基于相关性分析的多标签特征选择方法

针对现有大多数多标签特征选择算法未能有效去除特征空间冗余特征,同时也忽略了标签差异性的现状,提出一种基于相关性分析的多标签特征选择方法,利用特征之间的相关度对特征进行分组,解决了特征之间的相关性问题.根据样本所对应的标签属性对样本做一个正负类的聚类,对于正样本和负样本所构成的正类簇和负类簇单独确定其聚类个数,并计算原特征到正负类簇中各个类中心的距离,如此便产生了标签特定特征空间;将标签共享的特征空间和标签特定特征空间融合,考虑到多个标签之间的个性和关联性,解决了标签的差异性问题.实验测试表明,相较于现有的多标签特征选择算法,提出的基于相关性分析的多标签特征选择方法在各个分类指标上均有较优的表现,充分证明了该方法的有效性.     

一种新型混合特征选择方法及其在入侵检测中的应用

针对高维数据包含的不相关和冗余特征影响检测方法性能的问题,提出了集成filter和wrapper方法的混合特征选择新方法.采用基于信息增益的filter方法,删除不相关特征;采用基于改进的自适应遗传算法和评价函数的wrapper方法,获取最优特征子集.在入侵检测中的应用表明,该方法能降低特征选择的时间,检测率和虚警率均优于其它方法.     

Using maximum likelihood method to detect adaptive evolution of HCV envelope protein-coding genes

The basic process of adaptive evolution by natural selection is the replacement of one allele gene by an- other with a higher fitness in a population. Detecting the adaptive evolution would be helpful for better un- derstanding of bio-evolutionary mechani…     

基于二元语义GRP法的应急资源动员合作模式选择方法

为对应急资源动员合作模式的选择提供量化依据,建立了应急资源动员中地方政府合作的“介入—强度—方式”三维结构模型,从中选取4种典型合作模式,将模式选择问题转化为具有语言评价信息的多指标群决策问题。采用基于二元语义灰色关联投影的信息处理群决策方法,构建了合作模式的选择模型。依据传统投影分析方法的基本思想,通过计算备选方案对正理想方案和负理想方案的投影值,得到备选方案对正理想方案的相对贴近度,最终确定最优方案。在理论研究基础上,对京津冀地区应急资源动员合作模式进行实证分析,发现京津冀地区的应急资源动员合作更适合采取上级政府介入的强行政合作模式。基于专家评价数据的算例结果表明,该方法能够较好地反映合作模式与客观情况的匹配程度,在模式选择中发挥较好的作用。      

Episodic evolution of prolactin gene in primates

In the present study, we obtained exon 2--5 of prolactin (PRL) gene from four primate species by PCR and sequencing. Adding other genes available in GenBank, we calculate amino acid substitution rates for prolactin gene in primate. Comparison of nonsynonymous substitution rate to synonymous substitution rate ratios shows no evidence of positive selection for any lineage of primate prolactin gene.According to this and the facts that (i) no sites under positive selection are inferred by using maximum-likelihood method;(ii) among 32 amino acid replacement that occurred along the rapid evolutionary phase, only two are included in the 40 functionally important residues, indicating that amino acid replacement tends to occur in those functionally unimportant residues; (iii) partial of prolactin function is replaced by placental lactogen in primate at the rapid evolutionary phase of prolactin gene, we thus deem that it is relaxation of purifying selection to some extent rather than positive selection that enforces the rapid evolution of primate prolactin gene.     

基于特征对比的领域泛化算法自适应选择方法

在自动驾驶、医疗等领域,模型的泛化性是衡量其安全性的重要指标。领域泛化算法选择方法可以指导使用者快速准确地选出适合的模型训练算法。针对目前尚缺乏有效的算法选择方法的问题,提出一种基于对比学习的领域泛化算法的特征对比(feature contrast, FeCo)选择方法。依据正例和负例选择策略选择特征,采用点积的方式计算特征相似度,最后通过噪声对比估计(info noise contrastive estimation, InfoNC)计算得分。使用该得分评估了同一类特征的聚合程度和不同类特征的分离程度,在3个数据集共200个领域泛化模型上进行验证。实验结果表明,在所有的方法中FeCo是唯一结果稳定的方法,FeCo的结果和模型真实泛化误差的相关性最高可达0.89,且运行时间缩短超过60倍。     

