII mechanism when the relative amplitude of the OTC laser field is zero,and the first ionized electrons will quickly skim through the nucleus and share few energy with the second electron.As the relative amplitude increases,the V-shaped structure gradually disappears and electrons are concentrated on the diagonal in the electron correlation spectrum,indicating that the energy sharing after electrons collision is symmetric for OTC laser fields with large relative amplitudes.Our studies show that changing the relative amplitude of the OTC laser field can efficiently control the electron–electron collisions and energy exchange efficiency in the NSDI process.

High-throughput sorting of two-color fluorescent-labeled zebrafish embryos

The zebrafish embryos were widely employed in genetics,development and drug discovery studies as miniatured animal models.Sorting of two-color fluorescent embryos is often required in large-scale experiments but it is challenging to manually sort with high efficiency.Here,we reported a high-throughput sorting system for two-color fluorescent zebraflsh embryos.The embryos can be automatically loaded from a sample pool and sorted based on the average fluorescent intensity.The two-color fluorescent signals were split into two lines and detected by an area array camera.The system achieves the sorting of 100 embryos in less than 10 min with an accuracy of greater than 95%.

Effect of laser focus in two-color synthesized waveform on generation of soft x-ray high harmonics

Synthesis of multi-color laser pulses has been developed as a promising way to improve low conversion efficiency of high-order harmonic generation(HHG). Here we systematically study the effect of laser focus in a two-color waveform on generation of macroscopic HHG in soft x-rays. We find that the dependence of HHG yields on laser focus at low or high gas pressure is sensitive to the characteristics of single-atom harmonic response, in which “short”-or “long”-trajectory emissions can be selectively controlled by changing the waveform of two-color synthesized laser pulse. We uncover the phase-matching mechanism of HHG in the gas medium by examining the propagation of the two-color waveform and the evolution of time-frequency emissions of high-harmonic field. We further reveal that the nonlinear effects, such as geometric phase, atomic dispersion, and plasma defocusing, are responsible for modification of two-color waveform upon propagation. This work can be used to find better macroscopic conditions for generating soft x-ray HHG by employing two-color optimized waveforms.

Generation of quasi-chirp-free isolated attosecond pulses from atoms under the action of orthogonal two-color combined pulse of fundamental frequency and higher intensity second harmonic fields

The intrinsic chirp of high-order harmonic generation is an important factor limiting the production of ultrashort attosecond pulses.Based on numerically solving the time-dependent Schrodinger equation,the generation process of highorder harmonic from the He atom under the action of orthogonal two-color combined pulse of fundamental frequency and higher intensity second harmonic fields is studied.In this paper,we propose to achieve quasi-chirp-free isolated attosecond pulses by superimposing a higher second-harmonic field on the orthogonal direction of the fundamental frequency field.It is found that the high-energy part of its harmonic emission exhibits small chirp characteristics,which can be used to synthesize isolated attosecond pulses.Through the analysis of the wave packets evolution and the classical motion trajectories of the electron,it is demonstrated that the quasi-chirp-free harmonic can be attributed to the simultaneous return of electrons ionized at different times to the parent particle.The influence of the relative phase of the two pulses on the harmonics is further analyzed,and it is observed that this phenomenon is sensitive to the relative phase,but it can still generate isolated attosecond pulses within a certain phase.

基于图常量条件函数依赖的图修复规则发现

数据一致性是数据质量管理的一个重要内容。为了提升图数据一致性,大量关系型数据库中的数据依赖理论被引入到图数据库,包括图函数依赖、图关联规则等。图修复规则是最新提出的一种针对图数据的数据依赖规则,具有强大的修复能力,但目前尚无有效的挖掘算法。为了自动生成图修复规则并提高图数据修复的可靠性,提出一种将图常量条件函数依赖转化为图修复规则的方法(GenGRR)。通过图模式在图中匹配同构子图并映射成节点-属性二维表,从表中相应属性域中抽取错误模式把图常量条件函数依赖转化成图属性值修复规则;删去图模式中常量条件函数依赖RHS对应的节点与相连边生成图属性补充规则。基于最大公共同构子图筛选并验证生成图修复规则的一致性。在多个真实数据集上进行测试,验证相比图常量条件函数直接修复图数据,通过转化生成的图修复规则具有更好的修复效果。

