一种鲁棒的被动声目标识别方法

针对噪声环境下一些声目标识别技术性能严重下降的问题,提出了一种基于一阶补偿的径向基函数网络模型,并给出了该网络参数选择方法。实际应用表明,这种神经网络对噪声具有较强的鲁棒性,与传统的径向基函数网络相比,其识别性能等同于信噪比提高大约１０～１５ｄＢ。     

An Adaptive Identification and Control Scheme Using Radial Basis Function Networks

1.INTaODUCTIONThemultilayerfeedforwardneuralnetworks(MFNN)hasbeenwidelyappliedinmanyfieldssuchaspatternrecognition,nonlinearoptimization,systemidentificationandcontrol.Thesimpleback-propagation(BP)learningalgorithmisusuallyadoptedtoupdateitsweights.TheBPalgorithm,however,suffersdrawbacksofslowconvergenceandunpredictablesolutionsduringlearning.Althoughmailyapproachessuchasvariablestepsize,conjugategradient,pseudo-Newtonmethods[4]andothermethods[3,10],havebeendevelopedtoovercometheproble…     

广义函数及其在力学中的应用述评

简要介绍了广义函数的发展,给出了力学中常用的广义函数（即奇异函数）的定义及其微分、积分运算法则,分析了广义函数在力学各分支中的应用现状,通过用广义函数表示集中力作用下固支圆板的面分布集度来求解板弯曲挠度的实例,说明运用广义函数解决一些力学问题的优越性．     

Betaine homocysteine S-methyltransferase: just a regulator of homocysteine metabolism?

Betaine homocysteine methyltransferase (BHMT), a Zn²⁺-dependent thiolmethyltransferase, contributes to the regulation of homocysteine levels, increases in which are considered a risk factor for cardiovascular diseases. Most plasma homocysteine is generated through the liver methionine cycle, in which BHMT metabolizes approximately 25% of this non-protein amino acid. This process allows recovery of one of the three methylation equivalents used in phosphatidylcholine synthesis through transmethylation, a major homocysteine-producing pathway. Although BHMT has been known for over 40 years, the difficulties encountered in its isolation precluded detailed studies until very recently. Thus, the last 10 years, since the sequence became available, have yielded extensive structural and functional data. Moreover, recent findings offer clues for potential new functions for BHMT. The purpose of this review is to provide an integrated view of the knowledge available on BHMT, and to analyze its putative roles in other processes through interactions uncover to date. Received 26 May 2006; received after revision 3 July 2006; accepted 24 August 2006     

DNA methylation of fly genes and transposons

The use of anti-5-methylcytosine antibodies in affinity columns allowed the identification of methylated sequences in the genome of Drosophila melanogaster adults. In view of the presence of transposable elements amongst the identified sequences, it has been suggested that DNA methylation is involved in transposon control in the fly genome. On the contrary, a reanalysis of these data furnishes several intriguing elements that could raise new questions about the role that DNA methylation plays in the fly genome. The aim of the present paper is to discuss some features that emerge from the analysis of the identified methylated sequences. Received 26 January 2006; received after revision 8 May 2006; accepted 2 June 2006     

Lipid-lowering drugs

Although a change in life-style is often the method of first choice for lipid lowering, lipid-lowering drugs, in general, help to control elevated levels of different forms of lipids in patients with hyperlipidemia. While one group of drugs, statins, lowers cholesterol, the other group, fibrates, is known to take care of fatty acids and triglycerides. In addition, other drugs, such as ezetimibe, colesevelam, torcetrapib, avasimibe, implitapide, and niacin are also being considered to manage hyperlipidemia. As lipids are very critical for cardiovascular diseases, these drugs reduce fatal and nonfatal cardiovascular abnormalities in the general population. However, a number of recent studies indicate that apart from their lipidlowering activities, statins and fibrates exhibit multiple functions to modulate intracellular signaling pathways, inhibit inflammation, suppress the production of reactive oxygen species, and modulate T cell activity. Therefore, nowadays, these drugs are being considered as possible therapeutics for several forms of human disorders including cancer, autoimmunity, inflammation, and neurodegeneration. Here I discuss these applications in the light of newly discovered modes of action of these drugs. Received 5 September 2005; received after revision 29 December 2005; accepted 26 January 2006     

一种新的RBF神经网络非线性动态系统建模方法

将遗传算法与正交优选法结合 ,用来训练径向基函数 ( RBF)神经网络 ,并对基函数宽度进行自动地调整 ,得到了一种训练 RBF神经网络的新方法 .将其应用于连续流体搅拌反应槽 ( CFSTR)生化反应器的建模中 ,得到了令人满意的结果 .该算法提高了径向基函数神经网络的泛化能力和鲁棒性 ,研究表明是一种有效的“黑箱”动态建模方法     

模糊回归参数估计方法及应用于景气对策信号分析

模糊回归采用样本模糊数 (Xi,Yi)来对模糊回归参数进行估计 ,其中 Yi为观测模糊数 .本文认为一般模糊参数 A的估计方式不能真实地表达出样本所蕴含的信息 ,因而另行设立了一套模糊参数估计方法 ,此法对样本的解释方式将更为合理 ,且估计的过程也比线性规划简便 .本文还举出了该方法应用于景气对策信号分析的一个实例.     

一类新的离散时延递归RBF神经网络

提出了一类带有离散时间 FIR/ IIR滤波器的递归 RBF神经网络 ,用离散时间 FIR/ IIR滤波器代替通常的 RBF神经网络中的线性输出权值 ,以适用于离散动力学系统的辨识和控制以及混沌时间序列预测 .本文给出的学习算法简单 ,可以避免传统的递归算法的不稳定性 .将该类神经网络用于动力学系统的建模 ,收到很好的效果 .     

开关电容网络灵敏度分析的传递函数法

给出了一种开关电容网络(SCN)灵敏度分析的新方法。只要能写出SCN的z域改进节点电压方程,并且该方程系数矩阵在将z为e