arXiv subjects

Oleg I. Berngardt

Publications and source records attributed to Oleg I. Berngardt.

18 recordsLinked to original sources

arXiv2026

Physically-Informed Fuzzy Clustering of Vertical Sounding Ionograms

This paper presents a physically-informed fuzzy clustering of vertical sounding ionograms for automatically separating the ionogram into tracks suitable for further interpretation and determining their optimal number. The model is designed for use not only in conditions where the number of tracks is known, but also in disturbed ionospheric conditions where the number of tracks is preliminary unknown. The method is based on an expectation-maximization algorithm, used for clustering, and on parametrically specified distributions of distances from points to parametrically specified curves. The curves used as track models are close to model tracks in the parabolic ionospheric layer model. The resulting model of each track has six parameters: three standard ones (the critical frequency, the lower boundary of the layer, and its half-width), and three additional ones to take into account possible underlying layer effects. By sequentially increasing the number of tracks and optimizing their parameters, the model finds the optimal number of tracks on the ionogram by minimizing the modified Bayesian information criterion. The Sequential Least Squares Quadratic Programming algorithm is used to find the parameters of a single track. The width of each single track is assumed to be unknown constant found during fitting process. To improve the quality of ionogram clustering, automatic adaptive noise filtering is performed before clustering. This filtering is based on a combination of the DBSCAN and Gaussian Mixture algorithms. Also, to improve clustering quality on an ionosonde without hardware separation of the ordinary and extraordinary components, a preliminary approximate removal of points belonging to the extraordinary mode is performed.

physics.ao-ph↗

arXiv2024

Minimum number of neurons in fully connected layers of a given neural network (the first approximation)

This paper presents an algorithm for searching for the minimum number of neurons in fully connected layers of an arbitrary network solving given problem, which does not require multiple training of the network with different number of neurons. The algorithm is based at training the initial wide network using the cross-validation method over at least two folds. Then by using truncated singular value decomposition autoencoder inserted after the studied layer of trained network we search the minimum number of neurons in inference only mode of the network. It is shown that the minimum number of neurons in a fully connected layer could be interpreted not as network hyperparameter associated with the other hyperparameters of the network, but as internal (latent) property of the solution, determined by the network architecture, the training dataset, layer position, and the quality metric used. So the minimum number of neurons can be estimated for each hidden fully connected layer independently. The proposed algorithm is the first approximation for estimating the minimum number of neurons in the layer, since, on the one hand, the algorithm does not guarantee that a neural network with the found number of neurons can be trained to the required quality, and on the other hand, it searches for the minimum number of neurons in a limited class of possible solutions. The solution was tested on several datasets in classification and regression problems.

cs.LG↗

arXiv2023

Superclustering by finding statistically significant separable groups of optimal gaussian clusters

The paper presents the algorithm for clustering a dataset by grouping the optimal, from the point of view of the BIC criterion, number of Gaussian clusters into the optimal, from the point of view of their statistical separability, superclusters. The algorithm consists of three stages: representation of the dataset as a mixture of Gaussian distributions - clusters, which number is determined based on the minimum of the BIC criterion; using the Mahalanobis distance, to estimate the distances between the clusters and cluster sizes; combining the resulting clusters into superclusters using the DBSCAN method by finding its hyperparameter (maximum distance) providing maximum value of introduced matrix quality criterion at maximum number of superclusters. The matrix quality criterion corresponds to the proportion of statistically significant separated superclusters among all found superclusters. The algorithm has only one hyperparameter - statistical significance level, and automatically detects optimal number and shape of superclusters based of statistical hypothesis testing approach. The algorithm demonstrates a good results on test datasets in noise and noiseless situations. An essential advantage of the algorithm is its ability to predict correct supercluster for new data based on already trained clusterer and perform soft (fuzzy) clustering. The disadvantages of the algorithm are: its low speed and stochastic nature of the final clustering. It requires a sufficiently large dataset for clustering, which is typical for many statistical methods.

cs.LG↗

arXiv2023

Improving Classification Neural Networks by using Absolute activation function (MNIST/LeNET-5 example)

