Open Access
Access granted
Subscription Access
Vol 26, No 1 (2017)
- Year: 2017
- Articles: 9
- URL: https://journals.rcsi.science/1060-992X/issue/view/12240
Article
Deep learning: an overview and main paradigms
Abstract
In the present paper, we examine and analyze main paradigms of learning of multilayer neural networks starting with a single layer perceptron and ending with deep neural networks, which are considered regarded as a breakthrough in the field of the intelligent data processing. The baselessness of some ideas about the capacity of multilayer neural networks is shown and transition to deep neural networks is justified. We discuss the principal learning models of deep neural networks based on the restricted Boltzmann machine (RBM), an autoassociative approach and a stochastic gradient method with a Rectified Linear Unit (ReLU) activation function of neural elements.
1-17
Clearly defined architectures of neural networks and multilayer perceptron
Abstract
Neural networks with clearly defined architecture differ in the fact that they make it possible to determine the structure of neural network (number of neurons, layers, connections) on the basis of initial parameters of recognition problem. For these networks, the value of weights determined also analytically. In this paper, we consider the problem of networks with clearly defined architecture transformation into the classical schemes of multilayer perceptron architectures. Such possibility may allow us to combine the advantages of neural networks with clearly defined architecture with the capabilities of multilayer perceptron, that eventually may enable us to speed up and simplify the process of creating and training a neural network.
62-76
Computer simulations of association-based image search mechanisms basing on theory of active perception
Abstract
We discuss an approach to development of an associative memory model from the viewpoint of the theory of active perception. The theory of active perception allows one to develop the treebased memory model without the defects of the kd-tree and the vp-tree. Applications of the proposed model for solving problems of an image search by content from database are described. Also, we present the results of computer simulations directed at searching of similar and distorted images.
77-86
Iterative approach to solve the inverse diffraction problem under sharp focusing conditions
Abstract
In this paper, we consider an iterative approach to solve the inverse diffraction problem under sharp focusing conditions. It is shown that using this approach we can obtain a complex distribution in the entrance of the focusing system that creates desired intensity distribution in the focal area. As a result of the iteration process we obtained a complex distribution of transverse components of the initial field which reminds of Laguerre–Gauss mode with radial polarization and is focused in a light spot twice smaller than the diffraction limit.
18-25
Direct operation over the UWB optical signal transferred with pseudo-random carrier
Abstract
Modulation of white light emitted by the thermal source using simple optical systems capable to realize UWB information transfer are considered and realized. The experiments fulfilled permitted to estimate the minimum time resolution of the detection of signal with two-beams interferometer as 3.6 fs. For the effective time encoding method, the system based on Michelson echelon is proposed, its theory is developed using δ-wave approach for any angle of the beam incidence and scattering. Peculiarities of the echelon pulse response are analyzed. In addition, the simple method of digital encoding of the light flow with the pack of thin glasses of different thickness is demonstrated.
26-33
Optical properties of the glass composites with nano dimensional films dependence on the sol disperse phase characteristics studies
Abstract
It has been proved that the characteristics of sols affect the optical properties of composites with the glass surface and nanofilms. Film thickness and composite transmission are proportionally on the volume fraction of the sol disperse phase. The laser ablation threshold energy density under the irradiation of a nanosecond laser radiation pulse increases with increasing the packing density of the dispersed phase in the layer. And the ablation threshold energy density under the laser radiation microseconds pulse increases with increasing particles size of the dispersed phase and decreases with the increasing volume fraction of the dispersed phase and the particle packing density.
34-39
Development of the algorithm of adaptive construction of hierarchical neural network classifiers
Abstract
This paper presents the development of the algorithm for adaptive construction of hierarchical neural network classifiers based on automatic modification of the desired response of a perceptron with a small number of neurons in a single hidden layer. Improved versions of the algorithm are tested on standard benchmark problems Vowels and MNIST. A discussion of the results, strengths and weaknesses of the algorithm, directions of further work on its testing and improvement, is provided.
40-46
Ensembles of neural networks for forecasting of time series of spacecraft telemetry
Abstract
We analyzed different approaches to developing ensembles of neural networks in respect to their forecasting accuracy. We describe a two level model of ensembles of neural networks for forecasting of telemetry time series of spacecraft’s subsystems. A possibility of additional training of these ensembles of neural networks is examined. Our results show that use of ensembles of neural networks with dynamic weighing allows us to reduce the forecasting error.
47-54
Evolutionary algorithm for structural-parametric optimization of the remote photoplethysmography method
Abstract
We analyzed a problem of determination of an optimal combination of algorithms and parameters of the remote photoplethysmography method for heart rate estimation. To solve this problem we developed an evolutionary algorithm implementing an adaptive reinforcement learning. The optimal solution of the problem of the remote photoplethysmography was obtained with the aid of an original criterion combining the mean square error with the dutation of agent’s life cycle.
55-61