Search academic texts

Information × Registration Number 0213U001797, 0112U003024 , R & D reports Title Chemometric methods development for solving problem of gualitative chemical analysis popup.stage_title Head Pushkareva, Registration Date 12-02-2013 Organization Kharkov National University after named V.N. Karazin popup.description2 The purpose of the research project - exploring the applicability of artificial neural networks, developing recommendations for optimizing the parameters of artificial neural networks (the number of neurons, types of transfer functions, training methods) for reliable solving tasks of qualitative chemical analysis. Methods applied - algorithms of artificial neural networks, preliminary analysis of data sets, parametric and non-parametric methods of statistics and chemometrics. Within the scope of the project the urgent scientific problem of adapting artificial neural networks for the reliable identification and discrimination of objects in qualitative chemical analysis has been solved. The recommendations for the choice of optimal architecture and parameters of artificial neural networks have been developed. This allows: - to recommend the use of the Levenberg-Marquardt learning algorithm and hyperbolic tangent and linear activation functions; - to propose the rule for the determination of the optimal number of hidden neurons; - to propose the novel classification procedure based on a combination of the Kohonen and the probabilistic neural networks (the procedure does not require any a priori information about the number of classes and the patterns in the training set); - to recommend the probabilistic and dynamic neural networks as the most robust algorithms for the classification of objects; - to identify the geographical origins of vegetables and fruits on the base of their composition with the use of probabilistic neural network and statistical methods. Product Description popup.authors Пантелеймонов А. Пушкарьова Я Слєдзевська А. popup.nrat_date 2020-04-02 Close