基于轮廓与断面匹配的秦俑碎片拼接方法

针对兵马俑第3次发掘大量破碎俑片的虚拟复原需求,分析秦俑碎片拼接的特殊性,归纳陶制碎片的分类原则,提出一种结合碎片边缘轮廓和断裂面匹配的混合拼接方法。依据俑体碎片断裂面厚度,将其映射到空间曲线模板或空间曲面模板。基于积分不变量计算俑体碎片特征向量,采用一致性约束方法获取碎片初始匹配点对。利用筛选算法去除伪初始匹配点对,应用几何哈希算法寻找碎片最优匹配点对。其中,空间轮廓曲线匹配基于多边形弧串匹配法,空间曲面匹配基于多尺度积分不变量法。对破碎兵俑和陶马的拼接结果表明,该方法能正确得到碎片间的匹配关系,具有较好的鲁棒性。     

Multiplicative Noise Removal Using L1 Fidelity on Frame Coefficients

We address the denoising of images contaminated with multiplicative noise, e.g. speckle noise. Classical ways to solve such problems are filtering, statistical (Bayesian) methods, variational methods, and methods that convert the multiplicative noise into additive noise (using a logarithmic function), apply a variational method on the log data or shrink their coefficients in a frame (e.g. a wavelet basis), and transform back the result using an exponential function.     

串联重复序列比对的位置筛选方法

串联重复序列是基因组构建的困难片段,由于其重复单元之间的相似性与其拷贝数的不确定性,在序列比对时容易定位到多个候选位置,如何快速而准确地筛选出正确的比对位置是一项挑战。现有方法使用种子（从测序片段中选取的短序列）来定位并扩展候选比对位置,但挑选种子时未考虑串联重复序列特性。因此,提出了一种串联重复序列比对的位置筛选方法,其通过计算稀有kmer（长度为k的子序列）序列的相似性来筛选比对结果。此外,采用合并稀有kmer的策略加速计算,并利用基于编辑距离的模糊查找以提高过滤信息密度。实验结果表明,在模拟数据集上提高比对结果的召回率与准确率的同时,该方法比现有方法快约2倍,且具有良好的并行加速性能。     

A derived Markov process for modeling reaction networks

A reaction network arises when a set of reactants (chromosomes, chemicals, economic goods, or the like) recombine at specified rates to produce other reactants in the set. When the reactants are characterized in terms of "reactive regions" (schemata, active sites, building blocks), reaction networks can be modeled by classic stochastic urn models. The corresponding Markov processes are specified by matrices that, for realistic problems, are small enough to allow standard matrix operations and Monte Carlo estimates of important properties of the trajectory of the process, such as the expected time to first occurrence of some designated reactant.     

采用块间关系的彩色图像水印方法

提出一种空间域彩色图像的盲水印方法。该方法首先对YUV彩色空间中图像的亮度分量Y进行4×4分块,按每两个相邻图像块为一组对其分组。然后从挑选的图像块组中利用相邻图像块的亮度均值间的大小关系来实现水印的嵌入。而水印的提取是通过判别相邻两个图像块亮度均值间的大小关系来完成。该方法具有其实现简单、计算代价小和速度快的特点,同时实验结果表明在对抗诸如JPEG压缩、旋转、缩放、锐化和低通滤波等攻击有比其他几种方法有更好的性能和鲁棒性。     

一种基于双边滤波的图像边缘检测方法

提出了一种基于双边滤波的图像边缘检测方法。利用图像像素的空间邻近度和灰度相似度的乘积来代替传统的Canny算法中高斯滤波的权系数,用该乘积与原图像进行卷积运算,然后通过非极大值抑制和高低阈值的方法检测出图像的边缘。基于双边滤波的图像边缘检测方法不仅有效地去除了图像中的噪声,而且很好地检测出了图像的真实边缘。     

