Adaptive configuring of radial basis function network by hybrid particle swarm algorithm for QSAR studies of organic compounds

The configuring of a radial basis function network (RBFN) consists of selecting the network parameters (centers and widths in RBF units and weights between the hidden and output layers) and network architecture. The issues of suboptimum and overfitting, however, often occur in RBFN configuring. This paper presented a hybrid particle swarm optimization (HPSO) algorithm to simultaneously search the optimal network structure and parameters involved in the RBFN (HPSORBFN) with an ellipsoidal Gaussian function as a basis function. The continuous version of PSO was used for parameter training, while the modified discrete PSO was employed to determine the appropriate network topology. The proposed HPSORBFN algorithm was applied to modeling the inhibitory activities of substituted bis[(acridine-4-carboxamide)propyl]methylamines to murine P388 leukemia cells and the bioactivities of COX-2 inhibitors. The results were compared with those obtained from RBFNs with the parameters optimized by continuous PSO and by conventionally RBFN training the algorithm for a fixed network topology, indicating that the HPSO was competent for RBFN configuring in that it converged quickly toward the optimal solution and avoided overfitting.     

Determination of Operating Conditions of Ethylene/1-Octene Copolymerization Using Artificial Neural Network (ANN)

An artificial neural network (ANN) is applied to determine appropriate parameters in copolymerization of ethylene and 1-octene via metallocene catalytic system for producing a copolymer with desired chain microstructures. The polymerization parameters of interests are polymerization temperature, ethylene pressure, and the amount of hydrogen used. The ANN used is a feed-forward network with a back propagation learning method and has a 5-6-6-3 architecture. When comparing with both training and testing experimental data sets, it was found that ANN can provide a good guesstimation of polymerization parameters.     

Application of an artificial neural network (ANN) and piezoelectric chemical sensor array for identification of volatile organic compounds

A piezoelectric chemical sensor array was developed using four quartz crystals. Gas chromatographic stationary phases were used as sensing materials and the array was connected to an artificial neural network (ANN). The application of the ANN method proved to be particularly advantageous if the measured property (mass, concentration, etc.) should not be connected exactly to the signal of the transducers of the piezoelectric sensor. The optimum structure of neural network was determined by a trial and error method. Different structures were tried with several neurons in the hidden layer and the total error was calculated. The optimum values of primary weight factors, learning rate (η=0.15), momentum term (μ=0.9), and the sigmoid parameter (β=1) were determined. Finally, three hidden neurons and 900 training cycles were applied. After the teaching process the network was used for identification of taught analytes (acetone, benzene, chloroform, pentane). Mixtures of organic compounds were also analysed and the ANN method proved to be a reliable way of differentiating the sensing materials and identifying the volatile compounds.     

Application of an artificial neural network (ANN) and piezoelectric chemical sensor array for identification of volatile organic compounds

A piezoelectric chemical sensor array was developed using four quartz crystals. Gas chromatographic stationary phases were used as sensing materials and the array was connected to an artificial neural network (ANN). The application of the ANN method proved to be particularly advantageous if the measured property (mass, concentration, etc.) should not be connected exactly to the signal of the transducers of the piezoelectric sensor. The optimum structure of neural network was determined by a trial and error method. Different structures were tried with several neurons in the hidden layer and the total error was calculated. The optimum values of primary weight factors, learning rate (η=0.15), momentum term (μ=0.9), and the sigmoid parameter (β=1) were determined. Finally, three hidden neurons and 900 training cycles were applied. After the teaching process the network was used for identification of taught analytes (acetone, benzene, chloroform, pentane). Mixtures of organic compounds were also analysed and the ANN method proved to be a reliable way of differentiating the sensing materials and identifying the volatile compounds.     

Application of artificial neural networks for formulation and modeling of dye adsorption onto multiwalled carbon nanotubes

In this study, an artificial neural network (ANN) has been developed to predict the adsorption amount of dye (methylene blue) onto multiwalled carbon nanotubes. Batch experiments have been carried out to obtain experimental data. Important parameters in the adsorption system such as initial dye concentration, adsorbent dosage, temperature, pH and contact time have been used as the inputs of the network, while the output is the final concentration of dye in aqueous solution after adsorption. The neural network structure has been optimized by testing various training algorithms and different number of neurons in a hidden layer. An empirical equation for determination of final dye concentration in aqueous solutions after adsorption has been developed by using the weights of the optimized network. The results of the optimized ANN have been compared with conventional models in equilibrium and kinetic fields. According to error analysis and determination coefficient, the ANN was found to be the most appropriate model to describe this adsorption process. Sensitivity analysis showed that initial dye concentration, pH and contact time are the most effective parameters in this process. The influence percentages of these parameters on the output were 28, 24 and 24 %, respectively.     

