基于遗传算法的径向基神经网络及其在同时测定Fe Mn Cu Zn中应用

介绍了两种新的基于遗传算法的径向基神经网络(GA-Based RBFNN)训练算法.这两种算法均将遗传算法用于优化径向基神经网络的聚类中心和网络结构.第一种GA-Based RBFNN算法对所有训练样本采取二进制编码构成个体,优化径向基函数中心的选取和网络结构;第二种GA-Based RBFNN算法中,RBFNN采用自增长算法训练网络隐含层中心、采用十进制对距离因子ε编码构成染色体,优化网络.将两种GA-Based RBFNN算法应用于Fe、Mn、Cu、Zn同时测定的光谱解析,计算结果表明,本文的GA-Based RBFNN算法较通常的遗传算法与径向基人工神经网络(GA-RBFNN)联用,即在GA选择变量的基础上,再用RBFNN作数据解析的GA-RBFNN方法,在增强网络的泛化能力、提高预测的准确性等方面具有明显的优势.从这两种GA-Based RBFNN的比较看,第二种算法在性能上优于第一种算法.     

基于GA-RBFNN算法的列车车轮踏面损伤识别

为了实现列车车轮踏面损伤识别,提出了一种基于GA-RBFNN算法的货车车轮踏面损伤识别方法。该算法采用浮点数编码将RBFNN的中心参数和宽度进行了编码,利用GA的选择、交叉和变异操作优化网络参数,权值采用最小二乘法确定。利用该算法和BP算法、传统的RBFNN算法进行了剥离和擦伤识别的对比实验,结果表明：GA-RBFNN算法对剥离、擦伤和非损伤三类样本的测试集的识别率高于传统的RBFNN算法和BP算法,而且GA-RBFNN算法的进化代数远远小于BP算法和传统的RBFNN算法迭代次数。     

基于Fuzzy-RBFNN的多模医学图像融合方法研究

文中将模糊推理理论与径向基神经网络相结合构造了一个基于模糊推理的径向基神经网络（Fuzzy—RBFNN）应用于多模医学图像融合,并应用遗传算法训练网络获得网络参数,可自适应地完成多模医学图像融合。通过与基于梯度的金字塔融合方法的实验比较,证明了算法的有效性与可行性。     

基于软计算的汽车线控转向系统故障诊断

针对线控转向汽车的可靠性和安全性以及故障诊断方法的不足,提出了一种基于软计算的汽车线控转向故障诊断方法,该方法利用软计算中的粗糙集和粒子群优化的径向基神经网络进行结合。将粗糙集作为径向基神经网络的输入处理,对样本数据进行属性约简,约简后的属性集作为径向基神经网络的输入以达到缩短网络训练时间的目的。采用粒子群算法对径向基神经网络的基函数中心值和宽度进行编码和寻优,并使用得到的最优中心值和宽度组建径向基神经网络,使得径向基神经网络的样本训练误差相比未优化之前有一定程度的降低。然后使用训练好的神经网络对故障样本进行测试,测试结果表明,该方法加快了神经网络的训练速度,提高了神经网络的诊断准确度。     

改进的OLS算法选择RBFNN中心的方法

提出了一种优化选择径向基神经网络数据中心的算法,该算法结合了Kohonen网络的模式分类能力,将初步分类结果用做RBFNN的初始数据中心,然后采用OLS算法进行优化选择,对比仿真实验表明该算法效果比单独使用OLS算法生成的RBFNN性能更好。     

基于Kohonen网络和OLS算法的RBFNN中心选择方法

提出了一种优化选择径向基神经网络数据中心的算法,该算法结合了Kohonen网络的模式分类能力,将初步分类结果用作RBFNN的初始数据中心,然后采用OLS算法进行优化,对比仿真实验表明该算法效果比单独使用OLS算法生成的RBFNN性能更好。     

基于粒子群算法的RBF网络参数优化算法

针对神经网络的一些缺陷,研究神经网络基于粒子群优化的学习算法,将粒子群优化算法用于RBF神经网络的学习训练。提出了一种基于粒子群优化（PSO）算法的径向基（RBF）网络参数优化算法,首先利用减聚类算法确定网络径向基函数中心的个数,再用PSO算法优化径向基函数的中心及宽度,最后用PSO算法训练隐含层到输出层的网络权值,找到神经网络权值的最优解,以达到优化神经网络学习的目的。最后,通过一个实验与最小二乘法优化的神经网络进行了比较,验证了算法的有效性。     

