The Method of Virtual Experts in Mathematical Diagnostics

Mathematical Diagnostics (MD) deals with identification problems arising in different practical areas. Some of these problems can be described by mathematical models where it is required to identify points belonging to two or more sets of points. Most of the existing tools provide some identification rule (a classifier) by means of which a given point is assigned (attributed) to one of the given sets. Each classifier can be viewed as a virtual expert. If there exist several classifiers (experts), the problem of evaluation of experts’ conclusions arises. In the paper for the case of supervised classification the method of virtual experts (the VE-method) is described. Based on this method, a generalized VE method is proposed where each of the classifiers can be chosen from a given family of classifiers. As a result, a new optimization problem with a discontinuous functional is stated. Examples illustrating the proposed approach are provided. The work of the second author was supported by the Russian Foundation for Fundamental Studies (RFFI) under Grant No 03-01-00668.     

Deep ensemble-based hard sample mining for food recognition

Deep neural networks represent a compelling technique to tackle complex real-world problems, but are over-parameterized and often suffer from over- or under-confident estimates. Deep ensembles have shown better parameter estimations and often provide reliable uncertainty estimates that contribute to the robustness of the results. In this work, we propose a new metric to identify samples that are hard to classify. Our metric is defined as coincidence score for deep ensembles which measures the agreement of its individual models. The main hypothesis we rely on is that deep learning algorithms learn the low-loss samples better compared to large-loss samples. In order to compensate for this, we use controlled over-sampling on the identified ”hard” samples using proper data augmentation schemes to enable the models to learn those samples better. We validate the proposed metric using two public food datasets on different backbone architectures and show the improvements compared to the conventional deep neural network training using different performance metrics.     

基于BERT和LightGBM的文本关键词提取方法

传统的文本关键词提取方法忽略了上下文语义信息,不能解决一词多义问题,提取效果并不理想。基于LDA和BERT模型,文中提出LDA-BERT-LightG BM(LB-LightG BM)模型。该方法选择LDA主题模型获得每个评论的主题及其词分布,根据阈值筛选出候选关键词,将筛选出来的词和原评论文本拼接在一起输入到BERT模型中,进行词向量训练,得到包含文本主题词向量,从而将文本关键词提取问题通过LightG BM算法转化为二分类问题。通过实验对比了textrank算法、LDA算法、LightG BM算法及文中提出的LB-LightG BM模型对文本关键词提取的准确率P、召回率R以及F1。结果表明,当Top N取3～6时,F1的平均值比最优方法提升3.5%,该方法的抽取效果整体上优于实验中所选取的对比方法,能够更准确地发现文本关键词。     

特征增强和双线性特征向量融合的移动端工业货箱文本检测

在实际工业环境下,光线昏暗、文本不规整、设备有限等因素,使得文本检测成为一项具有挑战性的任务。针对此问题,设计了一种基于双线性操作的特征向量融合模块,并联合特征增强与半卷积组成轻量级文本检测网络RGFFD（ResNet18+GhostModule+特征金字塔增强模块（feature pyramid enhancement module, FPEM）+ 特征融合模块（feature fusion module,FFM）+可微分二值化（differenttiable binarization,DB））。其中,Ghost模块内嵌特征增强模块,提升特征提取能力,双线性特征向量融合模块融合多尺度信息,添加自适应阈值分割算法提高DB模块分割能力。在实际工厂环境下,采用嵌入式设备UP2 board对货箱编号进行文本检测,RGFFD检测速度达到6.5 f/s。同时在公共数据集ICDAR2015、Total-text上检测速度分别达到39.6 f/s和49.6 f/s,在自定义数据集上准确率达到88.9%,检测速度为30.7 f/s。     

双向特征融合的快速精确任意形状文本检测

现有的基于分割的场景文本检测方法仍较难区分相邻文本区域,同时网络得到分割图后后处理阶段步骤复杂导致模型检测效率较低.为了解决此问题,该文提出一种新颖的基于全卷积网络的场景文本检测模型.首先,该文构造特征提取器对输入图像提取多尺度特征图.其次,使用双向特征融合模块融合两个平行分支特征的语义信息并促进两个分支共同优化.之后,该文通过并行地预测缩小的文本区域图和完整的文本区域图来有效地区分相邻文本.其中前者可以保证不同的文本实例之间具有区分性,而后者能有效地指导网络优化.最后,为了提升文本检测的速度,该文提出一个快速且有效的后处理算法来生成文本边界框.实验结果表明:在相关数据集上,该文所提出的方法均实现了最好的效果,且比目前最好的方法在F-measure指标上最多提升了1.0％,并且可以实现将近实时的速度,充分证明了该方法的有效性和高效性.     

