Artificial Neural Networks

Comparison of Phytochemical Profiles of Wild and Cultivated American Ginseng Using Metabolomics by Ultra-High Performance Liquid Chromatography-High-Resolution Mass Spectrometry

American ginseng (Panax quinquefolius L.) has been recognized as a valuable herb medicine, and ginsenosides are the most important components responsible for the health-beneficial effects. This study investigated the secondary metabolites responsible for the differentiation of wild and cultivated American ginsengs with ultrahigh-performance liquid chromatography-high resolution mass spectrometry (UHPLC-HRMS)-based metabolomic approach. An in-house ginsenoside library was developed to facilitate data processing and metabolite identification. Data visualization methods, such as heatmaps and volcano plots, were utilized to extract discriminated ion features. The results suggested that the ginsenoside profiles of wild and cultivated ginsengs were significantly different. The octillol (OT)-type ginsenosides were present in greater abundance and diversity in wild American ginsengs; however, a wider distribution of the protopanaxadiol (PPD)-and oleanolic acid (OA)-type ginsenosides were found in cultivated American ginseng. Based on the tentative identification and semi-quantification, the amounts of five ginsenosides (i.e., notoginsenoside H, glucoginsenoside Rf, notoginsenoside R1, pseudoginsenoside RT2, and ginsenoside Rc) were 2.3-54.5 fold greater in wild ginseng in comparison to those in their cultivated counterparts, and the content of six ginsenosides (chicusetsusaponin IVa, malonylginsenoside Rd, pseudoginsenoside Rc1, malonylfloralginsenoside Rd6, Ginsenoside Rd, and malonylginsenoside Rb1) was 2.6-14.4 fold greater in cultivated ginseng compared to wild ginseng. The results suggested that the in-house metabolite library can significantly reduce the complexity of the data processing for ginseng samples, and UHPLC-HRMS is effective and robust for identifying characteristic components (marker compounds) for distinguishing wild and cultivated American ginseng.

Forecasting Daily Electricity Consumption in Thailand Using Regression, Artificial Neural Network, Support Vector Machine, and Hybrid Models

This article involves forecasting daily electricity consumption in Thailand. Electricity consumption data are provided by the Electricity Generating Authority of Thailand, the leading power utility state enterprise under the Ministry of Energy. Five forecasting techniques, including multiple linear regression, artificial neural network (ANN), support vector machine, hybrid models, and ensemble models, are implemented. The article proposes a hyperparameter tuning technique, called sequential grid search, which is based on the widely used grid search, for ANN and hybrid models. Auxiliary variables and indicator variables that can improve the models’ forecasting performance are included. From the computational experiment, the hybrid model of a multiple regression model to forecast the expected daily consumption and ANNs from the sequential grid search to forecast the error term, along with additional indicator variables for some national holidays, provides the best mean absolution percentage error of 1.5664% on the test data set.

Comment on "Reversible disorder-order transitions in atomic crystal nucleation"

Jeon et al. (Reports, 29 January 2021, p. 498) reported reversible fluctuations between disordered and crystalline states during nucleation. We argue that the adoption of "magic numbers" for separation of the size range into three parts is misleading when using the sigmoid functions and that their figure 3B was wrongly illustrated. The crystal shape effect was neglected during theoretical analysis.

Application of multiway-variate calibration to simultaneous voltammetric determination of three catecholamines

Three structurally similar catecholamines, adrenaline (AA), dopamine (DA) and noradrenaline (NA), which are neurotransmitters well known for their treatment of neural disorders and many other diseases, were investigated at a glassy carbon electrode (GCE) with the use of differential pulse stripping voltammetry (DPSV). The aim was to develop a simple, rapid and cost effective method of simultaneous analysis of these three substances in typical samples such as human urine. Each analyte showed typical reversible redox behaviour with the use of cyclic voltammetry in aqueous medium in the pH range of 4.0-7.9, and composite voltammograms of the three analytes were found to consist of three significantly overlapping peaks corresponding to the three individual substances. Thus, two- and three-way multi-variate data analysis methods were investigated. These included unfolding methods such as-partial least squares (UPLS), principal component regression (UPCR) and radial basis function-artificial neural networks (URBF-ANN), as well as trilinear models, such as parallel factor analysis (PARAFAC) and N-way PLS (NPLS). The results were compared with those obtained from two-way voltammetric data matrix with the use of conventional models, PLS, PCR and RBF-ANN. It was found that most of the three-way models, such as PARAFAC and UPLS, performed somewhat better than others on the basis of the %RPET (5.6∼5.9) and mean %Recovery (94∼102). The proposed methods were then applied for the determination of human urine samples spiked with the three catecholamines, and the results were satisfactory.

Use of artificial neural networks in near-infrared reflectance spectroscopy calibrations for predicting the inclusion percentages of wheat and sunflower meal in compound feedingstuffs

The use of near-infrared reflectance spectroscopy (NIRS) calibrations to predict the ingredient composition in compound feeds (i.e., inclusion percentage of each ingredient) is a complex task, regarding both the nature of the parameters to be predicted, since they are not well-defined chemical entities, and the heterogeneousness of the matrices/formulas in which each ingredient participates. The present paper evaluates the use of nonlinear regression methods, such as artificial neural networks (ANN), for developing NIRS calibrations to predict these parameters. Two of the most representative ingredients in the Spanish compound feed formulations (wheat and sunflower meal) were selected for evaluating ANN possibilities, using a large spectral library comprising a total of 7523 commercial compound feed samples; 7423 were used as training set and 100 as validation set. Three general models of networks were studied: multilayer perceptron with back-propagation training (BP), multilayer perceptron with Levenberg-Maquartd training (LM), and radial basis function nets (RBF); moreover, in accordance with a factorial design, more complex architectures were evaluated gradually, changing the number of hidden layers and hidden neurons, for the determination of the optimal network topology. For both ingredients, the best results were obtained using ANN with BP training, showing prediction error values (SEP) of 2.72% and 0.66% for wheat and sunflower meal, respectively. These SEP values showed a significant improvement (19%-49% for sunflower meal and wheat, respectively) in comparison with those obtained using calibrations developed with linear methods.