亚硝酰氢与羟基自由基反应的密度泛函理论研究

用密度泛函理论研究了HNO OH反应机理。在(U)B3LYP/aug-cc-pVTZ水平上,优化了反应通道上各驻点(反应物、中间体、过渡态和产物)的几何构型,获得了零点能校正后的反应势能曲线。研究表明:根据进攻方式的不同,有3个反应通道,反应通道不同则产物不同。反应的主要产物是NO H_2O,次要产物是NH_2 O_2;主要产物和次要产物与反应物的总能量之差(经零点能校正后)分别为-133.42 kJ/mol和150.44 kJ/mol。生成主要产物时主要反应通道的活化能为102.11 kJ/mol。     

A new similarity measure for collaborative filtering to alleviate the new user cold-starting problem

Collaborative filtering is one of the most successful and widely used methods of automated product recommendation in online stores. The most critical component of the method is the mechanism of finding similarities among users using product ratings data so that products can be recommended based on the similarities. The calculation of similarities has relied on traditional distance and vector similarity measures such as Pearson’s correlation and cosine which, however, have been seldom questioned in terms of their effectiveness in the recommendation problem domain. This paper presents a new heuristic similarity measure that focuses on improving recommendation performance under cold-start conditions where only a small number of ratings are available for similarity calculation for each user. Experiments using three different datasets show the superiority of the measure in new user cold-start conditions.     

基于邻域差分滤波生成式对抗网络的数据增强方法

针对工业产品样本缺乏且特征不明显而难以用于深度学习训练的问题,提出一种邻域差分滤波生成式对抗网络数据增强（NDF-GAN）方法。将邻域差分滤波器融合到生成对抗网络中,从样本中提取特征并进行样本重建,对样本进行数据增强。实验表明,提出的方法所生成的样本比现有两种模型质量更高,所生成样本与真实样本混合训练分类模型后获得更好的分类性能。因此,提出的NDF-GAN实现了对工业产品样本的数据增强。     

Product named entity recognition for Chinese query questions based on a skip-chain CRF model

As more and more commercial information can be obtained from the Internet, product named entity recognition plays an important role in market intelligence management. In this paper, a product named entity recognition method based on a skip-chain CRF model is proposed. This method considers not only the dependence between neighboring words but also the fact that product named entities are often connected by a connective. In this situation, the dependence between the words around the connective is more important than the dependence between neighboring words. This information improves the result of product named entity recognition as shown in the experiments. Experimental results on corpuses of mobile phone and digital camera demonstrate that the skip-chain CRF model works well and produces better results than the linear-chain CRF model.     

Applications of random sampling to virtual screening of combinatorial libraries

We describe statistical techniques for effective evaluation of large virtual combinatorial libraries (> 10(10) potential compounds). The methods described are used for computationally evaluating templates (prioritization of candidate libraries for synthesis and screening) and for the design of individual combinatorial libraries (e.g., for a given diversity site, reagents can be selected based on the estimated frequency with which they appear in products that pass a computational filter). These statistical methods are powerful because they provide a simple way to estimate the properties of the overall library without explicitly enumerating all of the possible products. In addition, they are fast and simple, and the amount of sampling required to achieve a desired precision is calculable. In this article, we discuss the computational methods that allow random product selection from a combinatorial library and the statistics involved in estimating errors from quantities obtained from such samples. We then describe three examples: (1) an estimate of average molecular weight for the several billion possible products in a four-component Ugi reaction, a quantity that can be calculated exactly for comparison; (2) the prioritization of four templates for combinatorial synthesis using a computational filter based on four-point pharmacophores; and (3) selection of reagents for the four-component Ugi reaction based on their frequency of occurrence in products that pass a pharmacophore filter.     

HQE: A hybrid method for query expansion

Query expansion methods have been extensively studied in information retrieval. This paper proposes a query expansion method. The HQE method employs a combination of ontology-based collaborative filtering and neural networks to improve query expansion. In the HQE method, ontology-based collaborative filtering is used to analyze semantic relationships in order to find the similar users, and the radial basis function (RBF) networks are used to acquire the most relevant web documents and their corresponding terms from these similar users’ queries. The method can improve the precision and only requires users to provide less query information at the beginning than traditional collaborative filtering methods.     