The paper discusses the use of the Absolute activation function in classification neural networks. An examples are shown of using this activation function in simple and more complex problems. Using as a baseline LeNet-5 network for solving the MNIST problem, the efficiency of Absolute activation function is shown in comparison with the use of Tanh, ReLU and SeLU activations. It is shown that in deep networks Absolute activation does not cause vanishing and exploding gradients, and therefore Absolute activation can be used in both simple and deep neural networks. Due to high volatility of training networks with Absolute activation, a special modification of ADAM training algorithm is used, that estimates lower bound of accuracy at any test dataset using validation dataset analysis at each training epoch, and uses this value to stop/decrease learning rate, and re-initializes ADAM algorithm between these steps. It is shown that solving the MNIST problem with the LeNet-like architectures based on Absolute activation allows to significantly reduce the number of trained parameters in the neural network with improving the prediction accuracy.

cs.LG↗

arXiv2022

Wrapped Classifier with Dummy Teacher for training physics-based classifier at unlabeled radar data

In the paper a method for automatic classification of signals received by EKB and MAGW ISTP SB RAS coherent scatter radars (8-20MHz operating frequency) during 2021 is described. The method is suitable for automatic physical interpretation of the resulting classification of the experimental data in realtime. We called this algorithm Wrapped Classifier with Dummy Teacher. The method is trained on unlabeled dataset and is based on training optimal physics-based classification using clusterization results. The approach is close to optimal embedding search, where the embedding is interpreted as a vector of probabilities for soft classification. The approach allows to find optimal classification algorithm, based on physically interpretable parameters of the received data, both obtained during physics-based numerical simulation and measured experimentally. Dummy Teacher clusterer used for labeling unlabeled dataset is gaussian mixture clustering algorithm. For algorithm functioning we extended the parameters obtained by the radar with additional parameters, calculated during simulation of radiowave propagation using ray-tracing and IRI-2012 and IGRF models for ionosphere and Earth's magnetic field correspondingly. For clustering by Dummy Teacher we use the whole dataset of available parameters (measured and simulated ones). For classification by Wrapped Classifier we use only well physically interpreted parameters. As a result we trained the classification network and found 11 well-interpretable classes from physical point of view in the available data. Five other found classes are not interpretable from physical point of view, demonstrating the importance of taking into account radiowave propagation for correct classification.

physics.geo-ph↗

arXiv2019

Interferometric calibration and the first elevation observations at EKB ISTP SB RAS radar at 10-12 MHz

The method for calibrating elevation measurements at EKB ISTP SB RAS radar obtained for the period 20/09/2019 - 18/11/2019 is presented. The calibration method is a modernization of the method for calibrating radar by meteor trails. The main difference of the method is the use of not a statistically processed FitACF data, but the full waveform of the signals scattered on the meteor trails. Using the full waveform makes it possible to more reliably distinguish meteor scattering from other possible scattered signal sources, and to determine meteor heights from the trail lifetime using the NRLMSIS-00 model. A comparison of the results with the results of E-layer calibration method shows a good agreement. The first examples of regular elevation observations at the EKB ISTP SB RAS radar are presented, and their preliminary interpretation is given.

physics.geo-ph↗

arXiv2019

Comparison of AATR and WTEC indices in the studies of the level of ionospheric disturbance

A comparative statistical analysis of AATR and WTEC indices was conducted based on data from the ISTP SB RAS GNSS receivers network. It is shown that at high levels of ionospheric disturbance (for WTEC > 0.1 TECU), the AATR index is proportional to the WTEC index with a factor of $1.5min^{-1}$. At small levels of ionospheric disturbance (for WTEC < 0.1 TECU), this proportionality is violated. It is shown that the contribution of daily dynamics of the background ionosphere to the AATR index is higher than to the WTEC index. This leads to a higher sensitivity of the WTEC index to disturbances. This also leads to violating the proportionality between WTEC and AATR indices at low levels of ionospheric disturbance. It is shown that at high latitudes the dynamics of the WTEC and AATR indices correlate significantly with the level of geomagnetic disturbance Kp. At mid-latitudes, the contribution of solar radiation variations (F10.7 index) and vertical seismic variations exceeds the influence of Kp variations. The program for calculating WTEC indices, used in the paper is put into open access.

