scholarly journals Neuroevolutionary reinforcing learning of neural networks

2022 ◽  
pp. 16-24
Author(s):  
Y. A. Bury ◽  
D. I. Samal
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Evolutionary Process ◽  
Control Device ◽  
Neural Network Learning ◽  
The Neural Network ◽  
Neural Network Structure ◽  
Living Nature ◽  
Developmental Features ◽  
Self Learning

The article presents the results of combining 4 different types of neural network learning: evolutionary, reinforcing, deep and extrapolating. The last two are used as the primary method for reducing the dimension of the input signal of the system and simplifying the process of its training in terms of computational complexity.In the presented work, the neural network structure of the control device of the modeled system is formed in the course of the evolutionary process, taking into account the currently known structural and developmental features of self-learning systems that take place in living nature. This method of constructing it makes it possible to bypass the specific limitations of models created on the basis of recombination of already known topologies of neural networks.

Neural Network-Based Speed Control of A Two-Mass-Model System

1999 ◽  
Vol 3 (5) ◽  
pp. 427-430 ◽  
Author(s):  
Rached Dhaouadi ◽  
◽  
Khaled Nouri
Keyword(s):  
Neural Network ◽  
Inverse Dynamics ◽  
Mass Model ◽  
Neural Network Learning ◽  
Inverse Control ◽  
Network Learning ◽  
The Neural Network ◽  
Control Scheme ◽  
On Line ◽  
Neural Network Structure

We present an application of artificial neural networks to the problem of controlling the speed of an elastic drive system. We derive a neural network structure to simulate the inverse dynamics of the system, then implement the direct inverse control scheme in a closed loop. The neural network learning is done on-line to adaptively control the speed to follow a stepwise changing reference. The experimental results with a two-mass-model analog board confirm the effectiveness of the proposed neurocontrol scheme.

scholarly journals Robot Manipulators Control with Guaranteed Stability Using Feedback Error Learning Neural Networks

1996 ◽  
Vol 8 (4) ◽  
pp. 383-391
Author(s):  
Ju-Jang Lee ◽  
◽  
Sung-Woo Kim ◽  
Kang-Bark Park
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Learning Rule ◽  
Neural Network Learning ◽  
Network Learning ◽  
The Neural Network ◽  
Control Schemes ◽  
Feedback Error Learning ◽  
Feedback Error ◽  
Error Learning

Among various neural network learning control schemes, feedback error learning(FEL)8),9) has been known that it has advantages over other schemes. However, such advantages are founded on the assumption that the systems is linearly parameterized and stable. Thus, FEL has difficulties in coping with uncertain and unstable systems. Furthermore, it is not clear how the learning rule of FEL is obtained in the minimization sense. Therefore, to overcome such problems, we propose neural network control schemes using FEL with guaranteed performance. The proposed strategy is to use multi-layer neural networks, to design a stabilityguaranteeing controller(SGC), and to derive a learning rule to obtain the tracking performance. Using multilayer neural networks we can fully utilize the learning capability no matter how the system is linearly parameterized or not. The SGC makes it possible for the neural network to learn without fear of instability. As a result, the more the neural network learning proceeds, the better the tracking performance becomes.

A Self-Learning Model for Cylinder Wakes Using Neural Networks

2002 ◽  
Author(s):  
G. Balasubramanian ◽  
D. J. Olinger ◽  
M. A. Demetriou
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Adaptive Estimation ◽  
Lattice Models ◽  
Spectral Element ◽  
The Self ◽  
Projection Algorithm ◽  
The Neural Network ◽  
Wake Patterns ◽  
Self Learning

Coupled map lattice models (CML) that combine a series of low-dimensional circle maps with a diffusion model have predicted qualitative features of the wake behind vibrating flexible cables. However, there are always unmodelled dynamics if a quantitative comparison is made with wake patterns obtained from laboratory or simulated wake flows. To overcome this limitation, self-learning features can be incorporated into the simple CML model to capture the unmodelled dynamics. The self-learning CML uses radial basis function neural networks as online approximators of the unmodelled dynamics. The neural network weights are adaptively varied using a combination of a multivariable least squares algorithm and a projection algorithm. The adaptive estimation scheme, derived from a new convective diffusive CML, seeks to precisely estimate the neural network weights at each timestep by mimimizing the error between the simulated and measured wake patterns. Studies of this approach are conducted using wake patterns from spectral element based NEKTAR simulations of freely vibrating cable wakes at Re = 100. It is shown that the neural network based self-learning CML precisely estimates the simulated wake patterns within several shedding cycles. The self-learning CML is also shown to be computationally efficient.

scholarly journals Development and training of neural networks for character recognition