Scalable prediction of non-functional properties in software product lines: Footprint and memory consumption

ContextA software product line is a family of related software products, typically created from a set of common assets. Users select features to derive a product that fulfills their needs. Users often expect a product to have specific non-functional properties, such as a small footprint or a bounded response time. Because a product line may have an exponential number of products with respect to its features, it is usually not feasible to generate and measure non-functional properties for each possible product.ObjectiveOur overall goal is to derive optimal products with respect to non-functional requirements by showing customers which features must be selected.MethodWe propose an approach to predict a product’s non-functional properties based on the product’s feature selection. We aggregate the influence of each selected feature on a non-functional property to predict a product’s properties. We generate and measure a small set of products and, by comparing measurements, we approximate each feature’s influence on the non-functional property in question. As a research method, we conducted controlled experiments and evaluated prediction accuracy for the non-functional properties footprint and main-memory consumption. But, in principle, our approach is applicable for all quantifiable non-functional properties.ResultsWith nine software product lines, we demonstrate that our approach predicts the footprint with an average accuracy of 94%, and an accuracy of over 99% on average if feature interactions are known. In a further series of experiments, we predicted main memory consumption of six customizable programs and achieved an accuracy of 89% on average.ConclusionOur experiments suggest that, with only few measurements, it is possible to accurately predict non-functional properties of products of a product line. Furthermore, we show how already little domain knowledge can improve predictions and discuss trade-offs between accuracy and required number of measurements. With this technique, we provide a basis for many reasoning and product-derivation approaches.     

DFT simulations and microkinetic modelling of 1-pentyne hydrogenation on Cu20 model catalysts

Adsorption and dissociation of H₂ and hydrogenation of 1-pentyne on neutral and anionic Cu₂₀ clusters have been investigated using the density functional theory and microkinetic modelling. Molecular adsorption of H₂ is found to occur strictly at atop sites. The H₂ dimer is activated upon adsorption, and the dissociation occurs with moderate energy barriers. The dissociated H atoms reside preferentially on 3-fold face and 2-fold edge sites. Based on these results, the reaction paths leading to the partial and total hydrogenation of 1-pentyne have been studied step-by-step. The results suggest that copper clusters can display selective activity on the hydrogenation of alkyne and alkene molecules. The hydrogenated products are more stable than the corresponding initial reactants following an energetic staircase with the number of added H atoms. Stable semi-hydrogenated intermediates are formed before the partial (1-pentene) and total (pentane) hydrogenation stages of 1-pentyne. The microkinetic model analysis shows that C₅H₁₀ is the dominant product. Increasing the reactants (C₅H₈/H₂) ratio enhances the formation of products (C₅H₁₀ and C₅H₁₂).     

基于矢量尺度的图像滤波

对于医学图像滤波来说,很重要的一点就是滤波后的图像应该尽可能地保留图像中的边缘和细节特征.但通常在滤波过程中,在消除噪声的同时会模糊图像中一些重要的结构信息.在最近几年中,基于尺度的滤波方法已经有效地应用在灰度图像的滤波中.现把基于尺度的方法推广到矢量(彩色)图像的滤波中.在传统滤波方法(矢量中值滤波、基本矢量方向滤波和方向距离滤波)基础上,相应地提出了3种基于球尺度的矢量滤波器.新的滤波方法能根据图像中像素的尺度信息,在图像边缘和细节附近,即区域边界执行较小的平滑,而在区域内部进行较大的平滑,从而能够自适应地控制滤波过程.实验结果表明,所提出的滤波方法与传统滤波方法相比,在消除噪声的同时更能够保留图像中的边缘和细节特征.     