基于K0honen网络和OLS算法的RBFNN中心选择方法

提出了一种优化选择径向基神经网络数据中心的算法,该算法结合了Kohonen网络的模式分类能力,将初步分类结果用作RBFNN的初始数据中心,然后采用OLS算法进行优化,对比仿真实验表明该算法效果比单独使用OLS算法生成的RBFNN性能更好。     

修正高斯模型神经网络的色谱重叠峰解析

目的：研究色谱重叠峰解析新方法。方法：构造了以修正高斯模型（EMG）为基函数的径向基函数神经网络（EMG＿RBFNN）,在网络学习算法中提出采用基于共享小生境技术的约束最优保留两阶段遗传算法：结构学习和参数最速梯度下降进化,从而使EMG＿RBFNN具有结构学习能力 ,使该神经网络能够适应组分数未知的色谱重叠峰解析。尤其是在将气相和高效液相色谱EMG模型参数τ、σ和tR之间的近似线性关系作为可行域约束条件引入算法后,极大地限制了算法可行的解空间,减少了局部最优解出现的概率,提高了算法运行效率。结果：将此新方法分别应用于计算机仿真色谱和45例中药材HPLC,算法可以以较高的精度解析出仿真色谱,实际色谱峰解析也有很高的解析精度。结论：该方法非常适用于组分数未知的各种色谱（重叠）峰的解析。     

改进的径向基函数网络的研究及应用

针对径向基函数网络和传统遗传算法的一些不足,提出引入一种自适应机制的浮点数编码的遗传算法,并将其与梯度下降法混合交互运算,作为径向基函数网络的学习算法,形成了基于改进遗传算法的径向基函数网络,它克服了径向基函数网络的学习算法上的缺陷。采用改进的遗传算法,无需计算梯度等,限制很少,还可用模型的预测性能作为优化目标。同时,也解决了单独利用遗传算法往往只能在短时间内寻找到接近全局最优解的近似解这一问题。最后将该算法应用到某地区电力负荷预测取得理想效果。     

Pointcuts, advice, refinements, and collaborations: similarities, differences, and synergies

Aspect-oriented programming (AOP) is a novel programming paradigm that aims at modularizing complex software. It embraces several mechanisms including (1) pointcuts and advice as well as (2) refinements and collaborations. Though all these mechanisms deal with crosscutting concerns, i.e., a special class of design and implementation problems that challenge traditional programming paradigms, they do so in different ways. In this article we explore their relationship and their impact on modularity, which is an important prerequisite for reliable and maintainable software. Our exploration helps researchers and practitioners to understand their differences and exposes which mechanism is best used for which problem.     

Household technology adoption,use, and impacts: Past,present, and future

Since the 1980s, researchers have been studying the phenomenon associated with technology being diffused to the household. In this paper, three themes in that stream of research, specifically adoption, use, and impacts, are explored. Key studies from prior research within each theme are discussed and directions for future research are offered. The directions for future research range from investigating adoption issues associated with the digital divide to understanding the impacts of new technology and social networking sites on individuals and families. The evolving nature of the technology continues to offer interesting research directions and challenges, with the study of unintended consequences of technology use presenting, perhaps, the greatest opportunities.

Quantifiers, Anaphora, and Intensionality

The relationship between Lexical-Functional Grammar (LFG) functional structures (f-structures) for sentences and their semanticinterpretations can be formalized in linear logic in a way thatcorrectly explains the observed interactions between quantifier scopeambiguity, bound anaphora and intensionality.Our linear-logic formalization of the compositional properties ofquantifying expressions in natural language obviates the need forspecial mechanisms, such as Cooper storage, in representing thescoping possibilities of quantifying expressions. Instead, thesemantic contribution of a quantifier is recorded as a linear-logicformula whose use in a proof will establish the scope of thequantifier. Different proofs can lead to different scopes. In eachcomplete proof, the properties of linear logic ensure thatquantifiers are properly scoped.The interactions between quantified NPs and intensional verbs such asseek are also accounted for in this deductive setting. A singlespecification in linear logic of the argument requirements ofintensional verbs is sufficient to derive the correct readingpredictions for intensional-verb clauses both with nonquantified andwith quantified direct objects. In particular, both de dictoand de re readings are derived for quantified objects. Theeffects of type-raising or quantifying-in rules in other frameworksjust follow here as linear-logic theorems.While our approach resembles current categorial approaches inimportant ways (Moortgat, 1988, 1992a; Carpenter, 1993; Morrill, 1994)it differs from them in allowing the greater compositional flexibility ofcategorial semantics (van Benthem, 1991)while maintaining a precise connection to syntax. As a result, we areable to provide derivations for certain readings of sentences withintensional verbs and complex direct objects whose derivation inpurely categorial accounts of the syntax-semantics interface appearsto require otherwise unnecessary semantic decompositions of lexicalentries.     