Connectivity ◽  
2020 ◽  
Vol 148 (6) ◽  
Author(s):  
R. D. Bukov ◽  
◽  
I. S. Shcherbyna ◽  
O. V. Nehodenko ◽  
Ye. S. Tykhonov
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Artificial Neural Networks ◽  
Character Recognition ◽  
Intelligent Systems ◽  
Recognition System ◽  
The Neural Network ◽  
Reference Images ◽  
Artificial Neural ◽  
Self Learning

This article discusses the problem of the application of neural networks for character recognition, as well as the problem of developing methods and algorithms for the synthesis of neural networks. To solve the problems of optimizing the character recognition system, highly intelligent systems based on artificial neural networks are often used. However, artificial neural networks are not a tool for solving problems of any type. They are unsuitable for tasks such as payroll, but they have an advantage for character recognition tasks that conventional personal computers do poorly or not at all. It has been proven that artificial neural networks can be used for predictive modeling, adaptive control and applications where they can be trained using a dataset. Experiential self-learning can occur in networks that can draw inferences from a complex and seemingly unrelated set of information. The application of neural networks for solving practical problems in the field of character recognition and their classification is shown. It has been established that images can denote objects of different nature: text symbols, images, sound samples. When training the network, various sample images are offered with an indication of which class they belong to. At the end of training the network, you can present previously unknown images and receive an answer from it about belonging to a certain class. The topology of such a network is characterized by the fact that the number of neurons in the output layer, as a rule, is equal to the number of conditioned classes. This establishes a correspondence between the output of the neural network and the class it represents. A method for training a neural network is proposed, according to which the person managing the network takes a direct part in training the network, it itself sets the reference images of all symbols, as well as distorted images of the standards (plagued copies).

scholarly journals Application of a Neural Network for the Analysis of the “Soyuz” Launch Vehicle Boosters’ Separation.

2020 ◽  
pp. 42-56
Author(s):  
M.M. Matushin ◽  
D.A. Makhalov
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Back Propagation ◽  
Automated Analysis ◽  
General Information ◽  
Network Configuration ◽  
Neural Network Learning ◽  
Network Learning ◽  
The Neural Network ◽  
Trained Neural Network

The paper discusses application of artificial intelligence (neural networks) technologies for automated analysis of dynamic processes of the “Soyuz” launch vehicle’s onboard systems. Cyclogram of strap-on boosters separa-tion as applied to this task, and telemetry measurement used to monitor this process are described. The general information about the construction of the used types of neural networks and about their learning using a back-propagation method is presented; the neural network configuration for solving the mentioned task, telemetry presentation format suitable for sup-plying power for the neural network, and features of the neural network learning are proposed. The approbation of the trained neural network for the analysis of launches of the “Soyuz-FG” and “Soyuz-2.1a” launch vehi-cles using telemetry in real-time and delayed modes was carried out.

scholarly journals Classification of companies with theassistance of self-learning neural networks

2010 ◽  
Vol 56 (No. 2) ◽  
pp. 51-58 ◽  
Author(s):  
V. Konečný ◽  
O. Trenz ◽  
E. Svobodová
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Financial Indicators ◽  
Financial Situation ◽  
Learning Network ◽  
The Neural Network ◽  
Major Attention ◽  
Subjective View ◽  
Self Learning

The article is focused on rating classification of financial situation of enterprises using self-learning artificial neural networks. This is such a situation where the sets of objects of the particular classes are not well-known. Otherwise, it would be possible to use a multi-layer neural network with learning according to models. The advantage of a self-learning network is particularly the fact that its classification is not burdened by a subjective view. With reference to complexity, this sorting into groups may be very difficult even for experienced experts. The article also comprises the examples which confirm the described method functionality and the neural network model used. A major attention is focused on the classification of agricultural companies. For this purpose, financial indicators of eighty one agricultural companies were used.

scholarly journals Convolutional Neural Networks Training for Autonomous Robotics

2021 ◽  
Vol 29 (1) ◽  
pp. 75-79
Author(s):  
Alexander Lozhkin ◽  
Konstantin Maiorov ◽  
Pavol Bozek
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Convolutional Neural Networks ◽  
Semiotic Analysis ◽  
Autonomous Robotics ◽  
New Approach ◽  
Neural Network Learning ◽  
The Neural Network ◽  
Gradient Descent Algorithm ◽  
Modern Mathematics

AbstractThe article discusses methods for accelerating the operation of convolutional neural networks for autonomous robotics learning. The analysis of the theoretical possibility of modifying the neural network learning mechanism is carried out. Classic semiotic analysis and the theory of neural networks is proposed to union. An assumption is made about the possibility of using the symmetry mechanism to accelerate the training of convolutional neural networks. A multilayer neural network to represent how space is an attempt has been made. The conclusion was based on the laws on the plane obtained earlier. The derivation of formulas turned out to be impossible due to the problems of modern mathematics. A new approach is proposed, which involves combining the gradient descent algorithm and the stochastic completion of convolutional filters by the principles of symmetries. The identified algorithms allow increasing the learning rate from 5% to 15%, depending on the problem that the neural network solves.

