Polymeric membrane ion-selective and cross-sensitive electrode-based electronic tongue for qualitative analysis of beverages

An electronic tongue based on the sensor array of polymeric membrane ion-selective electrodes combined with pattern recognition tools was applied to qualitative analysis of various brands of orange juice, tonic, and milk. The capability of this device to reliably discriminate between different brands of those products was presented. The tests of the system were performed using products of the same brand, but with different manufacture dates (and thus comparable by the term of taste). The fusion of two types of sensors-classical selective ones and partially selective in one versatile array, and working out the sensor array's response by means of principal component analysis and back propagation neural network methods allowed the discrimination between different brands of various beverages with very high accuracy (90-100%). The real performance of the electronic tongue was evaluated applying testing samples from another manufacture lot, than the samples used in the learning set.     

Development of an inorganic cations retention model in ion chromatography by means of artificial neural networks with different two-phase training algorithms

This paper describes development of artificial neural network (ANN) retention model, which can be used for method development in variety of ion chromatographic applications. By using developed retention model it is possible both to improve performance characteristic of developed method and to speed up new method development by reducing unnecessary experimentation. Multilayered feed forward neural network has been used to model retention behaviour of void peak, lithium, sodium, ammonium, potassium, magnesium, calcium, strontium and barium in relation with the eluent flow rate and concentration of methasulphonic acid (MSA) in eluent. The probability of finding the global minimum and fast convergence at the same time were enhanced by applying a two-phase training procedure. The developed two-phase training procedure consists of both first and second order training. Several training algorithms were applied and compared, namely: back propagation (BP), delta-bar-delta, quick propagation, conjugate gradient, quasi Newton and Levenberg-Marquardt. It is shown that the optimized two-phase training procedure enables fast convergence and avoids problems arisen from the fact that every new weight initialization can be regarded as a new starting position and yield irreproducible neural network if only second order training is applied. Activation function, number of hidden layer neurons and number of experimental data points used for training set were optimized in order to insure good predictive ability with respect to speeding up retention modelling procedure by reducing unnecessary experimental work. The predictive ability of optimized neural networks retention model was tested by using several statistical tests. This study shows that developed artificial neural network are very accurate and fast retention modelling tool applied to model varied inherent non-linear relationship of retention behaviour with respect to mobile phase parameters.     

Optimization of artificial neural networks used for retention modelling in ion chromatography

The aim of this work is the development of an artificial neural network model, which can be generalized and used in a variety of applications for retention modelling in ion chromatography. Influences of eluent flow-rate and concentration of eluent anion (OH-) on separation of seven inorganic anions (fluoride, chloride, nitrite, sulfate, bromide, nitrate, and phosphate) were investigated. Parallel prediction of retention times of seven inorganic anions by using one artificial neural network was applied. MATLAB Neural Networks ToolBox was not adequate for application to retention modelling in this particular case. Therefore the authors adopted it for retention modelling by programming in MATLAB metalanguage. The following routines were written; the division of experimental data set on training and test set; selection of data for training and test set; Dixon's outlier test; retraining procedure routine; calculations of relative error. A three-layer feed forward neural network trained with a Levenberg-Marquardt batch error back propagation algorithm has been used to model ion chromatographic retention mechanisms. The advantage of applied batch training methodology is the significant increase in speed of calculation of algorithms in comparison with delta rule training methodology. The technique of experimental data selection for training set was used allowing improvement of artificial neural network prediction power. Experimental design space was divided into 8-32 subspaces depending on number of experimental data points used for training set. The number of hidden layer nodes, the number of iteration steps and the number of experimental data points used for training set were optimized. This study presents the very fast (300 iteration steps) and very accurate (relative error of 0.88%) retention model, obtained by using a small amount of experimental data (16 experimental data points in training set). This indicates that the method of choice for retention modelling in ion chromatography is the artificial neural network.     

Neural networks convergence using physicochemical data

An investigation of the neural network convergence and prediction based on three optimization algorithms, namely, Levenberg-Marquardt, conjugate gradient, and delta rule, is described. Several simulated neural networks built using the above three algorithms indicated that the Levenberg-Marquardt optimizer implemented as a back-propagation neural network converged faster than the other two algorithms and provides in most of the cases better prediction. These conclusions are based on eight physicochemical data sets, each with a significant number of compounds comparable to that usually used in the QSAR/QSPR modeling. The superiority of the Levenberg-Marquardt algorithm is revealed in terms of functional dependence of the change of the neural network weights with respect to the gradient of the error propagation as well as distribution of the weight values. The prediction of the models is assessed by the error of the validation sets not used in the training process.     

毛细管电泳-小波神经网络法辅佐诊断乳腺癌的研究

采用毛细管电泳法测定了46个健康人和26个乳腺癌病人尿样中的13种正常核苷和修饰核苷,以小波神经网络作为模式识别工具对健康人和乳腺癌病人的分类作了研究,随机选取的训练集的识别率达到100％,相应的预测集判别率正确性在96％以上,与经典的前向多层神经网络相比,小波神经网络具有更强的信息提取和逼近能力.研究结果还表明,小波神经网络的预测能力强于主成分分析和线性判别分析,毛细管电泳法与小波神经网络的结合有望成为乳腺癌的辅助诊断手段.     