基于区域路标引导的月面大范围高效行驶导航技术

载人月球探测任务中,受月面环境的复杂性和车载系统配置的限制,仅依靠车载导航系统难以全自主实现精确导航,需要遥操作中心对载人车大范围移动进行智能化支持。针对载人车在复杂月面环境中进行远距离探测时的高效导航问题,提出了基于区域路标引导的月面大范围高效行驶导航方法,通过分析载人车导航相机成像区域,将全路线可视区域中的月坑构建为月面路标图;并利用载人车导航相机图像中的月坑构建导航相机路标图,使用子图匹配的方法确定载人车可视区域内的月坑和环月卫星影像中月坑的对应关系,从而完成载人车位姿的解算。仿真试验结果表明:提出方法可以实现大范围移动过程中的高效导航。

基于PathSim的MOOCs知识概念推荐模型

大规模开放在线课程提供大规模开放式在线学习平台,为推进现代教育发挥关键作用。然而,减少用户学习盲区和改善用户体验方面的研究仍具有挑战性:交互数据稀疏;难以扩展到大型推荐任务上;用户需求不单由用户喜好决定,还受到不同教师、课程影响;以统一的方式对课程学习事件中不同类型实体及关系进行建模并不妥靠。基于此,引入相关性度量,依据全图结构信息计算各边权重,提出采用相关性度量算法PathSim进行邻域采样的知识概念推荐模型PathSimSage。各实体间相关性得分可在本地离线计算,将神经网络与传播过程分离,保证神经网络的堆叠层数和传播过程的独立性,大幅减少模型所需训练时间。在公开的MoocCube数据集上进行了综合实验,PathSimSage降低了不相关的信息甚至噪声的影响,解决随机游走采样所引发的高度节点偏差问题,并在一定程度上缓解了过平滑效应。

航班延误特征可视分析方法

为分析航班延误发生规律,提出一种数学模型联合多视图协同的可视分析方法。对SEIR传染病传播模型进行调整,建立航班延误传播模型分析延误的传播特征,在此基础上,运用频繁子图挖掘算法提取延误频繁模式;设计基于中点分割的地图网络图、矩阵热力图和时序图,分析延误的时空分布特征;设计VA-FDC系统用于方法验证。实验结果表明,VA-FDC能够有效分析航班延误时空分布特征,依据航班延误传播模型准确描述延误传播特征,为相关部门有效措施的制定提供借鉴。

k阶采样和图注意力网络的知识图谱表示模型

知识图谱表示(KGE)旨在将知识图谱中的实体和关系映射到低维度向量空间而获得其向量表示。现有的KGE模型只考虑一阶近邻,这影响了知识图谱中推理和预测任务的准确性。为了解决这一问题,提出了一种基于k阶采样算法和图注意力网络的KGE模型。k阶采样算法通过聚集剪枝子图中的k阶邻域来获取中心实体的邻居特征。引入图注意力网络来学习中心实体邻居的注意力值,通过邻居特征加权和得到新的实体向量表示。利用ConvKB作为解码器来分析三元组的全局表示特征。在WN18RR、FB15k-237、NELL-995、Kinship数据集上的评价实验表明,该模型在链接预测任务上的性能明显优于最新的模型。此外,还讨论了阶数k和采样系数b的改变对模型命中率的影响。

基于复杂网络的突发事件下汽车产业供应链韧性研究

汽车产业供应链中,各企业间由于合作伙伴关系或者竞争关系而产生的交互作用影响整个供应链网络,加之其参与主体众多,更易受到各种突发事件的干扰影响网络韧性。文章以复杂网络模型为基础,收集中国经济金融研究数据库数据库汽车供应链板块中141家车企数据构建复杂网络关系矩阵模型,通过复杂网络特性分析对所构建的网络模型进行检验。采用目标干扰和随机干扰两种方式来模拟突发事件产生的供应链中断情景,由矩阵工厂和大型复杂网络分析工具对网络进行仿真实验,通过最大连通子图所表示的弹性性能和效率性能两项数据来观测网络稳定性,得出提升汽车产业供应链韧性的方法。

