APPLICATION OF ARTIFICIAL NEURAL NETWORKS IN THE PROBLEMS OF MANAGING A GENETIC ALGORITHM
D.A. Petrosov, R.A. Vashchenko, A.A. Stepovoi, N.V. Petrosova, A.N. Zelenina
In modern intelligent decision support systems, there is still a problem associated with improving performance in structural and parametric synthesis of large discrete systems with specified behavior based on genetic algorithms. Currently, there are two main areas of research that are designed for mathematical or hardware-based performance improvements. One of the ways to increase hardware performance is to use parallel computing, which includes GPGPU (General-purpose computing on graphics processing units) technology. In this paper, we consider the possibility of increasing the speed of intelligent systems using a mathematical tool of artificial neural networks by introducing a control module for the genetic algorithm directly when performing decision synthesis. The process of structural-parametric synthesis is controlled by predicting and assessing the state of the genetic algorithm (convergence, attenuation, finding the population at local extremes) using artificial neural networks. This allows you to change the parameters of the operators directly in the process of decision synthesis, changing their destructive ability relative to the binary string, which leads to a change in the trajectory of the population in the decision space, and as a result, should increase the speed of intelligent decision support systems.
Keywords: genetic algorithm, intelligent information systems, artificial neural networks, system analysis.
MODEL OF THE PROCESS OF MANAGING A GENETIC ALGORITHM USING AN ARTIFICIAL NEURAL NETWORK BECAUSE OF STRUCTURAL-PARAMETRIC SYNTHESIS OF LARGE DISCRETE SYSTEMS
D.A. Petrosov, Al Saedi Mohanad Ridha Ghanim, S.Y. Beletskaya
In intelligent decision support systems aimed at solving the problems of structurally parametric synthesis of models of large discrete systems with a given behavior, based on genetic algorithms, it is often required to increase speed using not only hardware, but also mathematical ones. In this paper, we consider the processes that arise when using an evolutionary procedure consisting of four genetic algorithms adapted to the task of synthesizing under the control of an artificial neural network. Each model that is part of the decision-making block fulfills its function in the task of structural-parametric synthesis of simulation models of large discrete systems. That is, it searches for solutions based on: models of elements that make up the synthesized object; interelement connections; initial parameters of the functioning of the elements; parameters of the elements of the synthesized system, which can change in the synthesized model during its operation. As a control, the use of an artificial neural network is considered, which makes adjustments to the functioning parameters of the operators of the genetic algorithm and (or) the connection of various combinations of evolutionary procedures depending on the convergence of the evolutionary procedure. When creating a process model, modern methodologies IDEF0 and IDEF3 were used, aimed at solving problems of system analysis.
Keywords: evolutionary procedures, structural-parametric synthesis, genetic algorithms, artificial neural networks, system analysis, simulation.
APPLICATION OF ANN IN HUMAN-MACHINE INTERFACES
UDC 004.5, 612.817.2
N.A.Budko, R.Y.Budko, A.Y.Budko
Currently, there are almost no areas of human activity that are not concerned with automation, which has received the greatest popularity over the past few years. To date, the methods that are based on the organization and functioning of biological neural networks have become most famous. The article provides an analytical review of the possibilities of using artificial neural networks (ANN) in the development of human-machine interfaces based on various physical principles of interaction with the human body. This interface provides user interaction with the machines it manages. Examples of the use of human-machine interfaces in household, medical and military areas are given. Efficiency is due to the flexibility, nonlinearity, speed and learning of systems based on neural networks. Thus, users can monitor the process with great precision, achieving the best result. The problems of using ANNs in control systems of technical objects based on the recognition of natural speech, tracking the direction of sight, analysis of the electrical activity of the brain and muscle fibers of a person are considered. The tasks of pre-processing information, classification, analysis of the result obtained by processing the neural network are described.
Keywords: :man-machine interface, artificial neural networks, control, electromyogram, electroencephalogram.
FORECESSION ELECTROMYOGRAPHY RECOGNITION AND GESTURES SELECTION FOR PROTESIS CONTROL
UDC 612.743, 612.817.2
R.Y. Budko, N.N. Chernov, N.A. Budko, A.Y. Budko
The relevance of this study is due to one of the main problems existing today in the field of building man-machine interfaces – is the creation of an effective management system that interacts directly with the user and external devices replacing functions (prostheses, wheelchairs, etc.). In this regard, this work is devoted to the study of the possibility of using physiological gestures from the daily life of a person to control the prosthesis with the safety of the forearm for at least one third. The leading approach to the study of this problem is the use of methods of statistical processing of experimental data, digital signal processing, machine learning algorithms and pattern recognition. This approach allows a comprehensive study of the electromyogram (EMG) of the forearm when making voluntary movements at different levels of the implementation of the myo-control system. The article presents the results of the EMG study recorded for 11 arbitrary movements from a group of subjects, describes the procedure for pre-processing the EMG and identifying characteristic features for signal recognition, discloses a method for classifying movements using an artificial neural network based on radial basic functions (RBF). Eight of the most suitable for classification movements were identified and ranked according to the classification accuracy: relaxation (like zero movement), hand opening, fist, hand flexion, hand supination, hand extension, hand pronation, pinch. The materials of the article are of practical value for building systems based on the human-machine interface, as well as for classification tasks in electrophysiology applications.