弹性反传神经网络法预测烷基苯的疏水性常数

 介绍了应用人工神经网络预测烷基苯分子疏水性常数的方法。该法同传统方法相比 ,具有操作简便 ,适用范围广的特点。基于误差反传神经网络 ,建立了分子连接性指数 (χ)、范德华表面积 (Aw)和疏水性常数 (logP)之间的数学模型。应用该模型对烷基苯分子的疏水性常数进行预测 ,其平均相对偏差为 0 6 7%。并且通过与标准误差反传算法和自适应学习算法相比较 ,发现弹性反传算法具有训练速度快 ,参数选择简单的特点。     

CmI奇宇称光谱能级的模式识别研究

应用新的模式识别方法ＰＣＡ－ＢＰＮ（ＰｒｉｎｃｉｐａｌＣｏｍｐｏｎｅｎｔＡｎａｌｙｓｉｓ－ＢａｃｋＰｒｏｐａｇａｔｉｏｎＮｅｔｗｏｒｋ）指认ＣｍＩ奇宇称未知能级，支持了前人应用传统的ＫＮＮ（ＫＮｅａｒｅｓｔＮｅｉｇｈｂｏｒｓ）等模式识别方法及对传神经网络方法（ＣｏｕｎｔｅｒＰｒｏｐａｇａｔｉｏｎｎｅｔｗｏｒｋ，ＣＰＮ）对大部分谱线的指认，进一步确认了这些组态的归属，鉴别了ＫＮＮ等与ＣＰＮ不同的预报     

Non‐targeted volatile profiles for the classification of the botanical origin of Chinese honey by solid‐phase microextraction and gas chromatography–mass spectrometry combined with chemometrics

A potential method for the discrimination and prediction of honey samples of various botanical origins was developed based on the non‐targeted volatile profiles obtained by solid‐phase microextraction with gas chromatography and mass spectrometry combined with chemometrics. The blind analysis of non‐targeted volatile profiles was carried out using solid‐phase microextraction with gas chromatography and mass spectrometry for 87 authentic honey samples from four botanical origins (acacia, linden, vitex, and rape). The number of variables was reduced from 2734 to 70 by using a series of filters. Based on the optimized 70 variables, 79.12% of the variance was explained by the first four principal components. Partial least squares discriminant analysis, naïve Bayes analysis, and back‐propagation artificial neural network were used to develop the classification and prediction models. The 100% accuracy revealed a perfect classification of the botanical origins. In addition, the reliability and practicability of the models were validated by an independent set of additional 20 authentic honey samples. All 20 samples were accurately classified. The confidence measures indicated that the performance of the naïve Bayes model was better than the other two models. Finally, the characteristic volatile compounds of linden honey were tentatively identified. The proposed method is reliable and accurate for the classification of honey of various botanical origins.     

A voltammetric electronic tongue made of modified epoxy-graphite electrodes for the qualitative analysis of wine

First results are presented of a new voltammetric electronic tongue which employs modified epoxy-graphite electrodes. This analytical tool has been applied to qualitative wine analysis, performing the classification of wine varieties, as well as recognition of the oxygenation effect. In the same way, studies related to the detection of some defects in wine production were also assessed, such as its vinegary taste in open-air contact or the use of excess sulphite preservative. The electronic tongue was formed by five voltammetric electrodes, four of them being bulk-modified with different substances: copper and platinum nanoparticles on one side, and polyaniline and polypyrrole powder on the other. The responses were preprocessed employing Principal Component Analysis (PCA) to visualize and identify distinct episodes. The resulting PCA scores were modelled with an artificial neural network that accomplishes final prediction with the qualitative classification of wines and/or detection of defects.     

神经网络法在使用裂解气相色谱鉴别中草药中的应用

将以误差反向传播为训练算法的前馈式人工神经网络（ＢＰ－ＡＮＮ）首次艇于中草药的裂解气相色谱谱图解析。重点考察了如何表征和提取复杂的裂解色谱图中有价值信息,用于主成分分析方法处理后输入到有数经优化的神经网络中。实验证明,该广阔示仅可以正确识别样品所属种类,耐用对一示同实验时间、数据残缺等原因造成的噪音具有优异的抗干扰能力。     

Classification of Cm I energy levels using PCA-BPN and PCA-NLM

Two pattern recognition (PR) techniques, principal component analysis-back propagation networks (PCA-BPN) and principal component analysis-nonlinear mapping (PCA-NLM), have been applied to the problem of classifying unknown energy levels of the first spectrum of curium (Cm I) according to their configurations. In comparison, with those reported by early PR techniques and counter propagation neural networks (CPN's), PCA-BPN has been demonstrated to possess much more prediction accuracy as to its performance on test sets. Obtained results further confirm the most previous assignments with these energy levels given by some early PR techniques and CPN. Moreover, the obtained results definitely reassign some energy levels' electronic configurations which were ambiguously conjectured in previous work.     