Multiple platforms design and product family process planning for combined additive and subtractive manufacturing

Industry 4.0 promotes the utilization of new exponential technologies such as additive manufacturing in responding to different manufacturing challenges. Among these, the integration of additive and subtractive manufacturing technologies can play an important role and be a game changer in manufacturing products. In addition, using product platforms improves the efficiency and responsiveness of manufacturing systems and is considered an enabler of mass customization. In this paper, a model to design multiple platforms that can be customized using additive and subtractive manufacturing to manufacture a product family cost-effectively is proposed. The developed model is used to determine the optimal number of product platforms, each platform design (i.e. its features set), the assignment of each platform to various product variants, and the macro process plans for customizing the platforms while minimizing the overall product family manufacturing cost.The multiple additive/subtractive platforms and their process plans are determined by considering not only the commonality between the product variants but also their various manufacturing cost elements and the customer demand of each variant. The design of multiple product family platforms and their process plans is NP-hard problem. A genetic algorithm-based model is developed to reduce the computational complexity and find optimal or near optimal solution. Two case studies are used to illustrate the developed multiple platform model. The model results were compared with a single platform model in literature and the results demonstrate the multiple platform model superiority in manufacturing product families in lower cost. The use of the developed model enables manufacturing product families cost efficiently and allows manufacturers to manage diversity in products and market demands.     

CH3S与CH2SH自由基反应的密度泛函理论研究

用量子化学从头算和密度泛函理论(DFT)对CH3S与CH2SH自由基反应进行了研究。在B3LYP/6-311+G(d,p)水平上优化了反应通道各驻点物种(反应物、中间体、过渡态和产物)的几何构型,用内禀反应坐标(IRC)计算和频率分析的方法对过渡态结构及连接性进行了验证。在QCISD(T)/6-311++G(d,p)水平上计算了各物种的单点能,并对总能量进行了零点能校正。研究结果表明,CH3S与CH2SH反应为多通道反应,有3条可能的反应通道。反应物首先通过C-S键相互作用形成链状碳-硫偶合中间体IM1,再经过裂解和脱氢气机理生成主要产物P1(CH3SH+CH2S)和次要产物P2和P3。根据势能面分析,所有反应均为吸热反应。相对于CH3S与CH2SH,各产物能量分别为(-175.2,335.5和331.5)kJ.mol-1。     

林产品贸易信息推送梯级过滤技术

目前信息推送服务广泛应用于各类电子商务网站,然而传统信息过滤技术在林产品贸易信息过滤的过程中,存在着不足。在总结林产品贸易信息过滤的特点的基础上,提出梯级过滤技术。该技术根据林产品贸易用户兴趣的不同、用户需求程度的不同、林产品贸易信息特征项权重的不同,在提出林产品贸易信息特征向量空间模型和分析用户兴趣特征向量结构的基础上,采用先进行用户显性需求的严格过滤,再进行用户显性需求的模糊过滤,再进行用户隐性需求的过滤的梯级过滤方法,从而达到准确满足用户需求的目的。实验结果证明该方法能够有效地过滤出用户满意的信息。     

单取代氧/硫杂紫菜嗪内氢迁移反应的理论研究

采用B3LYP/6-31G~(**)方法在Gaussian 03程序下,对氧杂紫菜嗪,硫杂紫菜嗪的稳定结构和能量进行优化,并寻找与内氢迁移反应相关的过渡态构型。计算结果表明,芳香性与各体系构象稳定性密切相关,但并不是决定构象稳定性的唯一因素;在中心空穴较小的紫菜嗪体系中,空间位阻,静电作用才是影响构象稳定性的关键因素。由于分子内氢键的存在,氧杂化会明显降低内氢迁移正负反应速率;而S杂化体系的IHAT速率降低则与静电斥力和空间位阻密切相关。     

Selection of Optimal Stopping Time for Nonlinear Diffusion Filtering

We develop a novel time-selection strategy for iterative image restoration techniques: the stopping time is chosen so that the correlation of signal and noise in the filtered image is minimized. The new method is applicable to any images where the noise to be removed is uncorrelated with the signal, under the assumptions that the filter used is suitable for the given type of data, and that neither the additive noise nor the filtering procedure alter the average gray value; no other knowledge (e.g. the noise variance, training data etc.) is needed.We analyse the theoretical properties of the method, then test the performance of our time estimation procedure experimentally, and demonstrate that it yields near-optimal results for a wide range of noise levels and for various filtering methods.