DIAGNOSTIC SYSTEM BASED ON AUTOENCODERS
Journal Title: Електроніка та системи управління - Year 2018, Vol 1, Issue 55
Abstract
It`s considered the problem of image processing which is used in diagnostic systems when it is necessary to process the results of ultrasound, computed tomography and magnetic resonance imaging. For the solution of this problem it`s often used artificial neural networks especially convolution neural networks. It`s considered the structure of convolutional neural networks especially types of layers. In the paper it is analyzed the creation of convolutional neural networks based on autoencoders. It`s considered the features of such neural network, the algorithm of learning and the important parameters which determine the function quality of image processing. The possible improvement of such topology is possible with help of restricted Boltzmann machine which can be used for pre-learning.
Authors and Affiliations
Victor Sineglazov, Sviatoslav Kostiuchenko
Keywords
Related Articles
- EP ID EP523333
- DOI 10.18372/1990-5548.55.12771
- Views 139
- Downloads 0
ACCURACY ESTIMATION OF FLIGHT TEST RESULTS ON THE BASIS OF APPROXIMATING THE CORRELATION FUNCTION
The solution of the problem is considered, as a result of which estimates of statistical characteristics after processing the results of flight tests by the double averaging method, both in time and in the ensemble of re...
HIGH-PRECISION ONBOARD FUEL LEVEL CONTROL AND MEASUREMENT SYSTEM
The fueling systems of modern aircraft are considered. The need for their further improvement is shown in accordance with the recommendations of ICAO. A brief overview of methods and means of measuring fuel level in tank...
RADIO ELECTRONIC EQUIPMENT FAILURES MODEL
The article deals with the problem of radio electronic equipment failures model building. The model adequacy checking is performed according to the simulation results. The simulation problem is associated with the case w...
STRUCTURAL SYNTHESIS OF HYBRID NEURAL NETWORKS ENSEMBLES
It is considered the structural synthesis of hybrid neural networks ensembles. It is chosen the ensemble topology as parallel structure with united layer. It is developed a hybrid algorithm for the problem solution which...
USING THE THEORY OF MARTINGALES TO PROVE THE SOUNDNESS OF THE ESTIMATES OF THE PARAMETERS OF LINEAR DYNAMIC SYSTEMS
The article is devoted to analytical methods of research of filtration algorithms in conditions of a priori uncertainty of information on statistical characteristics of state noise and measurement in linear dynamic syste...