Asynchronous rotation scan for synthetic aperture interferometric radiometer

Synthetic aperture interferometric technique has wide applications in optics, radio astronomy and microwave remote sensing areas. With the increasing demands of high resolution imaging observation, a new time-sharing sampling scheme of asynchronous rotation scan is proposed to meet the technical challenge of achieving a large equivalent aperture and overcome the operating barriers of space borne application. This configuration is basically composed by two asynchronously and concentrically rotating antenna groups, whose revolving radii and speeds are different. The synthetic aperture system with asynchronous rotation scanning scheme can effectively solve the trade-off problem of system complexity, and greatly simplify the system hardware at the cost of sacrificing a certain time resolution. The basic rules and design methods of asynchronous rotation scan are investigated The Gridding method is introduced to inverse the spiral sampling data for image reconstruction. The potential applications of geostationary orbit (GEO) earth observation and solar polar orbit (SPO) plasma cloud observation are explored with numerical simulations to validate the significance and feasibility of this new imaging configuration. Supported by the National Natural Science Foundation of China (Grant No. 40574070, 40671121, 40701100 and 40801136) and the National High-Tech Research Program of China (“863” Program) (Grant No. 2006AA12Z141)     

Gaussian sampling of lattices for cryptographic applications

Gaussian sampling is the major class of algorithms for solving the close vector problem(CVP)of lattices.In this paper we present a novel Gaussian sampling algorithm,which has the same cryptographic applications with original Gaussian sampling algorithms.Our novel Gaussian sampling algorithm has smaller deviations,meaning smaller space sizes of lattice based public-key ciphers.The shape of our novel algorithm is almost repeated implementations of original algorithm,with random repeating times.Major result is that the deviation can be reduced to 0.64～0.75 of that of original Gaussian sampling algorithm without clearly increasing the average time cost.     

机械电器产品检验抽样方法探讨

根据机电产品生产特点,对稳定生产状态下的生产条件进行细分,提出不同生产条件下的生产质量验收抽样方法。     

An Application of Silica‐Based Monolithic Membrane Emulsification Technique for Easy and Efficient Preparation of Uniformly Sized Polymer Particles

Summary: Uniformly sized polymer particles were prepared by an emulsification and polymerization technique utilizing a silica monolithic membrane, namely the “silica monolithic membrane emulsification technique”. In this paper, we utilized silica monolithic membrane as a device for the preparation of uniformly sized polymer particles. A mixture of monomers, diluents and oil‐soluble initiator was emulsified into a continuous medium through the silica monolithic membrane and polymerized. The particles obtained had a higher size uniformity than that of particles prepared by previously reported membrane emulsification techniques, such as the Shirasu Porous Glass (SPG) emulsification technique. Through the silica monolithic membrane emulsification technique, we could prepare particles having availability as a possible packing material for solid‐phase extraction (SPE) and high performance liquid chromatography (HPLC).

SEM photograph of silica particles prepared through capillary plate membrane.     

车间空气中氨浓度测定方法改进

用纳氏试剂分光光度法代替纳氏试剂目测比色法测定车间空气中氨浓度，并采用了单管直接抽取空气样品的方法，使操作简单。并对分析过程中的干扰因素进行了消除，经过作标准曲线、空白试验和计算方法检测限，证明此方法准确度较高。     

一种新的基于隐喻地图的RPA路径规划算法

智能化地制定机器人流程自动化(robotic process automation, RPA)执行路径有利于企业节约相关人力成本以及提高RPA的推广,提出基于改进深度双Q网络(double deep Q-learning algorithms, DDQN)算法进行RPA路径规划。首先针对存在RPA的作业环境即Web页面,不满足深度增强算法的探索条件的问题,借助隐喻地图的思想,通过构建虚拟环境来满足路径规划实验要求。同时为了提高DDQN算法探索效率,提出利用样本之间的位置信息的杰卡德系数,将其作为样本优先度结合基于排名的优先级(rank-based prioritization)构建新的采样方式。通过随机采用任务样本在虚拟环境上进行验证,证明其符合实验要求。进一步比较改进DDQN、深度Q网络(deep Q network, DQN)、DDQN、PPO以及SAC-Discrete算法的实验结果,结果显示改进算法的迭代次数更少、收敛速度更快以及回报值更高,验证了改进DDQN的有效性和可行性。     

