BP神经网络结合变量选择方法在牛奶蛋白质含量检测中的应用北大核心CSCD

牛奶中的蛋白质含量会影响牛奶的品质,利用高光谱图像的光谱特征信息研究对牛奶蛋白质含量预测的可行性。本文提出一种基于竞争性自适应重加权算法(competitive adaptive reweighted sampling, CARS)和连续投影算法(successive projections algorithm, SPA)结合多层前馈神经网络(back propagation, BP)的预测建模方法,实验以含有不同浓度蛋白质的牛奶为对象,利用可见光/近红外高光谱成像系统共采集到5种牛奶共计250组高光谱数据,通过实验对比选择采用标准化方法对获取到的吸收光谱预处理,然后采用CARS结合SPA筛选特征波长,得到18个特征波长,建立CARS-SPA-BP模型,经过试验,CARS-SPA-BP模型的训练集决定系数和测试集决定系数R;和R;分别达到0.971和0.968,训练集均方根误差(root mean square error of calibration,RMSEC)和测试集均方根误差(root mean square error of prediction,RMSEP)达到了0.033和0.034。研究发现,采用CARS结合SPA筛选的牛奶特征波长建立的多层前馈神经网络模型,其模型预测结果与全波长建模相比并没有明显降低,因此将CARS结合SPA用于波长筛选并且结合BP神经网络基本可以完成对牛奶蛋白质含量的预测。为验证CARS-SPA-BP模型的预测能力,在相同数据环境下,使用较为传统的偏最小二乘回归(partial least squares regression, PLSR)进行建模,实验结果表明,CARS-SPA-BP相较于PLSR,R;和RMSEP均有明显提升。研究表明,CARS-SPA-BP可充分利用牛奶光谱特征信息实现较高精度的牛奶蛋白质含量检测。     

A facile approach to fabricate and embed multifunctional nano ZnO into soap matrix and liquid cleansing products for enhanced antibacterial and photostability for health and hygiene applications

This research demonstrates, a facile approach to fabricate the nano ZnO system in an unique combination of surfactant-polyol-assembly (SPA) acting as a caging agent restricting the ZnO crystallite size in nano-regime. This SPA is suitable for health and hygiene products and such optimized technique is among the very few researches exploring the impact of embedding low concentrations of nano ZnO system into the matrix of sodium salt of long chain fatty acids (soap bar) and liquid cleansing personal care products. The fabricated nano ZnO in SPA and infused products were systematically characterized using various advanced and appropriate techniques. The hexagonal wurtzite structure of nano ZnO-SPA is evaluated based on XRD pattern which also exhibit an average crystallite size as 20.18 nm and high specific surface area as 52.99 m²/g. The SEM-supported morphological assessment confirms the formation of agglomerates of ultrafine ZnO rods and spherical particles. Novel nano ZnO having wideband gap energy (3.66 eV) embedded in soap bar act as a UV-blocker preventing the oxidation of unsaturated long chain fatty acids. Soap bar without ZnO experienced degradation and reduction in whiteness to 17.85% whereas 2.5 mg/g nano ZnO infused soap shows the reduction to 7.9% which clearly reflects the increased photostability of soap bar. The antibacterial efficacy of nano ZnO-SPA and infused products are investigated against Staphylococcus aureus (S. aureus) and Escherichia coli (E. coli) by Zone of Inhibition (ZOI) and European standard EN:1276. Infused products exhibited high antibacterial efficacy up to 4.43 log reduction equivalent to >99.99% germ kill.     

复杂系统的集对模式识别方法

人类社会和科学技术的不断发展必然导致复杂性的提高,如今复杂大系统的发展已经向传统的模式识别方法提出了挑战。本着对复杂大系统呼唤新的模式识别方法的响应,探讨了基于统一集和集对分析的新的模式识别方法,提出了利用该方法进行模式识别的判定原则。最后文章还举例证明了该方法的有效性。     

基于SPA联系数的同异反定量推理路径选择

基于集对分析（ＳＡＰ）的同异返定量推理已在学生成绩推算中得到应用。文章侧重研讨基于ＳＰＡ联系同异反定量推理路径的选择,给出其一般规则,并举例说明其应用。     

无模型技术在有源噪声控制中的应用

根据无模型自适应控制的基本原理,采用同时扰动算法对有源噪声控制系统中的测量误差信号进行直接处理,不再需要建立次级声学路径模型。分别就前馈控制和反馈控制两种情况加以研究。仿真结果表明,该算法不仅使主噪声信号得到明显的衰减,而且对系统次路径的变化也不敏感。     

一种适合于移动Ad Hoc网络的路由协议WPRP

针对无线Ad Hoc网络特点,提出了一种无线路径发现路由协议(Wireless Path Finding Routing Protocol,WPRP)。对WPRP协议进行了简要概述,分别介绍了采用最短路径算法(Short Path Algorithm,SPA)查找最短路径;递归环路检测算法(Recursive Loop Check Algorithm,RLCA)消除路由环路;路由校验实现节点间路由同步;自适应时间调节机制进一步减少网络冲突、降低网络开销等方法以及实现过程。OPNET仿真结果表明,与OLSR协议相比,WPRP具有较小的路由开销和较快的收敛速度,是一种动态路由协议,适用于无线Ad Hoc网络。     

Enhanced optical immunosensor based on surface plasmon resonance for determination of transferrin

Wavelength modulation surface plasmon resonance biosensors (SPR) using colloidal Au nanoparticles and double-linker sensing membrane enhancement are reported for determination of transferrin. The 2-mercaptoethylamine (MEA) was immobilized on the biosensor surface with traditional amine coupling method. The interaction between colloidal Au nanoparticles and MEA was investigated. The anti-transferrin was immobilized on the biosensor surface prepared with staphylococcal protein A (SPA). The interaction of the antibody and antigen was monitored in real time. The good response was obtained in the concentration range 1-20, 0.1-20 and 0.05-20 μg/mL for directly immune assay, double-linker assay and colloidal Au-amplified assay. The result clearly demonstrates that these methods may obtain significantly enhancement of sensitivity for the wavelength modulation SPR biosensor.     

LDPC码译码算法及性能分析

为了进一步降低低密度奇偶校验(LDPC)码译码算法的复杂度,基于经典置信传播(BP)译码算法,给出了对数域迭代后验概率对数似然比(APP LLR)算法。通过概率域的和积算法(SPA)和对数域的迭代APP LLR算法的性能仿真及分析可见,迭代APP LLR算法能以较小的性能损失换取复杂度的大幅降低。进一步选用迭代APP LLR算法,结合不同地形条件下的VHF频段信道模型,仿真了LDPC码编译码系统的性能。理论分析及仿真结果均表明,基于迭代APP LLR算法的LDPC码,实现简单,性能优异,具有良好的工程应用前景。     

智能卡SPA＆DPA攻击

智能卡是一种可防止被入侵设计的设备，为了防止攻击者藉由边际信道泄漏信息，如功率消耗、执行时间、故障时的输出与输入行为、电磁辐射、功率尖峰情形等信息攻击智能卡，必须采取一些防备措施。本文主要介绍攻击智能卡的简易功率分析（SPA），微分功率分析（DPA）。这些技术已被广泛地使用于窃取智能卡保护数据的技术上，提供这些技术的概略性观念，使我们知道问题所在，进一步想出对策，促使我们设计出更安全的智能卡。