Neural networks applied in chemistry. I. Determination of the optimal topology of multilayer perceptron neural networks

Artificial neural networks (ANNs) are comparatively straightforward to understand and use in the analysis of scientific data. However, this relative transparency may encourage their use in an uncritical, and therefore possibly unproductive, fashion. The geometry of a network is among the most crucial factors in the successful deployment of network tools; in this review, we cover methods that can be used to determine optimum or near‐optimum geometries. These methods of determining neural network architecture include the following: (i) trial and error, in which architectures chosen semirandomly are tested and modified by the user; (ii) empirical or statistical methods, in which an ANN's internal parameters are adjusted based on the model's performance; (iii) hybrid methods, such as fuzzy inference; (iv) constructive and/or pruning algorithms, that add and/or remove neurons or weights from an initial architecture, respectively, based on a predefined link between architecture and ANN performance; (v) evolutionary strategies, which search the topology space using genetic operators to vary the neural network parameters. Several case studies illustrate the development of neural network models for applications in chemistry and chemical engineering. Copyright © 2011 John Wiley & Sons, Ltd.     

Monitoring of fosinopril sodium impurities by liquid chromatography-mass spectrometry including the neural networks in method evaluation

In this paper, the mass spectrometry (MS) detection has been applied for screening of fosinopril sodium impurities which arise during forced stress study. Before MS analysis, liquid chromatographic method with suitable mobile phase composition was developed. The separation was done on SunFire 100 mm x 4.6 mm 3.5 microm particle size column. The mobile phases which consisted of methanol-ammonium acetate buffer-acetic acid, in different ratios, were used in a preliminary study. Flow rate was 0.3 mL min(-1). Under these conditions, percent of methanol, concentration of ammonium acetate buffer and acetic acid content were tested simultaneously applying central composite design (CCD) and artificial neural network (ANN). The combinations of experimental design (ED) and ANN present powerful technique in method optimization. Input and output variables from CCD were used for network training, verification and testing. Multiple layer perceptron (MLP) with back propagation (BP) algorithm was chosen for network training. When the optimal neural topology was selected, network was trained by adjusting strength of connections between neurons in order to adapt the outputs of whole network to be closer to the desired outputs, or to minimize the sum of the squared errors. From the method optimization the following mobile phase composition was selected as appropriate: methanol-10 mM ammonium acetate buffer-acidic acid (80:19.5:0.5 v/v/v). This mobile phase was used as inlet for MS. According to molecular structure and literature data, electrospray positive ion mode was applied for analysis of fosinopril sodium and its impurities. The proposed method could be used for screening of fosinopril sodium impurities in bulk and pharmaceuticals, as well as for tracking the degradation under stress conditions.     

Prediction of CO2 Solubility in Polymers by Radial Basis Function Artificial Neural Network Based on Chaotic Self-adaptive Particle Swarm Optimization and Fuzzy Clustering Method

To replace costly and time-consuming experimentation in laboratory, a novel solubility prediction model based on chaos theory, self-adaptive particle swarm optimization （PSO）, fuzzy c-means clustering method, and radial ba- sis function artificial neural network （RBF ANN） is proposed to predict CO2 solubility in polymers, hereafter called CSPSO-FC RBF ANN. The premature convergence problem is overcome by modifying the conventional PSO using chaos theory and self-adaptive inertia weight factor. Fuzzy c-means clustering method is used to tune the hidden centers and radial basis function spreads. The modified PSO algorithm is employed to optimize the RBF ANN connection weights. Then, the proposed CSPSO-FC RBF ANN is used to investigate solubility of CO2 in polystyrene （PS）, polypropylene （PP）, poly（butylene succinate） （PBS） and poly（butylene succinate-co-adipate） （PBSA）, respec- tively. Results indicate that CSPSO-FC RBF ANN is an effective method for gas solubility in polymers. In addition, compared with conventional RBF ANN and PSO ANN, CSPSO-FC RBF ANN shows better performance. The values of average relative deviation （ARD）, squared correlation coefficient （R2） and standard deviation （SD） are 0.1071, 0.9973 and 0.0108, respectively. Statistical data demonstrate that CSPSO-FC RBF ANN has excellent prediction capability and high-accuracy, and the correlation between prediction values and experimental data is good.     

Nondestructive Quantitative Analysis of Analgini Powder Pharmaceutical by Near-Infrared Spectroscopy and Artifical Neural Network Technique

ABSTRACT

This paper demonstrates the usefulness of near-infrared (NIR) spectra and artificial neural network (ANN) in nondestructive quantitative analysis of pharmaceuticals. Real data sets from near-infrared reflectance spectra of analgini powder pharmaceutical were used to build up an artificial neural network to predict unknown samples. The parameters affecting the network were discussed. A new network evaluation criterion, the degree of approximation, was employed. The overfitting was discussed. Owing to the good nonlinear multivariate calibration nature of ANN, the predicted result was reliable and precise. The relative error of unknown samples was less than 2.5%     

有机磷农药构效关系的主成分分析-人工神经网络研究

采用主成分分析法对样本数据集进行预处理，将得到的新的样本数据集输入人工神经网络，对有机磷农药的毒性参数进行预报。研究结果表明，主成分分析－人工神经网络的预报精度优于单纯的人工神经网络。     

Selection of quasi-optimal inputs in chemometrics modeling by artificial neural network analysis