Detecting selection using a single genome sequence of M. tuberculosis and P. falciparum

Selective pressures on proteins are usually measured by comparing nucleotide sequences. Here we introduce a method to detect selection on the basis of a single genome sequence. We catalogue the relative strength of selection on each gene in the entire genomes of Mycobacterium tuberculosis and Plasmodium falciparum. Our analysis confirms that most antigens are under strong selection for amino-acid substitutions, particularly the PE/PPE family of putative surface proteins in M. tuberculosis and the EMP1 family of cytoadhering surface proteins in P. falciparum. We also identify many uncharacterized proteins that are under strong selection in each pathogen. We provide a genome-wide analysis of natural selection acting on different stages of an organism's life cycle: genes expressed in the ring stage of P. falciparum are under stronger positive selection than those expressed in other stages of the parasite's life cycle. Our method of estimating selective pressures requires far fewer data than comparative sequence analysis, and it measures selection across an entire genome; the method can readily be applied to a large range of sequenced organisms.     

Detecting recent positive selection in the human genome from haplotype structure

The ability to detect recent natural selection in the human population would have profound implications for the study of human history and for medicine. Here, we introduce a framework for detecting the genetic imprint of recent positive selection by analysing long-range haplotypes in human populations. We first identify haplotypes at a locus of interest (core haplotypes). We then assess the age of each core haplotype by the decay of its association to alleles at various distances from the locus, as measured by extended haplotype homozygosity (EHH). Core haplotypes that have unusually high EHH and a high population frequency indicate the presence of a mutation that rose to prominence in the human gene pool faster than expected under neutral evolution. We applied this approach to investigate selection at two genes carrying common variants implicated in resistance to malaria: G6PD and CD40 ligand. At both loci, the core haplotypes carrying the proposed protective mutation stand out and show significant evidence of selection. More generally, the method could be used to scan the entire genome for evidence of recent positive selection.     

改进的mRmR特征选择方法在人体行为识别中的应用

在基于惯性传感器人体行为识别的研究中,特征选择的作用是挑选相关特征,以提高分类性能,去除冗余特征以降低计算复杂度。针对传统的过滤式特征选择方法仅使用一种相关度量系数筛选特征效果不佳的问题,提出一种改进的基于最大相关与最小冗余(mRmR)准则的特征选择方法。该方法在基于mRmR准则下,采用多种相关度量系数融合的方式,在考虑分类类别的条件下,分析待挑选特征与已选特征间的相关性对特征筛选可能产生的积极影响,以去除部分冗余、不相关特征,进而得到初选特征子集;然后利用二进制数对筛选后的特征编码,通过遗传算法搜索最优或次优特征子集。分别使用SVM和KNN分类器对7种日常行为进行分类。实验结果表明,与其他几种方法相比,该方法对实验分类的7种行为有最高的总体平均识别精度,通过SVM和KNN分类的各行为总体平均识别精度分别达到了97.02%和95.73%,与传统的mRmR方法相比,分别提高了13.72%和9.92%。     

利用极值蚁群优化的制粉出力建模变量选择算法

针对制粉出力难以直接测量,以及制粉系统包含变量多且各变量间耦合性强的问题,提出了一种利用极值蚁群优化的制粉出力建模变量选择法,并采用支持向量机根据所选变量建立了制粉出力的预测模型.该算法基于蚁群优化的正反馈原理,对蚂蚁搜索到的各个变量的相对重要性加以区分,并根据幂律分布选择重要性较小的变量进行变异,使得较差解不断得到改善,从而引导蚂蚁朝着最优解的方向搜索.采用制粉系统现场数据对所提算法、蚁群算法和蚁群遗传算法进行比较,结果表明,所提算法具有更快的收敛速度,且由其所选变量建立的制粉出力模型具有较高的预测精度.     

Inefficient positive selection of T cells directed by haematopoietic cells.