physics.geo-ph↗

arXiv2019

Noise level forecasts at 8-20MHz and their use for morphological studies of ionospheric absorption variations at EKB ISTP SB RAS radar

In this paper, a method is described for using 8-20MHz noise absorption effect for realtime detecting radiowave absorption periods. The method is based on two empirical autoregression models of noise dynamics. The first (rough) prediction model is based on radar measurements of daily minimal noise dynamics averaged over 28-days with specially calculated weight coefficients. The second (fine) prediction model uses real-time scaling of rough model. The scaling is based on the comparison of this model with the experimental noise observations during previous 5 days. The models are based on the whole EKB ISTP SB RAS radar dataset (2013-2018). The rough model allows one to estimate the boundary beyond which the noise variations can be associated with absorption periods with a high degree of certainty. A joint analysis of simultaneous data on neighboring radar beams and at several frequencies reduces the detection errors, and allows to identify absorption events with a higher degree of confidence. Use of fine model allows to estimate absorption. The technique is validated by frequency dependence of absorption during two-frequency measurements. The found frequency dependence has an average exponent of the order of -1.5, which is in good agreement with the literature data and the data obtained earlier in analysis of solar X-ray flares. The use of the detection technique at EKB radar shows that most probable absorption over absorption events is about -0.65dB. Analysis of absorption of different amplitudes shows that low-intensity absorption events (0..-1.3dB) have slight local time dependence and mostly observed at north directions. For the storng absorption events (stronger than -1.3dB) the local time dynamics correlates well with noise level dynamics, and usually fills the whole radar field-of-view.

physics.geo-ph↗

arXiv2017

Analysis of the fine structure of HF ground backscatter and ionospheric scatter signals based on EKB radar data

The analysis was carried out of the scattered signal in the cases of ground backscatter and ionospheric scatter. The analysis is based on the data of the decameter coherent EKB ISTP SB RAS radar. In the paper the signals scattered in each sounding run were analysed before their statistical averaging. Based on the analysis and on previously studied mechanisms, a model is constructed for ionospheric scatter and ground backscatter signals. Within the framework of the Bayesian approach and based on a large amount of the data the technique and algorithm for separating these two types of signals are constructed. The statistical analysis of the results was carried out based on the EKB ISTP SB RAS data.

physics.geo-ph↗

arXiv2017

The first results of monitoring observations of a meteor echo at ISTP SB RAS EKB radar: algorithms, validation, statistics

The paper presents the implementation of algorithms for automatic search for signals scattered at meteor tracks based on the ISTP SB RAS EKB radar data. The algorithm is divided into two phases - the detection of a meteoric echo and the determination of its parameters. In general, the algorithm is similar to the algorithms used for specialized meteor radars, but uses direct fitting of received signal quadrature components by a model when finding the parameters of meteor trail. Based on the analysis of maximum of the Geminid meteor shower 13.12.2016 the aspect sensitivity of the scattered signals detected by the algorithm was shown . This proves that the detected signals correspond to scattering by irregularities elongated with the direction to the radiant of the meteor shower. This also confirms that the source of signals detected by the algorithm are meteor traces. In this paper, we solve the inverse problem of reconstructing the vector of the neutral wind velocity from the radar data by the weighted least-squares technique. A comparison of the roses of horizontal neutral winds obtained in the model of two-dimensional (horizontal) wind and three-dimensional wind is made. Their similarity is shown and expected predominance of the zonal component over the meridional one is shown too. The implemented algorithm allows us to process the scattered signals in the real time. The detection algorithm has started its continuous operation at the ISTP SB RAS EKB radar at the end of 2016.

physics.geo-ph↗

arXiv2017

Attenuation of decameter sky noise during x-ray solar flares in 2013-2017 based on the observations at midlatitude radars