以子图融合为最小单位的混合精度推理

近几年卷积神经网络作为深度学习最重要的技术,在图像分类、物体检测、语音识别等领域均有所建树。在此期间,由多层卷积神经网络组成的深度神经网络横空出世,在各种任务准确性方面具有显著提升。然而,神经网络的权重往往被限定在单精度类型,使网络体积相较于特定硬件平台上的内存空间更大,且floating point 16、INT 8等单精度类型已无法满足现在一些模型推理的现实需求。为此,提出一种以子图为最小单位,通过判断相邻结点之间的融合关系,添加了丰富比特位的混合精度推理算法。首先,在原有单精度量化设计的搜索空间中增加floating point 16半精度的比特配置,使最终搜索空间变大,为寻找最优解提供更多机会。其次,使用子图融合的思想,通过整数线性规划将融合后的不同子图精度配置,根据模型大小、推理延迟和位宽操作数3个约束对计算图进行划分,使最后累积的扰动误差减少。最终,在ResNet系列网络上验证发现,所提模型精度相较于HAWQ V3的损失没超过1%的同时,相较于其他混合精度量化方法在推理速度方面得到了提升,在ResNet18网络中推理速度分别提升18.15%、19.21%,在ResNet50网络中推理速度分别提升13.15%、13.70%。

基于图挖掘的黑灰产运作模式可视分析

为分析黑灰产网络资产图谱数据中黑灰产团伙掌握的网络资产及其关联关系,提出一种基于图挖掘的黑灰产运作模式可视分析方法.首先,在网络资产图谱数据中锁定潜在团伙线索;其次,根据潜在线索、黑灰产业务规则挖掘由同一黑灰产团伙掌握的网络资产子图,并识别子图中的核心资产与关键链路;最后,基于标记核心资产和关键链路的黑灰产子图实现可视分析系统,从而直观发现黑灰产团伙掌握的网络资产及其关联关系,帮助分析人员制定黑灰产网络资产打击策略.经实验验证,该方法能有效、直观地分析和发现黑灰产团伙及其网络资产关联关系,为更好监测黑灰产网络运作态势提供必要的技术支持.

一种基于子图转述的问题生成方法

本文提出了一种子图转述的方法用于解决知识图谱问题生成中的未见谓词问题.传统的问题生成方法主要利用标注的问答数据(问题-逻辑形式对)生成问题,然而标注数据难以覆盖知识图谱中所有的谓词,如何对未见的谓词生成问题依然是一个挑战.本文提出了一种基于子图结构的语义解耦方法,通过将复杂问题对应的知识图谱子图分解为原子级子图,从而将包含未见谓词的多跳子图拆分为易于处理的单跳子图.并且本文设计了一种子图转述方法,通过对数据集中的谓词进行采样,得到子图描述文本,并在大规模无监督数据上训练得到子图转述器,能够为包含未见谓词的子图提供自然语言形式的表述,为生成问题提供了有效的信息.本文定量分析了在不同的难度级别下模型的性能表现,在GrailQA等数据集上的实验结果表明,本文的方法达到了最先进的性能.

子图匹配和强化学习增强的三维点云配准

针对低质量三维点云数据配准精度不足、效率低的问题,为了实现低质量点云的精确、快速配准,提出一种基于子图匹配和强化学习的点云配准方法.首先将三维点云配准转化为一系列离散的刚性变换连续作用结果,利用强化学习策略训练一个端到端的模型以迭代输出刚性变换动作;然后对于模型架构,采用双流主干网络分别提取源点云与目标点云的局部特征信息,设计交叉图注意力模块将源点云图和目标点云图中的相似节点关联起来,使用带选通向量的加权实现图节点的聚合,分别获取源点云图与目标点云图的全局特征表示;最后融合源点云图与目标点云图的全局特征,基于融合特征预测离散的刚性变换动作.强化学习策略的引入显著提高了点云配准算法的泛化性,在加入交叉图注意力模块后,点云配准的精度及效率也进一步被提升.在ModelNet40和ScanObjectNN这2个公共基准数据集上与最新的点云配准方法 ReAgent进行实验的结果表明,所提方法能够将旋转误差的均方差数值降低至少0.16,各向同性旋转误差数值也降低至少0.16,有效地提升低质量点云配准的精度.

基于国产DCU异构平台的图匹配算法移植与优化

子图匹配是一种基础的图算法,被广泛应用于社交网络、图神经网络等众多领域。随着图数据规模的增长,人们迫切需要高效的子图匹配算法。GENEVA是一种基于GPU的并行子图匹配算法,其利用区间索引的图存储结构和并行匹配优化方法,能够大幅度减少存储开销,提升子图匹配性能。但由于平台底层硬件架构和编译环境的不同,GENEVA无法直接应用到国产DCU异构平台。为了解决该问题,提出了GENEVA面向国产DCU的移植和优化方案。IO时间开销是GENEVA算法主要的性能瓶颈,文中采用锁页内存、预加载、调度器3种优化策略来突破该瓶颈。其中,锁页内存技术避免了从可分页内存到临时锁页内存的额外数据传输,在DCU平台上大幅度减少了IO传输的时间开销;预加载技术将IO数据传输与DCU核函数计算重叠,掩盖了IO时间开销;调度器在满足预加载需求的同时,减少了冗余数据的传输。在3个不同规模的真实数据集上进行实验,结果表明,采用优化策略后算法性能显著提高。在92.6%的测试用例上,经过优化的GENEVA-HIP算法在国产DCU平台的执行时间比移植前的GENEVA算法在GPU服务器的执行时间短。在较大规模的数据集上,优化的GENEVA-HIP算法在DCU平台上的执行时间相比移植前的GENEVA算法在GPU服务器的执行时间减少了52.73%。