基于AERF模型的油井结蜡预测

油井结蜡是一种在开发以及开采油田时对油井正常产出造成了负面影响的现象,该现象会引起油流通道堵塞,导致油井开采过程中出油量降低.对油井结蜡状况做出智能预警,完成油井设备提前修复,对油田提高产能效率、降低维护成本及智能化管理有非常关键的价值.为了解决油井正常数据和结蜡数据严重不平衡问题,本文引入了自适应合成抽样法(ADASYN)和最近邻规则欠抽样法(ENN)两种非均衡样本处理方法,分别对类别为结蜡的样本和非结蜡的样本进行处理,最终使用随机森林算法对新构成的数据集训练,构造出AERF智能模型来预测油井结蜡.实验结果表明,提出的AERF模型在油井的结蜡数据集中预测效果更佳,明显地提高了预测精度.     

中文知识库问答中的路径选择

路径选择是知识库问答任务的关键步骤,语义相似度常被用来计算路径对于问句的相似度得分。针对测试集中存在大量未见的关系,该文提出使用一种负例动态采样的语义相似度模型的训练方法,去丰富训练集中关系的多样性,模型性能得到显著提升。针对复杂问题候选路径数量组合爆炸问题,该文比较了两种路径剪枝方法,即基于分类的方法和基于集束搜索的方法。在包含简单问题和复杂问题的CCKS 2019-CKBQA评测数据集上,该方法能达到较优异的性能,测试集上单模型系统平均F₁值达到0.694,系统融合后达到0.731。     

基于改进的BSMOTE和时序特征的风机故障采样算法

针对风机数据集的不平衡问题,提出了一种BSMOTE-Sequence采样算法,在合成新样本时综合考虑空间和时间特征,并对新样本进行清洗,从而有效减少噪声点的生成。首先,根据每个少数类样本的近邻样本的类别比例,将少数类样本划分为安全类样本、边界类样本和噪声类样本。然后,对每个边界类样本都遴选出空间距离、时间跨度最接近的少数类样本集,利用线性插值法合成新样本,并过滤掉噪声类样本以及类间重叠样本。最后,以支持向量机（SVM）、卷积神经网络（CNN）、长短期记忆（LSTM）人工神经网络作为风机齿轮箱故障检测模型,F1-Score、曲线下面积（AUC）和G-mean作为模型性能评价指标,在真实风机数据集上把所提算法与常用的多种采样算法进行对比,实验结果表明：相比已有算法,BSMOTE-Sequence算法所生成样本的分类效果更好,使得检测模型的F1-Score、AUC和G-mean平均提高了3%,该算法能有效地适用于数据具有时序规律且不平衡的风机故障检测领域。     

基于局部概率抽样的标签噪声过滤方法

分类学习任务中,在获取数据的过程中会不可避免地产生噪声,特别是标签噪声的存在不仅使得学习模型更复杂,而且容易造成过拟合并导致分类器泛化能力的下降。标签噪声过滤算法虽然在一定程度上可以解决上述问题,但是仍然存在噪声识别能力较差、分类效果不够理想以及过滤效率低等问题。针对这些问题,提出一种基于标签置信度分布的局部概率抽样方法来进行标签噪声过滤。首先利用随机森林分类器对样本的标签进行投票,从而获取每个样本的标签置信度;然后根据标签置信度的大小,将样本划分为易识别样本和难识别样本;最后分别采用不同的过滤策略对样本进行过滤。实验结果表明,在标签噪声存在的情况下,所提方法在大多数案例上能够保持较高的噪声识别能力,并且在分类泛化性能上也具有明显优势。