基于核函数的非线性分类相关分析及其在化学模式识别中的应用

与统计分析和神经网络相比,基于结构风险最小的支持向量机有更好的分类性能。它用于非线性分类时,先将样本映射到更高维的特征空间,往往会增加复共线性与冗余信息,将影响样本分布,降低线性支持向量机分类器(LSVC)的预测性能。本研究提出非线性分类相关分析算法(NLCCA),利用核函数技术,无需了解非线性映射的算式,从特征空间的样本映像中提取分类相关成分,以消除冗余信息,改善样本分布。由此构建的NLCCA-LSVC集成分类器具有优良的预测性能。经模拟数据的测试,并实际用于两个复杂的化学模式识别问题,均取得令人满意的效果,也印证了算法的有效性。     

化学计量学非线性偏最小二乘算法进展评述

评述了偏最小二乘法(PLS)近年来非线性化的进展以及重要算法的原理、特性和应用。一类是对自变量进行非线性前处理的方法,包括二次变换、Chebychev变换和基于机理的变换;另一类是基于内部非线性映射的PLS,主要有多项式映射、样条函数映射、由遗传算法辅助实施的任意非线性映射和神经网络映射;介绍了具有类似功能的正交网模型。本文还就算法的普适性、稳健性和简易性,及其非线性化的演变进程予以评述、总结和展望。     

中药复杂组效关系的变结构神经网络辨识方法

针对中药复杂组效关系的辨识问题,研究了变结构多层前馈神经网络,推导出一种新型的变结构网络学习算法,成功地应用于中药川芎药效活性预测计算.该方法从一个规模较小的网络出发,当网络无法达到预定的学习精度时,自动增加隐含层神经元个数,并在原有学习结果的基础上确定新的网络参数,自适应地确定前馈神经网络结构,可用于处理复杂化学模式信息.计算机仿真实验结果表明,该方法能有效地确定多层前馈神经网络的最佳结构,提高网络学习效率和函数逼近精度,解决复杂非线性函数映射关系准确建模问题.     

用局部拟合主成分回归计算光度分析法测定黄连生物碱

针对具有样本数据非无匀分布和非线性特点的光度分析问题，提聘种局部拟合 主成分回归法，用于中药多组分计算测定。该方法根据待测样本与各已知样本光度 分析数据的欧式距离确定相应的权值，将部分权值较大的样本组成校正集，并用分 段线性拟合算法建立待测样本的校正预测模型，将其用于分析黄连的药根碱、巴巴 亭和小檗碱等三种生物碱，所得预测均方根误差分别为0.023,0.0400和0.052，优 于主成分回归法、偏最小二乘法以及人工神经元网络法所得结果。这表明，本方法 用于中药光度分析能获得较为准确的计算分析结果。     

便携式气相色谱仪中嵌入式谱峰识别算法设计

设计了嵌入式气相色谱谱峰识别算法,用于便携式气相色谱仪的数据分析处理,现场使用时能够不连接计算机就得出分析结果。该算法提出了用于谱峰识别的序列映射方法,减小了对斜率阈值的依赖,增强了谱峰分析的准确性和稳定性。通过实验对比可知,即使对于复杂组分的样本谱图,该嵌入式分析算法也能较好地分析提取出样品的谱峰数据,其分析结果与实验室色谱工作站软件分析结果非常接近,能够满足用户现场使用的分析需求,提高了便携仪器在现场使用的快捷性和便利性。     

火焰结构的平面激光诱导荧光技术观测

用平面激光诱导荧光（PLIF）技术测量了烧煤油的酒精喷灯单脉冲激光诱导荧光，由平面荧光图可清晰地看到氢氧基与C2、C3在火焰中的位置，同时，也看到了煤油火焰由发生、发展到熄灭过程中火焰结构的变化；PLIF技术是研究火焰结构的有力工具。     