Deflagration characteristics of freely propagating flames in magnesium hydride dust clouds

The flame propagation processes of MgH_(2)dust clouds with four different particle sizes were recorded by a high-speed camera.The dynamic flame temperature distributions of MgH_(2)dust clouds were reconstructed by the two-color pyrometer technique,and the chemical composition of solid combustion residues were analyzed.The experimental results showed that the average flame propagation velocities of 23μm,40μm,60μm and 103μm MgH_(2)dust clouds in the stable propagation stage were 3.7 m/s,2.8 m/s,2.1 m/s and 0.9 m/s,respectively.The dust clouds with smaller particle sizes had faster flame propagation velocity and stronger oscillation intensity,and their flame temperature distributions were more even and the temperature gradients were smaller.The flame structures of MgH_(2)dust clouds were significantly affected by the particle sinking velocity,and the combustion processes were accompanied by micro-explosion of particles.The falling velocities of 23μm and 40μm MgH_(2)particles were 2.24 cm/s and 6.71 cm/s,respectively.While the falling velocities of 60μm and 103μm MgH_(2)particles were as high as 15.07 cm/s and 44.42 cm/s,respectively,leading to a more rapid downward development and irregular shape of the flame.Furthermore,the dehydrogenation reaction had a significant effect on the combustion performance of MgH_(2)dust.The combustion of H_(2)enhanced the ignition and combustion characteristics of MgH_(2)dust,resulting in a much higher explosion power than the pure Mg dust.The micro-structure characteristics and combustion residues composition analysis of MgH_(2)dust indicated that the combustion control mechanism of MgH_(2)dust flame was mainly the heterogeneous reaction,which was affected by the dehydrogenation reaction.

采用局部子图嵌入的MOOCs知识概念推荐模型

大规模开放在线课程(MOOCs)在减少用户学习盲区和改善用户体验方面已经有大量的研究,尤其是基于图神经网络的个性化课程资源推荐,但现有工作主要集中在固定或同质图上,容易受到数据稀疏问题的影响且难以扩展。在局部子图上使用图卷积,并结合扩展的矩阵分解(MF)模型来解决这一问题。首先,将异构图分解为多个基于元路径的子图,结合随机游走采样方法实现在采样节点富有影响力邻域的同时捕获实体之间复杂的语义关系,并在局部邻域上进行图卷积平滑各节点表示,实现高可扩展性;然后,使用注意力机制适应性地融合不同子图的上下文信息,更全面地构建用户偏好;最后,通过扩展矩阵分解优化模型参数,获得推荐列表。为了验证提出模型的性能,在公开的MOOCs数据集上进行对比实验,相较于最优基线,性能提升了2%,内存计算需求降低了近500%,缓解数据稀疏问题的同时仍具有较强的可扩展性。

基于可控制性度量的图神经网络门级硬件木马检测方法

随着全球化的不断深入,第三方知识产权(IP)核应用越来越广泛。随着硬件木马攻击技术逐渐成熟,使得在芯片设计阶段植入硬件木马成为可能。因此,在芯片设计过程中面临IP核被植入木马的严重威胁,现有研究所提的硬件木马检测方法具有依赖黄金参考电路、需要完备的测试向量、大量的样本进行学习等特征。面向IP核的硬件木马检测需求,提出一种基于可控制性度量的图神经网络检测方法。该方法以门级网表作为输入,首先以可控制性值为指导,得到可疑的门节点,用于缩小搜索范围;然后利用可疑门节点生成对应的子图,利用图卷积神经网络从子图中提取特征,实现对子图的分类和检测,最终识别硬件木马。实验结果表明,该方法无须测试激励和黄金模型,利用硬件木马的隐蔽特性与结构特征相结合的方法提升硬件木马的检测准确率,平均真阳率为100%,假阳率为0.75%,在保证较高真阳率的同时可有效降低假阳率,达到较好的检测效果。