scholarly journals APPLICATION OF THE EVOLUTIONARY PARADIGM TO DESIGNING ARCHITEСTURE OF A NEURAL NETWORK FOR RECOGNIZING THE DISTORTED TEXT

2018 ◽  
pp. 45-50
Author(s):  
Y. A. Bury ◽  
D. I. Samal
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Intelligent Systems ◽  
Evolutionary Process ◽  
Neural Systems ◽  
Present Moment ◽  
Wide Range ◽  
Natural Analogues ◽  
Scope Of Application ◽  
Living Nature

The paper presents an attempt to apply of evolutionary methods to the design and training of a system for recognizing distorted text.Over the past decades, artificial neural networks are widely used in many areas of artificial intelligence, such as forecasting, optimization, data analysis, pattern recognition and decision making. Nevertheless, the traditional heuristic approaches to design of multi-layer neural networks are based on the recombination of already existing neural network architectures.This approach allows us to solve a wide range of problems, but implies compliance with specific conditions for the quality work of algorithms.The natural analogues of such intelligent systems in living nature, however, are universal enough to adapt to virtually any habitat.Despite their extreme complexity and limited ability to study their structures, it is known that these structures were formed as a result of the evolutionary process. And if today it is impossible to determine the exact architecture of the links in biological neural systems, then at least one can try to reproduce the very process of their formation in order to obtain a more universal algorithm than those developed to the present moment.In opposite to them we form the final structure of the core of the classification system by evolutionary process, taking into account the knowledge about the features of the development and construction of the nervous system of vertebrates.Applying of the approach makes it possible to abstract from the limitations of existing neural network algorithms, caused by the scope of application of specific types of their structures.

SCORING MODELING BASED ON NEURAL NETWORKS FOR DETERMINING A BANK BORROWER'S RATING

2020 ◽  
Vol 2020 (10) ◽  
pp. 54-62
Author(s):  
Oleksii VASYLIEV ◽  
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Network Architecture ◽  
Statistical Data ◽  
Activation Function ◽  
Decision Making Process ◽  
Neural Network Architecture ◽  
Acceptable Accuracy ◽  
The Neural Network ◽  
Sigmoid Activation Function

The problem of applying neural networks to calculate ratings used in banking in the decision-making process on granting or not granting loans to borrowers is considered. The task is to determine the rating function of the borrower based on a set of statistical data on the effectiveness of loans provided by the bank. When constructing a regression model to calculate the rating function, it is necessary to know its general form. If so, the task is to calculate the parameters that are included in the expression for the rating function. In contrast to this approach, in the case of using neural networks, there is no need to specify the general form for the rating function. Instead, certain neural network architecture is chosen and parameters are calculated for it on the basis of statistical data. Importantly, the same neural network architecture can be used to process different sets of statistical data. The disadvantages of using neural networks include the need to calculate a large number of parameters. There is also no universal algorithm that would determine the optimal neural network architecture. As an example of the use of neural networks to determine the borrower's rating, a model system is considered, in which the borrower's rating is determined by a known non-analytical rating function. A neural network with two inner layers, which contain, respectively, three and two neurons and have a sigmoid activation function, is used for modeling. It is shown that the use of the neural network allows restoring the borrower's rating function with quite acceptable accuracy.

scholarly journals Discretization and machine learning approximation of BSDEs with a constraint on the Gains-process

2021 ◽  
Vol 0 (0) ◽  
Author(s):  
Idris Kharroubi ◽  
Thomas Lim ◽  
Xavier Warin
Keyword(s):  
Neural Network ◽  
Machine Learning ◽  
Neural Networks ◽  
Differential Equations ◽  
Numerical Experiments ◽  
Optimization Problem ◽  
Learning Approach ◽  
The Neural Network ◽  
Machine Learning Approach ◽  
Mesh Grid

AbstractWe study the approximation of backward stochastic differential equations (BSDEs for short) with a constraint on the gains process. We first discretize the constraint by applying a so-called facelift operator at times of a grid. We show that this discretely constrained BSDE converges to the continuously constrained one as the mesh grid converges to zero. We then focus on the approximation of the discretely constrained BSDE. For that we adopt a machine learning approach. We show that the facelift can be approximated by an optimization problem over a class of neural networks under constraints on the neural network and its derivative. We then derive an algorithm converging to the discretely constrained BSDE as the number of neurons goes to infinity. We end by numerical experiments.

Sign in / Sign up

Export Citation Format

Share Document