A classification-based privacy-preserving decision-making for secure data sharing in Internet of Things assisted applications

The introduction of the Internet of Things (IoT) paradigm serves as pervasive resource access and sharing platform for different real-time applications. Decentralized resource availability, access, and allocation provide a better quality of user experience regardless of the application type and scenario. However, privacy remains an open issue in this ubiquitous sharing platform due to massive and replicated data availability. In this paper, privacy-preserving decision-making for the data-sharing scheme is introduced. This scheme is responsible for improving the security in data sharing without the impact of replicated resources on communicating users. In this scheme, classification learning is used for identifying replicas and accessing granted resources independently. Based on the trust score of the available resources, this classification is recurrently performed to improve the reliability of information sharing. The user-level decisions for information sharing and access are made using the classification of the resources at the time of availability. This proposed scheme is verified using the metrics access delay, success ratio, computation complexity, and sharing loss.     

面向智慧能源的供需平衡数据可视化挖掘方法

为满足能源数据监测和综合分析领域的业务需要,提出面向智慧能源的供需平衡数据可视化挖掘方法。构造数据迁移与分类视图,确定每个供需平衡数据点位置坐标,并采用正交投影可视化技术将数据挖掘算法核心结构展示在视图中(支持用户对其进行相应调整);引入数据挖掘算法,提取数据可视化挖掘结果参数,实现面向智慧能源供需平衡数据的可视化挖掘。实验结果显示:智慧能源供需数据可视化挖掘结果综合评分数值范围为64.70~82.04分,充分说明该方法具备较好的可视化挖掘性能,可应用在多维典型用能场景中。     

物理化学课程“热化学”主题学习历程研究——化学师范生知识建构的文本分析

为深入了解地方性师范院校化学师范生知识建构习惯和水平，探讨他们在大学专业基础课程中的学习历程，以物理化学课程中以“融合科学读写特色”的“热化学”学案作为研究载体，研究对象在学案引导下开展自学活动，并在阅读教材和文献基础上进行科学写作。采用量表工具针对大学生自学效果和反思活动进行评价，并采用SOLO分类评价科学写作作品，采用KWL工具评价学生的科学阅读收获。结果发现研究对象学习动机不强，不擅长在新的学科知识学习中联系已有专业课程知识，基于文本学习的建构活动水平极低，可能是因为他们在应试背景下的基础知识并未形成体系，先修课程中并未培养元认知学习策略。     

Design criteria for molecular mimics of fragments of the β-turn. 1. Cα atom analysis

Peptides represent an extensive class of biologically active molecules. They may be used as leads in the development of novel therapeutic agents provided the pharmacophoric information present within them can be translated into non-peptide analogs that lack the peptide backbone and are stable to proteolysis. This is the rationale for peptidomimetic drug design. Frequently, the -turn has been implicated as a conformation important for biological recognition of peptides. Empirical evidence from known peptidomimetics, coupled with a theoretical model of peptide binding and the observation that glycine and proline residues are common within the -turn, has suggested the design of molecules to mimic placement of between two and four of the side-chains. The moderate number of different -turn conformations, combined with the combinatoric nature of side-chain selection complicates the procedure. In this paper, cluster analysis has been used to classify the arrangement of C_{_} atoms about the various fragments of the -turn. Recombination of the observed patterns provides a general model for the -turn which may be used as an effective screen for potential peptidomimetic scaffolds in chemical databases.     

An Analytical Method for the Separation and Determination of As(III) and As(V) in Seepage and Acid Mine Drainage Water

To avoid changes in the original As species distribution in natural water after sampling, a method of immediate separation of As(V) by anion exchange at the sampling site was developed. The procedure consists of two steps. The total concentration of arsenic is determined in one part of the water sample acidified on site. Another part of the water samples is pressed through a column filled with an anion exchanger. The As(III) species that is not redox-stable remains in the effluent of the sorbents column and can be analyzed with conventional methods after stabilization by addition of conc. HNO₃. As(V) is sorbed by the exchanger material. The As(V) concentration can be calculated as the difference between As_sol and As(III), neglecting very low contents of methylated species. Oxidation of Fe(II) by air followed by co-precipitation of arsenic with iron hydroxide was applied in field experiments to minimize the As concentration in seepage and mining water.