Motility,mixing, and multicellularity

A fundamental issue in evolutionary biology is the transition from unicellular to multicellular organisms, and the cellular differentiation that accompanies the increase in group size. Here we consider recent results on two types of “multicellular” systems, one produced by many unicellular organisms acting collectively, and another that is permanently multicellular. The former system is represented by groups of the bacterium Bacillus subtilis and the latter is represented by members of the colonial volvocalean green algae. In these flagellated organisms, the biology of chemotaxis, metabolism and cell–cell signaling is intimately connected to the physics of buoyancy, motility, diffusion, and mixing. Our results include the discovery in bacterial suspensions of intermittent episodes of disorder and collective coherence characterized by transient, recurring vortex streets and high-speed jets of cooperative swimming. These flow structures markedly enhance transport of passive tracers, and therefore likely have significant implications for intercellular communication. Experiments on the Volvocales reveal that the sterile flagellated somatic cells arrayed on the surface of Volvox colonies are not only important for allowing motion toward light (phototaxis), but also play a crucial role in driving fluid flows that transport dissolved molecular species. These flows, generated by the collective beating of flagella, confer a synergistic advantage with regard to transport of nutrients and chemical messengers. They allow these species to circumvent a nutrient acquisition bottleneck which would exist if transport were purely diffusive, and thereby evolve to larger multicellular individuals. In both cases, a higher level of organization, specialization and complexity counteract the higher costs inherent to larger groups.     

Mandalas,screws, pears,and Klein bottles

The goal of this article is to present an informal introduction and tutorial on the production of aesthetically pleasing solid models. The article is intended for the nonmathematical reader interested in computer art. Simple generating formulas and recipes are included.     

Children, robotics, and education

Popular interest in robotics has increased astonishingly in the last few years. Robotics is seen by many as offering major new benefits in education at all levels. Before rushing to exploit this popularity, educators should ask serious questions about the universality and longevity of the robotics phenomenon. Is it a fashion? To be useful, the energy released by robotics must be sustained and universal, and the means of exploiting it must be systematic. Universities define their own robotics curriculum, but most schools lack both the resources and the freedom to do this, and must work with a national curriculum. If it can be shown that robotics has sustained potential in education, it seems inevitable that new ways need to be found to integrate it into the school curriculum.     

Complexity,convexity, and unimodality

A class of polygons termedunimodal is introduced. LetP = P₁,p ₂,...,p _n be a simplen-vertex polygon. Given a fixed vertex or edge, several definitions of the distance between the fixed vertex or edge and any other vertex or edge are considered. For a fixed vertex (edge), a distance measure defines a distance function as the remaining vertices (edges) are traversed in order. If for every vertex (edge) ofP a specified distance function is unimodal thenP is a unimodal polygon in the corresponding sense. Relationships between unimodal polygons, in several senses, andconvex polygons are established. Several properties are derived for unimodal polygons when the distance measure is the euclidean distance between vertices of the polygons. These properties lead to very simple 0(n) algorithms for solving a variety of problems that occur in computational geometry and pattern recognition. Furthermore, these algorithms establish that convexity is not the key factor in obtaining linear-time-complexity for solving these problems. The paper closes with several open questions in this area.     

Recipes,Algorithms, and Programs

In the technical literature of computer science, the concept of an effective procedure is closely associated with the notion of an instruction that precisely specifies an action. Turing machine instructions are held up as providing paragons of instructions that "precisely describe" or "well define" the actions they prescribe. Numerical algorithms and computer programs are judged effective just insofar as they are thought to be translatable into Turing machine programs. Nontechnical procedures (e.g., recipes, methods) are summarily dismissed as ineffective on the grounds that their instructions lack the requisite precision. But despite the pivotal role played by the notion of a precisely specified instruction in classifying procedures as effective and ineffective, little attention has been paid to the manner in which instructions "precisely specify" the actions they prescribe. It is the purpose of this paper to remedy this defect. The results are startling. The reputed exemplary precision of Turing machine instructions turns out to be a myth. Indeed, the most precise specifications of action are provided not by the procedures of theoretical computer science and mathematics (algorithms) but rather by the nontechnical procedures of everyday life. I close with a discussion of some of the rumifications of these conclusions for understanding and designing concrete computers and their programming languages.