Alexander Kugaevskikh

ITMO University | SPbNRU ITMO

The article presents the result of the developed architecture experimental verification of a recurrent neural network for predicting oil and gas consumption by the parameters of the Venturi flow meter. The prediction quality, measured using the determination coefficient, is at the level of 0,90 for oil flow and 0,92 for gas flow. In addition to the...

This article describes the use of convolutional neural networks to screening first stage of the COVID on exhale spectra. A distinctive feature is the use of the glow-dicharge optical spectroscopy. The hypothesis put forward about the use of spectra images, and not the spectra themselves, for classification was confirmed. Accuracy was 87%. However,...

This paper presents the result of designing the architecture of a neural network on bio-inspired neurons, whose task is to work out the mechanism for recognizing an illusory contour using the example of “Kanizsa’s figures”. The neural network made it possible to achieve invariance to the number of corners of the figure and does not lose recognition...

The article presents a new model of the MT neuron (neuron of the middle temporal region), which allows motion detecting and determining its direction and speed without the use of recurrent communication. The model is based on signal accumulation and is organized using a space-time vector that sets the weighting coefficients. The space-time vector i...

This article is dedicated to modeling the end-stopped neuron. This type of neuron gives the maximum response at the end of the line and is used to refine the edge. The article provides an overview of different models of end-stopped neurons. I have proposed a simpler and more accurate model of an end-stopped neuron based on the use of Gabor filters...

The paper presents a new model of the MT-neuron (Middle temporal area neuron), which allows detecting movement and determining its direction and speed, without using recurrent connection. The model is based on the accumulation of the signal and is organized using a space-time vector that sets the weight coefficients. Despite the combinatorial redun...

In this paper discusses the current situation in Russia and the world in the field of development of sign languages translation system. The main problems are formulated, and ways to solve them are given. One of the most important unresolved tasks is the task of recognizing the gestures of the deaf. To effectively solve it, an approach based on the...

This article is dedicated to modeling the end-stopped neuron. This type of neuron gives the maximum response at the end of the line and is used to refine the edge. The article provides an overview of different models of end-stopped neurons. I have proposed a simpler and more accurate model of an end-stopped neuron based on the use of Gabor filters...

—This paper describes the architecture of a neural network for edge detection. Different filters for first-layer neurons are compared. Neural network learning based on a cosine measure algorithm shows much worse results than an error backpropagation algorithm. Optimal parameters for the first-layer neuron operation are given. The proposed architect...

In this paper I proposed generalized formula of Gabor filter. It was made according to the extensive review of sources. I carried out comparing of different coefficients influencing the form of a kernel on a DC component metrics. Thus the optimum formula was defined. The significant parameters (wavelength and scale of the filter) influencing qualit...

В современных условиях точные измерения выхода на газоконденсатных скважинах затруднены и, как следствие, затруднено прогнозирование, обычно измеряют параметры на участках сбора, но при этом нельзя получить четкую картину по каждой конкретной скважине. Тем не менее, на основании истории изменения некоторых параметров скважины можно предсказать ее в...

В работе представлен новый подход к распознаванию иероглифических текстов со слабо изученной лингвистикой. Классическая нейронная сеть была модифицирована для улучшения качества распознавания. Применение фильтра Габора позволило улучшить выявление границ символов, а совмещение сегментации и распознавания в одной нейросетевой архитектуре улучшить во...

In training of any neural network there is a problem of an assessment of quality of training that is defined by training of separate layers or neurons. In case of large number neurons it is difficult to trace flows of neurons activation. This paper shows the visualization mechanism of neural network training, the way of neurons activation allowing...

В работе предложен новый алгоритм сегментации, основанный на агломеративном подходе. Основной принцип состоит в построении пирамиды сегментации, каждый уровень которой представлен в виде неориентированного графа. На первом уровне вершинами графа являются пиксели изображения, на втором и последующих – выделенные сегменты. Также строятся деревья влож...

В статье рассмотрена возможность применения аппарата искусственных нейронных сетей на примере неокогнитрона для задачи распознавания древнеегипетских иероглифов.