Instrumentation spectra used for chemometrics analysis are often too unwieldy to model, as many of the inputs do not contain important information. Several mathematical methods are used for reducing the number of inputs to the significant ones only. Artificial neural networks (ANN) modeling suffers from difficulties in training models with a large number of inputs. However, using a non-random initial connection weight algorithm and local minima avoidance and escape techniques can overcome these difficulties. Once the ANN model is trained, the analysis of its connection weights can easily identify the more relevant inputs. Repeating the process of training the ANN model with the reduced input set and the selection of the more relevant inputs can proceed until a quasi-optimal, small, set of inputs is identified. Two examples are presented—finding the minimal set of wavelengths in benchmark diesel fuel NIR spectra, and in spectra generated in a recent work, modeling of “artificial nose” sensor array. In the last example, 1260 inputs were reduced to optimal sets of <10 inputs. Causal index calculation can analyze the influence of each of selected wavelengths on the predicted property. Some of the resulting minimal sets are not unique, depending on the ANN architecture used in the training. The accuracy of the resulting ANN models is usually better, and more robust, than the original large ANN model.     

中药复杂组效关系的变结构神经网络辨识方法

针对中药复杂组效关系的辨识问题,研究了变结构多层前馈神经网络,推导出一种新型的变结构网络学习算法,成功地应用于中药川芎药效活性预测计算.该方法从一个规模较小的网络出发,当网络无法达到预定的学习精度时,自动增加隐含层神经元个数,并在原有学习结果的基础上确定新的网络参数,自适应地确定前馈神经网络结构,可用于处理复杂化学模式信息.计算机仿真实验结果表明,该方法能有效地确定多层前馈神经网络的最佳结构,提高网络学习效率和函数逼近精度,解决复杂非线性函数映射关系准确建模问题.     

Estimation of Average Comonomer Content of Ethylene/1-Olefin Copolymers Using Crystallization Analysis Fractionation (Crystaf) and Artificial Neural Network (ANN)

Summary: An artificial neural network (ANN) with a 4-3-3-1 architecture was developed to estimate average comonomer content of ethylene/1-olefin copolymers from crystallization analysis fractionation (Crystaf) results. The ANN was trained with a back propagation algorithm. It was found that average comonomer contents predicted from ANN agree well with experimental results for both training and testing data sets. The developed ANN was also used to systematically investigate the effects of chain microstructures and Crystaf operating conditions on Crystaf calibration curves.     

Prediction of retention times for anions in linear gradient elution ion chromatography with hydroxide eluents using artificial neural networks

The feasibility of using an artificial neural network (ANN) to predict the retention times of anions when eluted from a Dionex AS11 column with linear hydroxide gradients of varying slope was investigated. The purpose of this study was to determine whether an ANN could be used as the basis of a computer-assisted optimisation method for the selection of optimal gradient conditions for anion separations. Using an ANN with a (1, 10, 19) architecture and a training set comprising retention data obtained with three gradient slopes (1.67, 2.50 and 4.00 mM/min) between starting and finishing conditions of 0.5 and 40.0 mM hydroxide, respectively, retention times for 19 analyte anions were predicted for four different gradient slopes. Predicted and experimental retention times for 133 data points agreed to within 0.08 min and percentage normalised differences between the predicted and experimental data averaged 0.29% with a standard deviation of 0.29%. ANNs appear to be a rapid and accurate method for predicting retention times in ion chromatography using linear hydroxide gradients.     

Co‐evolution of non‐linear PLS model components

The issue of outer model weight updating is important in extending partial least squares (PLS) regression to modelling data that shows significant non‐linearity. This paper presents a novel co‐evolutionary component approach to the weight updating problem. Specification of the non‐linear PLS model is achieved using an evolutionary computational (EC) method that can co‐evolve all non‐linear inner models and all input projection weights simultaneously. In this method, modular symbolic non‐linear equations are used to represent the inner models and binary sequences are used to represent the projection weights. The approach is flexible, and other representations could be employed within the same co‐evolutionary framework. The potential of these methods is illustrated using a simulated pH neutralisation process data set exhibiting significant non‐linearity. It is demonstrated that the co‐evolutionary component architecture can produce results which are competitive with non‐linear neural network‐based PLS algorithms that use iterative projection weight updating. In addition, a data sampling method for mitigating overfitting to the training data is described. Copyright © 2007 John Wiley & Sons, Ltd.     

Application of Artificial Neural Networks in Multifactor Optimization of Selectivity in Capillary Electrophoresis

ABSTRACT

A methodology based on the coupling of experimental design and artificial neural networks (ANNs) was proposed in the optimization of selectivity in capillary electrophoresis. The effect of the buffer composition, concentration, SDS concentration, ethanol percentage and the applied voltage on the separation of six choice solutes was examined by using orthogonal design. Feedforward-type neural networks with faster back propagation (BP) algorithm were applied to model the separation process, and then optimization of the experimental conditions was carried out in the modeled neural network with 5-7-1 structure, which had been confirmed to be able to provide the maximum performance. It was demonstrated that by combining ANN modeling with experimental design, the number of experiments necessary to search and find optimal separation conditions can be reduced significantly. Because of its general validity, the new proposed approach can also be applied in other separation conditions.