基于样本优化选取的光谱重建方法研究

针对光谱反射率重建中已有样本选取方法的不足,提出了一种基于核模糊C聚类的样本优化选取方法。该方法综合考虑了光谱反射率空间的广泛性和色度空间的相似性,较大程度满足了光谱重建的精度。首先采用已有样本选取法在光谱反射率空间选取C个样本作为聚类初始点,再将原光谱转化到色度空间进行聚类,同时引入核函数将二维色度空间映射到三维特征空间,使得特征数据线性可分,从而达到更好的划分效果。实验结果表明,使用该方法选取训练样本进行光谱反射率重建能够进一步提高光谱重建精度,色度评价和光谱评价结果均好于已有方法。     

基于近红外光谱的果蔬脆片品质评价方法研究

本文选取苹果脆片为研究对象,结合最新果蔬脆片研究成果和支持向量机机器学习方法,以苹果脆片近红外光谱吸收值数据样本为基础,根据不同苹果样本、不同时序的红外光谱吸收值数据特征信息,产生不同的映射关系,将红外光谱吸收值映射到对应样本在对应时刻的各项品质指标(如水分含量),然后利用支持向量机(Support Vector Machine,SVM)方法对近红外光谱吸收值的数据样本进行训练,通过调整模型参数来体现不同苹果个体间的生理差异,以取得更接近于实际情况的品质指标预测效果。模拟仿真实验表明该模型的平均相关系数(R~2)达到93.81%,均方误差(MSE)为0.0057。该研究可为果蔬脆片的评价体系提供新的参考方法。     

基于Person VII核函数的支持向量机及其在化学模式分类中的应用

化学模式分类问题通常是非线性的，而且比较复杂，难以用经典统计方法建立分类判别模型。以支持向量机（SVM）构建的分类器具有更好的分类性能。对于非线性分类，SVM通过核函数将其映射到高维特征空间中，然后再进行线性分类。因此，核函数往往是决定SVM非线性分类性能的关键。实际应用时，一般通过选择几种核函数，并对其参数进行优化，然后根据分类器的预测性能来决定，训练过程非常耗时，而且结果难以保证最优。为此，采用一种通用性的核函数，即PersonVg核函数（PUKF），它可取代目前常用的几种核函数，可避免SVM非线性分类器训练过程的核函数选择问题。本研究将基于PUKF的SVM分类器应用于两个化学模式分类问题，均取得了较好的结果。对于多类分类，设计了一种子分类器的构造方法，它在分类性能保持较好的情况下，简化了多类分类器结构，大大降低了计算量。     

边缘保持滤波技术在空间色域映射中的应用

正色域映射是跨媒体图像复制中的一项关键技术,主要用以解决图像颜色信息在不同色域大小的颜色设备之间的准确传输。目前已有不少成熟的色域映射算法,其中,按照映射原理可以将色域映射算法分为逐点色域映射和空间色域映射两大类。逐点色域映射方法由于不考虑图像中像素间的颜色关系,因此会造成映射图像的细节丢失。空间色域映射方法考虑了每个映射点的空间颜色关系,是一种更为合理和更符合人类视觉特性的映射方法。但是,目前空间色域映射算法需要进行很多细节补偿,这将会产生视觉上较为明显的光晕效应。     

Relationship between the retention characteristics on an alumina column and physico-chemical parameters of some environmental pollutants

The retention behavior of 16 environmental pollutants was studied on alumina and porous graphitized carbon (PGC) columns using n-hexane as eluent. The relationship between the logarithm of the capacity factors determined on the alumina column and the physico-chemical characteristics of the solutes was elucidated by principal component analysis (PCA) followed by two-dimensional nonlinear mapping. The 12 original variables can be reduced to four with only a 10% loss of information. The logarithm of the capacity factor formed a cluster with the hydrogen donor and acceptor properties of the solutes and their Taft's constant on the two-dimensional nonlinear map of PC loadings indicating that both steric and electronic parameters play a considerable role in the retention mechanism on alumina support. Alcohols, aromatics and chlorinated alkanes formed separate clusters on the two-dimensional nonlinear map of PC variables suggesting that their retention mechanism may be different on the alumina column. Solutes were not retained on the PGC surface proving that under normal-phase conditions the retention capacity of alumina can be higher than that of PGC.     