Comparison of multivariate preprocessing techniques as applied to electronic tongue based pattern classification for black tea

In an electronic tongue, preprocessing on raw data precedes pattern analysis and choice of the appropriate preprocessing technique is crucial for the performance of the pattern classifier. While attempting to classify different grades of black tea using a voltammetric electronic tongue, different preprocessing techniques have been explored and a comparison of their performances is presented in this paper. The preprocessing techniques are compared first by a quantitative measurement of separability followed by principle component analysis; and then two different supervised pattern recognition models based on neural networks are used to evaluate the performance of the preprocessing techniques.     

CmⅠ奇宇称光谱能级的模式识别研究

应用新的模式识别方法PCA－BPN（PrincipalComponentAnalysis－BackPropagationNetwork）指认CmⅠ奇宇称未知能级，支持了前人应用传统的KNN（KNearestNeighbors）等模式识别方法及对传神经网络方法（CounterPropagationNetwork，CPN）对大部分谱线的指认，进一步确认了这些组态的归属；鉴别了KNN等与CPN不同的预报结果，纠正CPN的某些错误分类，并以可视非线性映照分类器加以佐证     

神经网络方法在血管紧张素转换酶抑制剂定量构效关系建模中的应用

对20个ACEI化合物用量子化学方法进行结构优化并计算出10个参数,用9种不同隐含层节点数的BP神经网络研究了ACEI的定量构效关系,建立了节点为10/6/1的三层BP神经网络模型。结果表明:以量化理论计算所得参数可以构建合理的ACEI定量构效关系模型,神经网络模型M6的r2=0.995,S=0.050,6个验证集化合物的残差平方和为0.002,预测能力明显强于多元线形回归模型,亦优于同类文献报道,可作为ACEI研发领域中预测先导化合物活性的理论工具。     

饱和醇结构-保留定量相关的人工神经网络模型

以拓扑指数为结构描述符，用基于Levenberg-Marquardt优化的BP神经网络建立了醇类化合物的结构与色谱保留值的相关性模型，用于未知醇类化合物在SE－30和OV－3两根色谱柱上保留指数的同时预测，其学习速率优于文献中普通BP神经网络法，预测准确度与普通BP神经网络法接近，但优于多元线性回归法，因而是一种较好的预测有机化合物气相色谱保留指数的方法。     

Simultaneous determination of fermented milk aroma compounds by a potentiometric sensor array

The paper reports on the application of an electronic tongue for simultaneous determination of ethanol, acetaldehyde, diacetyl, lactic acid, acetic acid and citric acid content in probiotic fermented milk. The αAstree electronic tongue by Alpha M.O.S. was employed. The sensor array comprised of seven non-specific, cross-sensitive sensors developed especially for food analysis coupled with a reference Ag/AgCl electrode. Samples of plain, strawberry, apple-pear and forest-fruit flavored probiotic fermented milk were analyzed both by standard methods and by the potentiometric sensor array. The results obtained by these methods were used for the development of neural network models for rapid estimation of aroma compounds content in probiotic fermented milk.The highest correlation (0.967) and lowest standard deviation of error for the training (0.585), selection (0.503) and testing (0.571) subset was obtained for the estimation of ethanol content. The lowest correlation (0.669) was obtained for the estimation of acetaldehyde content. The model exhibited poor performance in average error and standard deviations of errors in all subsets which could be explained by low sensitivity of the sensor array to the compound. The obtained results indicate that the potentiometric electronic tongue coupled with artificial neural networks can be applied as a rapid method for the determination of aroma compounds in probiotic fermented milk.     

An Extreme Learning Machine Algorithm to Predict the In-flight Particle Characteristics of an Atmospheric Plasma Spray Process

A robust single hidden layer feed forward neural network (SLFN) is used in this study to model the in-flight particle characteristics of the atmospheric plasma spray (APS) process with regard to the input processing parameters. The in-flight particle characteristics influence the structure and properties of the APS coating and, thus, are considered important parameters to comprehend the manufacturing process. The training times of traditional back propagation algorithms, mostly used to model such processes, are far slower than desired for implementation of an on-line control system. Use of slow gradient based learning methods and iterative tuning of all network parameters during the learning process are the two major causes for such slower learning speed. An extreme learning machine (ELM) algorithm, which randomly selects the input weights and biases and analytically determines the output weights, is used in this work to train the SLFNs. Performance comparisons of the networks trained with ELM algorithm and standard error back propagation algorithms are presented. It is found that networks trained with ELM have good generalization performance, much shorter training times and stable performance with regard to the changes in number of hidden layer neurons. The trends represent robustness of the trained networks and enhance reliability of the application of the artificial neural network in modelling APS processes.