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。     

采用射频直采技术的超宽带软件无线电平台

针对现代无线电系统对超宽带、可重构、多功能等需求日益增长的问题,基于射频直接采样技术,构建了一套符合PXI(PCI extensions for Instrumentation)标准的通用化超宽带软件无线电平台,硬件上以数模转换芯片AD9173和模数转换芯片ADC083000为核心,直接采样率高达3 Gsample/s;采用20 nm工艺制程的UltraScale Kintex FPGA,以IP(Intellectual Property)化开发理念设计FPGA固件,实现用户可编程。该平台能够满足0.1~1.4 GHz超宽带软件无线电应用需求,瞬时带宽高达1.3 GHz。     

某通信设备激励模块自检偶发报故的分析与解决措施

简单介绍了某通信设备激励模块工作原理,针对某批次激励模块自检偶发报故的问题,通过对故障原因的详细分析,发现采样时间与指示信号有效时间临界的问题。提出了在充分考虑指示信号有效时间范围的前提下,找到最佳采样时间段,对软件采样时间的设计进行优化的解决措施。     

基于历史梯度平均方差缩减的协同参数更新方法

随机梯度下降算法(SGD)随机使用一个样本估计梯度,造成较大的方差,使机器学习模型收敛减慢且训练不稳定。该文提出一种基于方差缩减的分布式SGD,命名为DisSAGD。该方法采用历史梯度平均方差缩减来更新机器学习模型中的参数,不需要完全梯度计算或额外存储,而是通过使用异步通信协议来共享跨节点的参数。为了解决全局参数分发存在的“更新滞后”问题,该文采用具有加速因子的学习速率和自适应采样策略：一方面当参数偏离最优值时,增大加速因子,加快收敛速度;另一方面,当一个工作节点比其他工作节点快时,为下一次迭代采样更多样本,使工作节点有更多时间来计算局部梯度。实验表明：DisSAGD显著减少了循环迭代的等待时间,加速了算法的收敛,其收敛速度比对照方法更快,在分布式集群中可以获得近似线性的加速。     

Electron microprobe analysis of micro amount powder samples: A case study on high purity coal

For the major, minor and trace element analysis of the inorganic compounds of a Ruhr-Saar coal different preparation techniques are investigated with X-ray fluorescence analysis, electron microprobe and classical wet chemical methods minimizing sample weight at standard preparation times and analytical accuracies. Considering accuracy as well as preparation efforts, determinations by electron microprobe on small sample amounts (<50 mg) proved to be superior to the other methods.Abbreviations AAS atomic absorbtion spectrometry - EMA electron microprobe analysis - IC ion chromatography - ISE ion sensitive electrode - PHOT photometry - WCA wet chemical analysis - XRF X-ray fluorescence analysis - LTA low temperature plasma ashing - HTA high temperature ashing - l.o.i. loss on ignition     

Coupling of thermal desorption in modified closeable sampling columns with wide-bore capillary gas chromatography and mass spectrometric detection

In contrast to usability of Curie-point pyrolysis at 700°C directly attached to gas chromatography-mass spectrometry (GC-MS) for determination of organic wood preservatives in waste wood samples the investigation method reported here consists of thermal desorption at temperatures about 260°C in connection with GC-MS for peak identification or GC with flame ionization detection for quantitative analyses. So-called “modified closeable sampling columns” are used as batch-reactor in thermal desorption experiments. Desorbed vapours can be introduced on capillary columns without sample discrimination and without a disturbing lost of resolution. In this manner a lot of individual polycyclic aromatic hydrocarbons were determinated in waste wood samples, especially in railway sleepers.     

