基于网格密度和距离信息特征的聚类算法

摘要: 提出的基于网格密度和距离信息特征的聚类方法（GDD）通过构建基于距离的跃迁函数将局域密度波动特征与距离分布信息联系在一起,根据计算出的跃迁函数值扩展和增长聚类簇,从而避免了多数基于网格和密度的聚类算法存在的单调性搜索聚类缺陷。结合具体的跃迁函数在不同测试集上的实验结果表明：GDD算法不仅能够发现任意形状的簇和对噪音数据不敏感,且具有线性于网格数目的时间复杂性,能够回避密度分布不均对聚类结果的影响,更适合于对大规模真实数据集的聚类。     

An adaptive spatial clustering method for automatic brain MR image segmentation

In this paper, an adaptive spatial clustering method is presented for automatic brain MR image segmentation, which is based on a competitive learning algorithm – self-organizing map (SOM). We use a pattern recognition approach in terms of feature generation and classifier design. Firstly, a multi-dimensional feature vector is constructed using local spatial information. Then, an adaptive spatial growing hierarchical SOM (ASGHSOM) is proposed as the classifier, which is an extension of SOM, fusing multi-scale segmentation with the competitive learning clustering algorithm to overcome the problem of overlapping grey-scale intensities on boundary regions. Furthermore, an adaptive spatial distance is integrated with ASGHSOM, in which local spatial information is considered in the clustering process to reduce the noise effect and the classification ambiguity. Our proposed method is validated by extensive experiments using both simulated and real MR data with varying noise level, and is compared with the state-of-the-art algorithms.     

Photonics: lasers producing tailored beams

Compact lasers that can produce a range of beam patterns are important for progress in several areas, including the improvement of optical tweezers, ultra-high-density optical memory and microfluidics. Here we engineer photonic crystals to generate semiconductor lasers that produce a range of beam patterns while maintaining stable single-mode oscillation. Our results could contribute to the realization of compact lasers that are capable of producing diverse beam patterns on demand.     

Resonance in the electron-doped high-transition-temperature superconductor Pr0.88LaCe0.12CuO4-delta

In conventional superconductors, the interaction that pairs the electrons to form the superconducting state is mediated by lattice vibrations (phonons). In high-transition-temperature (high-T(c)) copper oxides, it is generally believed that magnetic excitations might play a fundamental role in the superconducting mechanism because superconductivity occurs when mobile 'electrons' or 'holes' are doped into the antiferromagnetic parent compounds. Indeed, a sharp magnetic excitation termed 'resonance' has been observed by neutron scattering in a number of hole-doped materials. The resonance is intimately related to superconductivity, and its interaction with charged quasi-particles observed by photoemission, optical conductivity, and tunnelling suggests that it might play a part similar to that of phonons in conventional superconductors. The relevance of the resonance to high-T(c) superconductivity, however, has been in doubt because so far it has been found only in hole-doped materials. Here we report the discovery of the resonance in electron-doped superconducting Pr0.88LaCe0.12CuO4-delta (T(c) = 24 K). We find that the resonance energy (E(r)) is proportional to T(c) via E(r) approximately 5.8k(B)T(c) for all high-T(c) superconductors irrespective of electron- or hole-doping. Our results demonstrate that the resonance is a fundamental property of the superconducting copper oxides and therefore must be essential in the mechanism of superconductivity.     

子阵约束下实矩阵反问题的最小二乘解

提出了子阵约束下实矩阵反问题的最小二乘问题,给出了解的表达式．考虑了解集合对给定矩阵的最佳逼近问题,证明了最佳逼近问题解的存在性与唯一性,给出了求最佳逼近解的数值方法．将所得结果应用于解决子阵约束下实矩阵特征反问题．     

毛冠鹿与3种麂属动物的GAPDH基因序列的比较分析

麂亚科（Muntiacinae）动物包括麂属（Muntiacus）和毛冠鹿属（Elaphodus）,分布于东南亚、中国及印度,动物形态相似,但具有令人瞩目的细胞遗传学特性．本文以赤麂（Muntiacus muntjak）、黑麂（Muntiacus crinifrons）、小麂（Muntiacus reevesi）和毛冠鹿（Elaphodus cephalophus）4种麂亚科动物为材料,以3-磷酸甘油醛脱氧酶（GAPDH,管家基因）为靶基因,对毛冠鹿在麂亚科中分类地位进行分子鉴定．提取这4种动物的基因组DNA,根据人的GAPDH基因设计引物,进行PCR扩增．获得了约为300bp的扩增产物．将纯化后的扩增产物克隆到质粒pMD18＋TVector中,转化DH-5a菌,挑选阳性克隆。用M13-47／RV—M通用引物进行测序．测序的结果经clustal软件进行了同源性比较．结果显示：麂属动物种问GAPDH基因间的DNA差异在2．52％～4．37％之间,毛冠鹿与麂属3种动物的差异在1．79％-2．87％之间．这些变异以碱基的转换为主．本文结果支持毛冠鹿为独立属1种的分类观念．     

按序选取的并行存储系统数据分布方式研究

针对采用轮转方式的并行存储系统在增加新存储节点后的重均衡操作中移动数据量大、资源消耗高的问题,提出了按序选取的数据分布方式.它通过多次模拟数据重均衡操作来完成并行存储系统的数据分布,从而使得系统更易于节点扩展.基于此,又提出了优化按序选取方式,这样可以通过改变数据抽取规则来提高按序选取方式的访问并行度.理论分析和模拟实验表明,采用所提方式的并行存储系统,数据重均衡操作均可达到零无效移动率,从而最大程度地减少数据的移动量,降低系统的资源耗费.     

肠外营养在危重病人中的应用

目的探讨肠外营养在危重病人中的应用效果.方法将22例危重病患者给予常规抗感染及对症治疗,同时给予肠外营养支持治疗.17例对照组给予常规抗感染及对症治疗.结果病人的肝肾功能肠外营养后与营养前相比有意义(P＜0.05及P＜0.01).病人状况根据KPS评分有意义.对照组治疗前后肝肾功能则改善不明显.结论肠外营养对危重病人病情及肝肾功能有明显改善作用.     

基于覆盖的程度粗糙集模型

在基于覆盖的粗糙集模型的基础上,引入分类误差,建立了基于覆盖的程度粗糙集模型,并讨论了该模型的一些相关性质.