Based on a joint analysis of the data from 10 midlatitude decameter radars the effects are investigated during 80 x-ray flares in the period 2013-2017. For the investigation nine mid-latitude SuperDARN radars of the northern hemisphere and Ekaterinburg coherent decameter radar of ISTP SB RAS are used. All the radars work in the same 8-20MHz frequency band and have similar hardware and software. During the analysis the temporal dynamics of noise from each of the radar direction and for each flare is investigated separately. As a result, on the basis of about 13000 daily measurements we found a strong anticorrelation between noise power and x-ray flare intensity, that allows to use short-wave sky noise to diagnose the ionospheric effects of x-ray solar flares. It is shown that in 88.3\% of cases an attenuation of daytime decameter radio noise is observed during solar flare, correlating with the temporal dynamics of the solar flare. The intensity of decameter noise anticorrelates well (the Pearson correlation coefficient better than -0.5) with the shape of the X-ray flare in the daytime (for solar elevation angle $>0$) in 33\% of cases, the average Pearson correlation over the daytime is about -0.34. Median regression coefficient between GOES 0.1-0.8nm x-ray intensity and daytime sky-noise attenuation is about $-4.4\cdot10^{4}\ [dB\cdot m^{2}/Wt]$. Thus, it has been shown that measurements of the decameter sky noise level at midlatitude decameter radars can be used to study the ionospheric absorption of high-frequency waves in the lower ionosphere during x-ray solar flares. This can be explained by the assumption that the most part of decameter sky noise detected by the radars can be interpreted as produced by ground sources at distances of the first propagation hop (\textasciitilde{} 3000 km).

physics.geo-ph↗

arXiv2014

Seismo-ionospheric effects associated with 'Chelyabinsk' meteorite during the first 25 minutes after its fall

This paper presents the properties of ionospheric irregularities elongated with Earth magnetic field during the first 25 minutes after the fall of the meteorite 'Chelyabinsk' experimentally observed with EKB radar of Russian segment of the SuperDARN. It is shown that 40 minutes before meteor fall the EKB radar started to observe powerful scattering from irregularities elongated with the Earth magnetic field in the F-layer. Scattering was observed for 80 minutes and stopped 40 minutes after the meteorite fall. During 9-15 minutes after the meteorite fall at ranges 400-1200 km from the explosion site a changes were observed in the spectral and amplitude characteristics of the scattered signal. This features were the sharp increase in the Doppler frequency shift of the scattered signal corresponding to the Doppler velocities about 600 m/s and the sharp increase of the scattered signal amplitude. This allows us to conclude that we detected the growth of small-scale ionospheric irregularities elongated with the Earth magnetic field at E-layer heights. Joint analysis with the seismic data and numerical modeling shows that the observed effect is connected with the passage of secondary acoustic front formed by supersonic seismic ground wave from the 'Chelyabinsk' meteorite. As a possible explanation the growth of elongated ionospheric irregularities may be caused by the passage of the high-speed acoustic wave in the ionosphere in the presence of high enough background electric field.

physics.geo-ph↗

arXiv2014

Optimal signals for spectral measurements at EKB SuperDARN radar: Theory and Experiment

The requirements for an optimal sequence for SuperDARN radar are stated. Obtained quasioptimal signals up to 26-pulse sequence are based on optimal and nearly optimal Golomb rulers. It is shown that standard 7-pulse SuperDARN sequence is one of the quasioptimal sounding signals, but 8-pulse kascan SuperDARN signal is not. The characteristics are calculated for quasioptimal sounding sequences up to 26-pulse one. It is shown that the most effective signals for the spectral measurements are 10, 12, 14, 18, 20, 24 and 26 pulse sequences. We present the results of the first spectral measurements at the EKB SuperDARN radar with quasioptimal 8- and 9-pulse sequences, as well as with the optimal 10-and 12-pulse sequences. A comparison with the results obtained with standard 7-pulse SuperDARN sequence were made. The continuity of results in amplitude and velocity is shown, the expected improvement in the spectral width are also demonstrated.

physics.geo-ph↗

arXiv2014

Dynamics of vertical ionospheric inhomogeneities over Irkutsk during 06:00-06:20UT 11/03/2011 caused by the Tohoku earthquake