New In-situ Sampling and Analysis of the Production of CeO<Subscript>2</Subscript> Powders from Liquid Precursors using a Novel Wet Collection System in a rf Inductively Coupled Thermal Plasma Reactor. Part 1: Reactor System and Sampling Probe

A new reactor and a novel in-situ sampling technique were developed for the study of the synthesis of CeO₂ powders produced from dissolved cerium nitrate salts. The conical reactor minimized particle recirculation and provided a highly symmetrical and undisturbed plasma flow suitable for the analysis of the phenomena affecting the formation of CeO₂ powders. Both a calorimetric study of the reactor and a thermodynamic analysis of CeO₂ formation were conducted. The sampling probe is described and near-isokinetic sampling was achieved. The sampled particles were collected using a miniature wet collection system, i.e. a mist atomizer and a custom-made spray chamber. A numerical simulation of the velocity and temperature fields of the plasma gas in the reactor was done using Fluent. A comprehensive droplet-to-particle formation mechanism presented elsewhere is revisited and expanded based on calorimetry, thermodynamics of CeO₂ formation, numerical simulations and collected particles. No traces of other oxidation states other than CeO₂ were found.     

用环形扩散管和滤纸联用采样技术采集空气中氨和铵盐

建立了环形扩散管和滤纸采样夹联用，分形态同时采集空气中所态氨和颗粒态铵盐的方法。在同一气流中，采用涂渍１．５％草酸乙醇水溶液的环形扩散管采集气态氨，用浸渍上述试剂的玻璃纤维纸和慢速定量滤纸分别采集颗粒安和第一层滤纸上的铵盐挥发产生的氨气。用靛酚蓝比色法分别测定氨和铵盐。当采气流速为１．０Ｌ／ｍｉｎ时，采样效率高于９８．２％。将本法测得的氨气和铵盐的总量与标准采样方法的测定结果比较。无显著性差异（Ｐ     

Thermodynamical approach to the traveling salesman problem: An efficient simulation algorithm

We present a Monte Carlo algorithm to find approximate solutions of the traveling salesman problem. The algorithm generates randomly the permutations of the stations of the traveling salesman trip, with probability depending on the length of the corresponding route. Reasoning by analogy with statistical thermodynamics, we use the probability given by the Boltzmann-Gibbs distribution. Surprisingly enough, using this simple algorithm, one can get very close to the optimal solution of the problem or even find the true optimum. We demonstrate this on several examples.We conjecture that the analogy with thermodynamics can offer a new insight into optimization problems and can suggest efficient algorithms for solving them.The author acknowledges stimulating discussions with J. Piút concerning the main ideas of the present paper. The author is also indebted to P. Brunovský, J. erný, M. Hamala, . Peko, . Znám, and R. Zajac for useful comments.     

Stability of workroom air volatile organic compounds on solid adsorbents for thermal desorption gas chromatography

The storage stability of the occupationally frequently occurring compounds, methylethylketone, methylisobutylketone, benzene, toluene, tetrachloroethylene, n-butylacetate, -pinene, β-pinene, limonene and n-decane, has been investigated on the adsorbents Tenax TA, Chromosorb 106 and Carbotrap using thermally desorbable tube type samplers, commonly utilized in ambient and workroom atmospheric measurements. Fifty and 500 ng of each compound were loaded on the various adsorbents tubes, stored at both ambient (20 °C) and refrigerated (4 °C) temperatures and analysed by means of thermal gas chromatography with mass spectrometric detection on days 0, 7, 14 and 28 after exposure. A 90% storage recovery was chosen as acceptance criteria for storage stability, and statistical testing by Student's t-test, analysis of variance and Bonferroni post hoc tests were employed to investigate the effect of the categorical variables storage time, storage temperature and analyte loading on the different adsorbents. Chromosorb 106 showed the overall best behaviour with recoveries of 90% or better for all analytes during the 28-day test period. Tenax TA and Carbotrap yielded lower recoveries and were more influenced by variations in storage time, storage temperature and analyte loading. Refrigerated temperatures were best avoided for storage on Tenax TA, but may increase the recovery of some compounds on Carbotrap (e.g. n-butylacetate). The blank build-up on the adsorbents was also investigated, and Carbotrap and Tenax TA showed no signs of artefact development over time. Chromosorb 106, however, contained inherently more artefacts that build up over time, which in spite of the excellent storage capability, may limit its use in field studies where long storage times are normal.