摘要：Artificial intelligence(AI)is the core technology of technological revolution and industrial *** one of the new intelligent needs in the AI 2.0 era,financial intelligence has elicited much attention from the academia and *** our current dynamic capital market,financial intelligence demonstrates a fast and accurate machine learning capability to handle complex data and has gradually acquired the potential to become a'financial brain.'In this paper,we survey existing studies on financial ***,we describe the concept of financial intelligence and elaborate on its position in the financial technology ***,we introduce the development of financial intelligence and review state-of-the-art techniques in wealth management,risk management,financial security,financial consulting,and ***,we propose a research framework called FinBrain and summarize four open issues,namely,explainable financial agents and causality,perception and prediction under uncertainty,risk-sensitive and robust decision-making,and multi-agent game and mechanism *** believe that these research directions can lay the foundation for the development of AI 2.0 in the finance field.

摘要：针对摇摆运动下自然循环流动不稳定性的非线性演化特性进行了分析。应用非线性时间序列分析方法对不同流动状态的时间序列进行幅度谱分析、吸引子重构,并基于相空间重构理论,计算包括关联维数(CD)、Kolmogorov熵(K熵)和最大李雅普诺夫指数(MLE)在内的几何不变量的值,根据计算结果分析了摇摆运动下两相自然循环系统流动不稳定性的非线性演化特性。分析结果表明:随着无量纲功率的增大,系统几何不变量的值先增加后减小,系统由极限环运动经倍周期分岔发展成混沌振荡,最终回到稳定流动;系统非线性特征先增强后减弱,是由于热驱动力、流动阻力和摇摆引起的驱迫外力之间的相互反馈及耦合程度不同所致。

摘要：对摇摆条件下两相自然循环系统不规则复合型脉动进行了非线性时序分析。通过相空间重构计算了不规则复合型脉动时间序列的关联维、K2熵和最大Lyapunov指数值,在几何不变量的计算结果基础上,结合密度波型脉动和波谷型脉动的对比,分析了不规则复合型脉动的混沌特征和产生机理。分析结果表明:不规则复合型脉动为典型的混沌运动,热驱动力、流动阻力和摇摆引起的附加外力的相互作用和耦合导致了混沌的出现。

摘要：The effect of porosity on compressive,bending,and tensile properties of the porous tantalum scaffolds fabricated by electron beam powder bed fusion(EB-PBF)was *** porous tantalum scaffolds with porosity from 69%to 77.8%were obtained by varying the designed porosity and adjusting the processing *** is found that the pores and unfused powder decrease with the increase of deposited energy *** decrease of porosity leads to an improvement in mechanical *** relevancy between compressive/bending/tensile yield strength and relative density can be described appropriately by exponential model,while the relationship between elastic modulus and relative density is in good agreement with the Gibson-Ashby *** the porous tantalum scaffolds exhibit good ductility in compressive,bending and tensile *** fragmentation of struts is observed during the compression process,but cracks are formed on the strut surface after 90°bending,mainly due to the high sensibility to defects caused by the oxide.

摘要：无拖曳卫星的本体姿态、卫星本体与测试质量间的相对位移及相对姿态的联合控制受到外部扰动、输入噪声、测量噪声及饱和约束、输入耦合以及状态耦合等因素的影响,控制器的设计面临挑战.本文采用基于扩张状态的卡尔曼滤波对系统状态和系统扰动进行实时估计,引入自抗扰控制策略进行了控制器设计.针对无拖曳控制子系统设计了测量饱和受限下的扩张状态估计算法,并进行了信息融合.在设计控制律时不仅考虑了对外部扰动的补偿,还将系统状态间的耦合关系看成内部扰动进行补偿,使得被控系统等价为“积分串联型系统”,在此基础上实现了无拖曳卫星的联合控制.数值仿真验证了方法的有效性和合理性.

摘要：Flexible and Personalizable battery is a promising candidate for energy storage, but suffers from the weldablity and large-scale producibility of the electrode. To address the issues, we design a nickel foam catalyzed electroless deposition (NFED) derived 3D-metal-pattern embroidered electrodes. This is the first attempt to utilize this type of electrode in battery field. It is found that the current collector can be embroidered on any selected areas of any complex-shape electrodes, with high controllability and economical feasibility. As a result, the electronic conductivity of the flexible electrodes can be improved by nearly one order of magnitude, which can be easily and firmly weldded to the metal tab using the industry generic ultrasonic heating process. The embroidered electrodes could substantially promote the electrochemical performance under bending deformation, with both Li-S and Li-Li FePO4batteries as the models. This innovation is also suitable to embroider all the VIII group elements on any electrodes with personalized shapes, which is widely attractive for the development of next generation flexible and personalizable energy storage devices.