According to the data of quasi-vertical sounding at Usolie-Tory path (midpoint - 52.3N, 103E, 120km distance), the dynamics of vertical structure of the midscale inhomogeneities of the plasma frequency was reconstructed. These irregularities are caused by Tohoku earthquake 11/03/2011 (38.3N, 142.4E). Observed horizontal velocities of propagation in the ionosphere and the spatial scales of the wave-like disturbance modes, generated by earthquakes, are calculated. As a result of numerical simulation involving the data from TALAYA seismic station (TLY, 51.7N, 103.7E) and comparison with the experiment it is shown that the vertical ionospheric irregularities of 5-40km range observed in first 20 minutes after the beginning of observations of effects (from 06:00 to 06:20 UT) are qualitatively explained by travelling of acoustic shock wave cone ( Mach cone ) from supersonic ground source - seismic wave to the heights up to 200km. It is shown that the most likely source of the shock wave were Z and E components of the seismic oscillations observed at TLY station. Irregularities observed after 06:20UT were apparently linked with other mechanisms. It is shown that the existing temporary CHIRP ionosonde temporal resolution (60 seconds) and the ionogram inversion technique used are not enough for detailed diagnosis of fine spatial vertical structures of this process.

physics.geo-ph↗

arXiv2011

Improving the spatial resolution by effective subtraction technique at Irkutsk incoherent scatter radar: the theory and experiment

We describe a sounding technique that allows us to improve spatial resolution of Irkutsk Incoherent Scatter Radar without loosing spectral resolution. The technique is based on transmitting of rectangle pulses of different duration in various sounding runs and subtracting correlation matrixes. Theoretically and experimentally we have shown, that subtraction of the mean-square parameters of the scattered signal for different kinds of the sounding signal one from another allows us to solve the problem within the framework of quasi-static ionospheric parameters approximation.

physics.space-ph↗

arXiv2009

Towards the dispersion relation for ionacoustic instabilities in weakly inhomogeneous ionospheric plasma at altitudes 80-200km and its low-frequency solution

In the paper within the approximation of the two-fluid magnetohydrodynamics and geometrooptical approximation the dispersion relation was found for ionacoustic instabilities of the ionospheric plasma at 80-200km altitudes in three-dimensional weakly irregular ionosphere. Low freqeuncy solution was found. The difference between obtained and standard solution becomes significant at altitudes above 140 km. As the analysis shown in this case the solution grows with time. The conditions for existence of such solution are the presence of co-directed electron density gradients and electron drifts and perpendicularity of line-of-sight to the magnetic field. The necessary conditions regularly exist at the magnetic equator. Detailed analysis has shown that this solution corresponds to well-known 150km equatorial echo and explains some of its statistical characteristics observed experimentally.

physics.plasm-ph↗

arXiv2002

Towards a new technique of incoherent scatter signal processing

This paper offers a new technique of incoherent scattering signal processing. The technique is based on the experimentally observed comb structure of the spectral power of separate realizations. The technique implies determining the positions and amplitudes of peaks in separate realizations, the formation - on their basis - of the spectral power of an individual realization not distorted by the smoothing function, and a subsequent summation of such spectra for the realizations. The technique has been tested using data from the Irkutsk incoherent scatter radar, both for the case of the incoherent scattering from thermal irregularities of plasma and for the case of the aspect scattering from instabilities elongated with the geomagnetic field.

physics.ao-ph↗

arXiv2001

Least Sqaure Method for Sum of the Functions Satysfying the Differential Equations with Polynomial Coefficients

We propose a linear algorithm for determining two function parameters by their linear combination. These functions must satisfy the first order differential equations with polynomial coefficients and our parameters are the coefficients of these polynomials. The algorithm consists of sequential solution by least squares method of two linear problems - first, differential equation polynomial coefficients determining for linear combination of two given functions and second - determining functions parameters by these polynomial coefficients. Numerical modeling carried by this scheme gives an good accordance under weak normal noise (with dispersion (<5%)).

math.NA↗