Stochastic proximity embedding

We introduce stochastic proximity embedding (SPE), a novel self-organizing algorithm for producing meaningful underlying dimensions from proximity data. SPE attempts to generate low-dimensional Euclidean embeddings that best preserve the similarities between a set of related observations. The method starts with an initial configuration, and iteratively refines it by repeatedly selecting pairs of objects at random, and adjusting their coordinates so that their distances on the map match more closely their respective proximities. The magnitude of these adjustments is controlled by a learning rate parameter, which decreases during the course of the simulation to avoid oscillatory behavior. Unlike classical multidimensional scaling (MDS) and nonlinear mapping (NLM), SPE scales linearly with respect to sample size, and can be applied to very large data sets that are intractable by conventional embedding procedures. The method is programmatically simple, robust, and convergent, and can be applied to a wide range of scientific problems involving exploratory data analysis and visualization.     

A PacI/SwaI map of the Pseudomonas aeruginosa PAO chromosome.

The PacI/SwaI macrorestriction map of the 5.9 Mb Pseudomonas aeruginosa PAO (DSM 1707) genome was constructed by fragment size comparisons of complete single and double digests and hybridization with probes of known genomic localization. P. aeruginosa PAO contains five recognition sites for PacI and six sites for SwaI. An integrated DpnI/PacI/SpeI/SwaI map was obtained by fragment pattern and Southern blot hybridization analyses of all combinations of double digestions and by two-dimensional pulsed-field gel electrophoresis mapping protocols.     

Generalised technique for calculation of plane director profiles in bounded nematic liquid crystals

We propose an improved conformal mapping technique for analytical calculation of two-dimensional profiles of the nematic liquid crystal (NLC) director in a cell with a simply connected cross-sectional region. We consider the case of the strong anchoring and the piecewise constant director pretilt on the piecewise smooth curve which bounds the region. Obtained expressions for the director profile explicitly depend on the conformal mapping which maps the region onto the upper half plane of a complex plane. An advantage of our method in comparison with the standard conformal mapping technique is that it does not require the knowledge of the inverse mapping and the calculation of the integral in the Poisson formula. Proposed technique allows to take into account topological defects in the bulk of the NLC on the symmetry axis of the region. As an example of how the method can be used, we find an analytical expression for the director profile in a horizontal cylindrical groove partly filled with the NLC. We consider the case where a disclination line parallel to the axis of the groove occurs in the bulk of the NLC. The equilibrium position of the disclination line is found.     

A protein molecular weight map of ES2 clear cell ovarian carcinoma cells using a two-dimensional liquid separations/mass mapping technique

A molecular weight map of the protein content of ES2 human clear cell ovarian carcinoma cells has been produced using a two-dimensional (2-D) liquid separations/mass mapping technique. This method uses a 2-D liquid separation of proteins from whole cell lysates coupled on-line to an electrospray ionization-time of flight (ESI-TOF) mass spectrometer to map the accurate intact molecular weight (M(r)) of the protein content of the cells. The two separation dimensions involve the use of liquid isoelectric focusing as the first phase and nonporous silica reversed-phase high-performance liquid chromatography (HPLC) as the second phase of separation. The detection by ESI-TOF-MS provides an image of pI versus M(r) analogous to 2-D gel electrophoresis. Each protein is then identified based upon matrix-assisted laser desorption/ionization (MALDI)-TOF-MS peptide mapping and intact M(r) so that a standard map is produced against which other ovarian carcinoma cell lines can be compared. The accurate intact M(r) together with the pI fraction, and peptide map serve to tag the protein for future interlysate comparisons. An internal standard is also used to provide a means for quantitation for future interlysate studies. In the ES2 cell line under study it is shown that nearly 900 M(r) bands are detected over 17 pI fractions from pH 4 to 12 and a M(r) range up to 85 kDa and that around 290 of these bands can be identified using mass spectrometric based techniques. The protein M(r) is detected within an accuracy of 150 ppm and it is shown that many of the proteins in this human cancer sample are modified compared to the database. The protein M(r) map may serve as a highly reproducible standard Web-based method for comparing proteins from related human cell lines.     