摘要：工业计算机断层扫描成像技术(Computed Tomography,CT)检测叠片电芯极片时,在CT图像中分割出极片是关键步骤。然而,由于叠片电芯的极片结构复杂、长宽比大,现有分割方法在速度与精度上难以满足大规模生产检测的需求。为解决这个问题,提出一种融合条形注意力机制的电芯极片CT图像分割网络。该网络主要由主干网络、条形空洞金字塔池化模块(Strip Atrous Spatial Pyramid Pooling,S-ASPP)和解码器模块组成。其中,主干网络设计为轻量型网络MobilenetV2,用于减少网络参数量并提高分割速度;S-ASPP模块中将全局池化设计为条形池化,用于减少条形特征信息的丢失,有效解决欠分割的问题;解码器模块的条形注意力机制用于关注图像中的边缘和细节信息,从而改善边缘模糊的问题。在电芯极片CT数据集上进行学习验证,实验结果表明,本文网络对电芯极片的平均像素分类准确率(mAcc)、平均交互比(mIoU)、参数量和分割速度分别达到了98.62%,89.97%,5.814M,56.94 frame/s。相比DANet,DeepLabV3+,U-Net,HRNet和Segnext网络,本文网络的分割精度最高,分割速度优于DANet,DeepLabV3+,U-Net和HRNet网络,略逊色Segnext网络。综合考虑评价指标mIoU,mAcc,FPS,Parameters(M),本文方法在保持较低计算成本的同时,显著提升了极片分割的精度和效率,优于目前主流的分割网络。

摘要：相机内参及相机与激光雷达外参标定是两者融合感知的前提,为解决平面标定方法存在的过拟合问题,并提高相机与低线束激光雷达的外参标定精度,提出一种基于三维标定板的相机内参及相机/激光雷达外参的联合标定方法。首先通过三维标定板校准相机内参以解决内参标定过程中的过拟合问题;然后设计了一种基于线曲率的平面提取方法来提高单帧点云的平面提取精度;最后通过图像与激光点云中对应的平面特征计算两个传感器之间的外参。实验结果表明,所提方法相比平面标定方法能得到更精确相机内参,与velo2cam_calibration算法相比,外参标定的平均平移误差减少了39.4%,平均旋转误差减少了37.1%。

摘要：隐式反馈具有数据获取成本小、形式广泛的特点,因此在现代推荐系统中被广泛使用.由于用户的隐式反馈通常是稀疏,不平衡,且含义不明确的.因此,想要准确学习用户和物品之间的复杂交互具有挑战性.传统的基于矩阵分解的推荐方法只能建模用户-物品之间的相似性.同时,矩阵分解使用点积运算作为相似度评估方式,而点积运算不满足三角不等式,即不能将用户-物品相似性传递到用户-用户以及物品-物品的相似性建模.因此,矩阵分解不足以在隐式反馈中充分建模用户和物品的关系.尽管现在有基于隐式反馈使用欧式距离来度量用户-物品相似度的度量学习方法,使得对应的推荐方法能够满足三角不等式.但是,现有的度量方法通常会将每个用户或者物品表示为度量空间中的单个点,进而在单个空间内通过用户-物品之间的距离来表征用户-物品之间的相似性.由于在不同的环境下,用户对于同一种类型的物品的偏好也可能存在差异.基于单个空间的用户、物品嵌入向量有可能无法满足用户具有的多种偏好和物品具有的多种属性,进而限制了推荐系统的性能.为了充分刻画用户和物品,我们尝试从多个侧面对于用户和物品进行表示,并提出了一个基于多空间的度量学习(MML)框架.通过设计整合多个空间相似性的度量方式,我们将用户和物品投影到多个空间中进行细粒度的表示.另外,我们设计了一种经过校准的优化策略,包括经过校准的最大间隔损失函数和经过校准的采样方法.在保持多空间度量学习表示能力的同时,确保框架的有效性.最后,模型通过训练好的用户、物品向量,对于稀疏的用户-物品交互矩阵进行填补.在动态更新空间权重的同时,可以赋予模型新的训练视角,最终实现端到端的训练.通过四个真实世界推荐数据集上进行的大量实验表明,MML可以在Recall和nDCG衡量指标上将目前最优的对比算法提高40%以上.

摘要：对摇摆条件下的单相强迫循环流量脉动特性进行实验研究,并通过受力分析研究影响流量脉动的主要因素。研究表明:摇摆运动能否造成强迫循环流量出现周期性的脉动主要取决于泵驱动压头与附加惯性力的相对大小。摇摆幅度或频率越大,流量脉动幅度越大;循环泵驱动压头越大,流量脉动幅度越小;当驱动压头与附加惯性力的比值达到一定值后,摇摆运动发生前后流量不再出现明显变化。