Intrathymic differentiation of alpha beta TCR+ T cells depends on positive selection of CD4+CD8+ thymocytes by thymic major histocompatibility complex (MHC) molecules. Positive selection allows the maturation of only those T cells capable of restricted antigen recognition in the context of the hosts' MHC alleles. Studies of normal or T-cell receptor-transgenic mice engrafted with MHC-different bone marrow or thymuses support the conclusion that positive selection is directed by MHC molecules expressed on non-haematopoietic cells, presumably thymic epithelial cells. Here we, present contrary evidence that class I MHC molecules expressed by haematopoietic cell types direct positive selection of CD8+ T cells, though at a reduced rate compared with positive selection directed by thymic epithelial cells. The identity of cell types that direct positive selection bears directly on mechanistic models of the process, including the idea that thymic epithelial cell MHC molecules uniquely present specialized peptides that mediate positive selection, and the notion that thymic epithelial cells express unique differentiation-inducing cell surface molecules.     

战略性新兴产业选择评价指标与综合决策模型

分析了部分省市战略性新兴产业选择决策方面存在的问题,明确了战略性新兴产业选择的标准,提出了指标体系构建的一般流程和建立战略性新兴产业评价指标体系应当遵循的原则,运用层次分析法（AHP）和专家调查相结合,建立了战略性新兴产业评价指标体系,确定了各指标的权重。对多指标加权灰靶综合决策评估模型的算法步骤进行梳理,并根据所建立的战略性新兴产业评价指标体系和多指标加权灰靶综合决策评估模型对某省6个备选的战略性新兴产业进行综合评价,明确了各产业的优先顺序,所得结果可以作为政府进行战略性新兴产业选择决策的依据。     

Thymic selection threshold defined by compartmentalization of Ras/MAPK signalling

A healthy individual can mount an immune response to exogenous pathogens while avoiding an autoimmune attack on normal tissues. The ability to distinguish between self and non-self is called 'immunological tolerance' and, for T lymphocytes, involves the generation of a diverse pool of functional T cells through positive selection and the removal of overtly self-reactive thymocytes by negative selection during T-cell ontogeny. To elucidate how thymocytes arrive at these cell fate decisions, here we have identified ligands that define an extremely narrow gap spanning the threshold that distinguishes positive from negative selection. We show that, at the selection threshold, a small increase in ligand affinity for the T-cell antigen receptor leads to a marked change in the activation and subcellular localization of Ras and mitogen-activated protein kinase (MAPK) signalling intermediates and the induction of negative selection. The ability to compartmentalize signalling molecules differentially in the cell endows the thymocyte with the ability to convert a small change in analogue input (affinity) into a digital output (positive versus negative selection) and provides the basis for establishing central tolerance.     

带有多属性偏好决策模型的应用

多属性决策方法是决策者在管理和项目选择中的重要辅助工具,在分析多属性决策问题的基础上,依据LINMAP法的基本思想,在给出群体一致度和非一致度的基础上,构造了一个估计正理想点和权重的线性规划模型,通过计算每个方案与正理想点之间的距离来确定最优方案.     

小麦耐甘露醇变异细胞系的离体筛选及植株再生Ⅰ

目的 离体筛选小麦耐甘露醇变异细胞系并再生植株。方法 以耐盐系小麦850512r的幼穗和幼胚为外植体诱导胚性愈伤组织。以所得到的胚性愈伤组织为起始材料，采用一步和多步正筛选法离体筛选耐甘露醇变异细胞系。结果 成功获得小麦耐甘露醇变异细胞系，并再生植株。结论 小麦耐甘露醇变异细胞系的离体筛选并再生植株成功，证明在细胞水平上筛选并获得具有潜在耐水分胁迫特性的小麦新品系是完全可能的。     