Proteome analysis of rat polymorphonuclear leukocytes: a two-dimensional electrophoresis/mass spectrometry approach

The development of a two-dimensional (2-D) map of rat polymorphonuclear (PMN) leukocytes is here reported for the first time. The map is built up by utilizing a wide immobilized pH gradient (IPG), pH 3-10, in the first dimension and also a narrower IPG pH 4.5-8.5 gradient. In addition, the map is constructed by adopting the most recent protocols in 2-D mapping, which call for reduction and alkylation of the sample prior to the start of any electrophoretic step, including the IPG dimension. Fifty-two major protein spots have been so far identified by utilizing both matrix assisted laser desorption/ionization-time of flight (MALDI-TOF) and electrospray quadrupole (Q)-TOF mass spectrometry. A large number of house-keeping and cytoskeleton proteins were detected, together with proteins which are specific to PMN organelles or related to PMN functions such as phagocytosis and chemotaxis. The results obtained demonstrate the possibility of obtaining a single 2-D gel based proteomic map of PMN with representative proteins from different cellular compartments, also including membrane components, allowing the study of PMN protein expression on a proteome-wide scale. The aim of this project is to build an extensive database of such proteins, to be utilized for future studies where the expression of PMN proteins is used as a disease- or drug treatment marker.     

二能级体系的动态局域化

综述了我们在二能级体系的动态局域化方面所作的工作.通过把受驱二能级体系的量子动力学映射到在二维谐性势场和与之垂直的周期性磁场中粒子的经典动力学,使得对局域化机制理解更为深刻.应用这种方案就可以得到在线性偏振场作用下二能级体系的局域化条件.同时解析地分析了在脉冲场和圆偏振场驱动下二能级体系的局域化,发现圆偏振场诱导的局域化可以由复平面上单位圆中的椭圆完全刻画.     

Nitrosopeptide mapping: a novel methodology reveals s-nitrosylation of dexras1 on a single cysteine residue

S-Nitrosylation of specific cysteine residues is a reversible signaling mechanism of nitric oxide (NO) generated by NO synthase (NOS) enzymes. In some proteins, evidence has accumulated that more than one cysteine can be S-nitrosylated; however, it is difficult to distinguish S-nitrosylation on separate cysteine residues. We report a novel simple, sensitive, and specific procedure for nitrosopeptide mapping. Dexras1 is a monomeric G protein whose guanine nucleotide exchange activity is augmented by NO; the identity and number of its S-nitrosylated cysteines is unknown. We describe the radiolabeling of S-nitrosylated cysteine residues in Dexras1. A nitrosopeptide map, generated by two-dimensional peptide chromatography, reveals that only a single cysteine is S-nitrosylated following NO exposure. Mutagenesis of Cys11 abolished the effect of NO donors on Dexras1, implicating this residue in the NO-mediated activation of Dexras1.     

Prediction of Molecular Properties Using Molecular Topographic Map

Prediction of molecular properties plays a critical role towards rational drug design. In this study, the Molecular Topographic Map (MTM) is proposed, which is a two-dimensional (2D) map that can be used to represent a molecule. An MTM is generated from the atomic features set of a molecule using generative topographic mapping and is then used as input data for analyzing structure-property/activity relationships. In the visualization and classification of 20 amino acids, differences of the amino acids can be visually confirmed from and revealed by hierarchical clustering with a similarity matrix of their MTMs. The prediction of molecular properties was performed on the basis of convolutional neural networks using MTMs as input data. The performance of the predictive models using MTM was found to be equal to or better than that using Morgan fingerprint or MACCS keys. Furthermore, data augmentation of MTMs using mixup has improved the prediction performance. Since molecules converted to MTMs can be treated like 2D images, they can be easily used with existing neural networks for image recognition and related technologies. MTM can be effectively utilized to predict molecular properties of small molecules to aid drug discovery research.     

Receptor Mapping by Comparative Molecular Field Analysis of Phospholipase A2 Inhibitors

In structure-based design of phospholipase A₂ inhibitors, comparative molecular field analysis is performed on known phospholipase A₂ (PLA₂) inhibitors to build a pseudo-receptor map of the enzyme, which is compared to maps from probe searches with proton and methane. The pseudo-receptor map resembled in steric properties to the map from a probe search. Hence known ligands can be used to deduce the shape and chemical properties of the binding site of a receptor. Useful information for drug design can be obtained from such receptor mapping.