Keywords: :electromyogram, prosthesis, biocontrol, human-machine interface, machine learning, artificial neural networks.
NON-STATIONARY TIME SERIES FORECASTING BASED ON
MULTIWAVELET POLYMORPHIC NETWORK
S.N. Verzunov, N.M Lychenko
There are many methods and models for forecasting non-stationary time series. How-ever, the problem of the accuracy and adequacy of the forecast of non-stationary time series has not been solved yet. In this paper, a new forecast model, based on a multiwavelet network with additional customizable parameters, which is called polymorphic, is proposed. The effi-ciency of the proposed model is compared with the well-known time series forecast models like autoregressive integrated moving average model, multilayer perceptron and hybrid model in which both models are combined. Three well-known real data sets (the Wolf’s sunspot data, the Canadian lynx data and the British pound/US dollar exchange rate data) were taken as empirical data. The comparison showed that forecast model based on the proposed multi-wavelet polymorphic network has a smaller prediction error for each series. This is achieved by introducing additional customizable parameters into the wavelet network, which allow to better adapt to the non-stationary nature of time series. Moreover, for the wavelet network to perform well in the presence of linearity, were used linear connections between the wavelet neurons of input and output layers. The proposed technology can be used to predict the time series gen-erated by dynamic processes of a different nature.
Keywords: :forecasting, non-stationary time series, multiwavelet network, additional customizable parameters, ARIMA-model, artificial neural networks, hybrid model.
FUZZY MATHEMATICAL MODELS FOR PREDICTION AND EARLY DIAGNOSIS OF OCCUPATIONAL DISEASES OF AGRICULTURAL WORKERS IN CONTACT WITH PESTICIDES
F. A. Surkov, N. V. Petkova, S. F. Sukhovskiy
This article deals with the problem of forecasting prices for real estate in the long and medium term for management decisions. The real estate market is one of the most dynamic areas of the Russian economy. Rapidly changing factors and price dynamics require a thorough study of new advanced methods using innovative technologies. Forecasting is an integral part of the mass valuation of real estate, it is impossible to plan future expenses or to build economic development plans. The price situation described by the average prices in the residential real estate market is a fundamental object for evaluation and forecasting in the study of the residential real estate market. Based on average prices, prices are managed in the residential real estate market. These indicators are considered when forecasting the market price of real estate, which is important in the development of the subjects of the real estate market auxiliary techniques for the selection of strategic actions for the development and improvement of the housing sector. Mass valuation of real estate as a complex system requires not only the definition of the parameters characterizing the price of real estate, and the identification of dependencies that link these parameters, but also the construction of a forecast of real estate prices in the future. Market conditions are constantly changing, and time has a direct impact on all market processes and decision-making. Seasonal calibration of prices for real estate objects is executed. The idea of using artificial neural networks that meet the modern requirements of real estate valuation is analyzed and proposed. A mathematical model based on harmonic series (Fourier series) and a neural network model are constructed and analyzed. A comparative analysis of the growth trends in the value of real estate.
Keywords: time series, real estate valuation, Fourier series, statistical methods, artificial neural networks.
THE SHORTCOMINGS IN THE USE OF ARTIFICIAL NEURAL NETWORKS FOR SOLVING PROBLEMS OF BINARY CLASSIFICATION
N. I. Murashkin, V. N. Kostrova
The article aims at identifying problems of the use of artificial neural networks for solving problems of binary classification. For solving problems of binary classification it is expected to classify samples that are already available to a certain class. A leading approach to the study of this problem is the algorithm of Levenberg-Marquardt, which allows to optimize the parameters of nonlinear regression models. As optimization criterion is adopted to the root mean square error of the model on the training set. It is proposed to accelerate the computation to apply the method to the elastic distribution. The materials of the article are of practical value to professionals who use artificial neural networks for classification tasks.
Keywords: binary tasks, artificial neural networks, algorithm, algorithm of Levenberg-Marquardt, algorithm of Gauss-Newton, method of M. Riedmiller and G. Brown.