Dampening of death pathways by schnurri-2 is essential for T-cell development

Generation of a diverse and self-tolerant T-cell repertoire requires appropriate interpretation of T-cell antigen receptor (TCR) signals by CD4(+?) CD8(+) double-positive thymocytes. Thymocyte cell fate is dictated by the nature of TCR-major-histocompatibility-complex (MHC)-peptide interactions, with signals of higher strength leading to death (negative selection) and signals of intermediate strength leading to differentiation (positive selection). Molecules that regulate T-cell development by modulating TCR signal strength have been described but components that specifically define the boundaries between positive and negative selection remain unknown. Here we show in mice that repression of TCR-induced death pathways is critical for proper interpretation of positive selecting signals in vivo, and identify schnurri-2 (Shn2; also known as Hivep2) as a crucial death dampener. Our results indicate that Shn2(-/-) double-positive thymocytes inappropriately undergo negative selection in response to positive selecting signals, thus leading to disrupted T-cell development. Shn2(-/-) double-positive thymocytes are more sensitive to TCR-induced death in vitro and die in response to positive selection interactions in vivo. However, Shn2-deficient thymocytes can be positively selected when TCR-induced death is genetically ablated. Shn2 levels increase after TCR stimulation, indicating that integration of multiple TCR-MHC-peptide interactions may fine-tune the death threshold. Mechanistically, Shn2 functions downstream of TCR proximal signalling compenents to dampen Bax activation and the mitochondrial death pathway. Our findings uncover a critical regulator of T-cell development that controls the balance between death and differentiation.     

Calcineurin sets the bandwidth for discrimination of signals during thymocyte development

At critical times in development, cells are able to convert graded signals into discrete developmental outcomes; however, the mechanisms involved are poorly understood. During thymocyte development, cell fate is determined by signals originating from the alphabeta T-cell receptor. Low-affinity/avidity interactions between the T-cell receptor and peptide-MHC complexes direct differentiation to the single-positive stage (positive selection), whereas high-affinity/avidity interactions induce death by apoptosis (negative selection). Here we show that mice deficient in both calcineurin and nuclear factor of activated T cells (NFAT)c2/c3 lack a population of preselection thymocytes with enhanced ability to activate the mitogen-activated protein kinase (Raf-MEK-ERK) pathway, and fail to undergo positive selection. This defect can be partially rescued with constitutively active Raf, indicating that calcineurin controls MAPK signalling. Analysis of mice deficient in both Bim (which is required for negative selection) and calcineurin revealed that calcineurin-induced ERK (extracellular signal-regulated kinase) sensitization is required for differentiation in response to 'weak' positive selecting signals but not in response to 'strong' negative selecting signals (which normally induce apoptosis). These results indicate that early calcineurin/NFAT signalling produces a developmental period of ERK hypersensitivity, allowing very weak signals to induce positive selection. This mechanism might be generally useful in the discrimination of graded signals that induce different cell fates.     

Positive Darwinian selection in human population: A review

This paper reviews a large number of genes under positive Darwinian selection in modern human populations, such as brain development genes, immunity genes, reproductive related genes, percep- tion receptors. The research on the evolutionary property of these genes will provide important insight into human evolution and disease mechanisms. With the increase of population genetics and com- parative genomics data, more and more evidences indicate that positive Darwinian selection plays an indispensable role in the origin and evolution of human beings. This paper will also summarize the methods to detect positive selection, analyze the interference factors faced and make suggestions for further research on positive selection.     

基于遗传算法的入侵检测系统特征选择方法研究

在入侵检测系统中,分类器所选取的特征对系统的性能有很大的影响,大量冗余和不相关特征的存在会降低系统的正确性和实时性,因此如何选取出最优特征组合成为研究的热点问题。在研究当前各种特征选择方法的基础上,提出了一种基于遗传算法的特征组合选择方法。使用遗传算法搜索特征空间,依据Fisher准则计算各种特征组合的分类能力,根据计算结果对特征组合进行选择、交叉、变异,通过多次反复迭代最终选取出最优的特征组合。在实验中分别使用全部特征和选取出的最优特征组合的进行分类验证,最终证明选取出的最优特征组合能够使入侵检测系统在保持高检测率和低误报率的同时具有较高的检测效率,提高了系统的整体性能。     

水平移动式抛丸机叶片受力的有限元分析

为提高水平移动式抛丸机前曲叶片的使用寿命,运用ANSYS软件,建立了抛丸机叶片的有限元模型,以SOLID45和SURF154单元共同划分叶片网格,采用节点载荷加载方式。结果显示：叶片总体为低应力分布,但存在明显的应力集中点,叶片受弹丸正压力的最大值位于叶片边缘。叶片的失效形式是应力集中引起的疲劳断裂和正压力过大导致的表面磨损。提高叶片的使用寿命,除选用合适的材料提高叶片的耐磨性外,还需要设计合理的叶片形状,以降低最大正压力的影响。