All issues

Deep learning’s power stems from complex architectures; however, these can lead to overfitting, where models memorize training data and fail to generalize to unseen examples. This paper proposes a novel probabilistic approach to mitigate this issue. We introduce two key elements: Truncated Log-Uniform Prior and Truncated Log-Normal Variational Approximation, and Automatic Relevance Determination (ARD) with Bayesian Deep Neural Networks (BDNNs). Within the probabilistic framework, we employ a specially designed truncated log-uniform prior for noise. This prior acts as a regularizer, guiding the learning process towards simpler solutions and reducing overfitting. Additionally, a truncated log-normal variational approximation is used for efficient handling of the complex probability distributions inherent in deep learning models. ARD automatically identifies and removes irrelevant features or weights within a model. By integrating ARD with BDNNs, where weights have a probability distribution, we achieve a variational bound similar to the popular variational dropout technique. Dropout randomly drops neurons during training, encouraging the model not to rely heavily on any single feature. Our approach with ARD achieves similar benefits without the randomness of dropout, potentially leading to more stable training.

To evaluate our approach, we have tested the model on two datasets: the Canadian Institute For Advanced Research (CIFAR-10) for image classification and a dataset of Macroscopic Images of Wood, which is compiled from multiple macroscopic images of wood datasets. Our method is applied to established architectures like Visual Geometry Group (VGG) and Residual Network (ResNet). The results demonstrate significant improvements. The model reduced overfitting while maintaining, or even improving, the accuracy of the network’s predictions on classification tasks. This validates the effectiveness of our approach in enhancing the performance and generalization capabilities of deep learning models.

This paper formulates and solves a practical problem of data recovery regarding the distribution of languages on regional level in context of China. The necessity of this recovery is related to the problem of the determination of the linguistic diversity indices, which, in turn, are used to analyze empirically and to predict sources of social and economic development as well as to indicate potential conflicts at regional level. We use Ethnologue database and China census as the initial data sources. For every language spoken in China, the data contains (a) an estimate of China residents who claim this language to be their mother tongue, and (b) indicators of the presence of such residents in China provinces. For each pair language/province, we aim to estimate the number of the province inhabitants that claim the language to be their mother tongue. This base problem is reduced to solving an undetermined system of algebraic equations. Given additional restriction that Ethnologue database introduces data collected at different time moments because of gaps in Ethnologue language surveys and accompanying data collection expenses, we relate those data to a single time moment, that turns the initial task to an ’ill-posed’ system of algebraic equations with imprecisely determined right hand side. Therefore, we are looking for an approximate solution characterized by a minimal discrepancy of the system. Since some languages are much less distributed than the others, we minimize the weighted discrepancy, introducing weights that are inverse to the right hand side elements of the equations. This definition of discrepancy allows to recover the required variables. More than 92% of the recovered variables are robust to probabilistic modelling procedure for potential errors in initial data.

The short-term prediction of road traffic condition is one of the main tasks of transportation modelling. The main purpose of which are traffic control, reporting of accidents, avoiding traffic jams due to knowledge of traffic flow and subsequent transportation planning. A number of solutions exist — both model-driven and data driven had proven to be successful in capturing the dynamics of traffic flow. Nevertheless, most space-time models suffer from high mathematical complexity and low efficiency. Artificial Neural Networks, one of the prominent datadriven approaches, show promising performance in modelling the complexity of traffic flow. We present a neural network architecture for traffic flow prediction on a real-world road network graph. The model is based on the combination of a recurrent neural network and graph convolutional neural network. Where a recurrent neural network is used to model temporal dependencies, and a convolutional neural network is responsible for extracting spatial features from traffic. To make multiple few steps ahead predictions, the encoder-decoder architecture is used, which allows to reduce noise propagation due to inexact predictions. To model the complexity of traffic flow, we employ multilayered architecture. Deeper neural networks are more difficult to train. To speed up the training process, we use skip-connections between each layer, so that each layer teaches only the residual function with respect to the previous layer outputs. The resulting neural network was trained on raw data from traffic flow detectors from the US highway system with a resolution of 5 minutes. 3 metrics: mean absolute error, mean relative error, mean-square error were used to estimate the quality of the prediction. It was found that for all metrics the proposed model achieved lower prediction error than previously published models, such as Vector Auto Regression, LSTM and Graph Convolution GRU.

The article reports the findings of an investigation into pollution migration processes at the municipal solid waste (MSW) landfill located in the water protection zone of Lake Seliger (Tver Region). The distribution of pollutants is investigated and migration parameters are determined in field and laboratory conditions at the landfill site. A mathematical model describing physical and chemical processes of substance migration in soil strata is constructed. Pollutant migration is found to be due to a variety of factors. The major ones, having a significant impact on the migration of MSW ingredients and taken into account mathematically, include convective transport, diffusion and sorption processes. A modified mathematical model differs from its conventional counterparts by considering a number of parameters reflecting the decrease in the concentration of ammonium and nitrate nitrogen ions in ground water (transpiration by plant roots, dilution with infiltration waters, etc.). An analytical solution to assess the pollutant spread from the landfill is presented. The mathematical model provides a set of simulation models helping to obtain a computational solution of specific problems, vertical and horizontal migration of substances in the underground flow. Numerical experiments, analytical solutions, as well as field and laboratory data was studied the dynamics of pollutant distribution in the object under study up to the lake. A long-term forecast for the spread of landfill pollution is made. Simulation experiments showed that some zones of clean groundwater interact with those of contaminated groundwater during the pollution migration from the landfill, each characterized by a different pollutant content. The data of a computational experiments and analytical calculations are consistent with the findings of field and laboratory investigations of the object and give grounds to recommend the proposed models for predicting pollution migration from a landfill. The analysis of the pollution migration simulation allows to substantiate the numerical estimates of the increase in $NH_4^+$ and $NO_3^-$ ion concentration with the landfill operation time. It is found that, after 100 years following the landfill opening, toxic filtrate components will fill the entire pore space from the landfill to the lake resulting in a significant deterioration of the ecosystem of Lake Seliger.

The paper reviews possibilities to predict buckling of thin cylindrical shells with non-destructive techniques during operation. It studies shallow shells made of high strength materials. Such structures are known for surface displacements exceeding the thickness of the elements. In the explored shells relaxation oscillations of significant amplitude can be generated even under relatively low internal stresses. The problem of the cylindrical shell oscillation is mechanically and mathematically modeled in a simplified form by conversion into an ordinary differential equation. To create the model, the researches of many authors were used who studied the geometry of the surface formed after buckling (postbuckling behavior). The nonlinear ordinary differential equation for the oscillating shell matches the well-known Duffing equation. It is important that there is a small parameter before the second time derivative in the Duffing equation. The latter circumstance enables making a detailed analysis of the obtained equation and describing the physical phenomena — relaxation oscillations — that are unique to thin high-strength shells.

It is shown that harmonic oscillations of the shell around the equilibrium position and stable relaxation oscillations are defined by the bifurcation point of the solutions to the Duffing equation. This is the first point in the Feigenbaum sequence to convert the stable periodic motions into dynamic chaos. The amplitude and the period of relaxation oscillations are calculated based on the physical properties and the level of internal stresses within the shell. Two cases of loading are reviewed: compression along generating elements and external pressure.

It is highlighted that if external forces vary in time according to the harmonic law, the periodic oscillation of the shell (nonlinear resonance) is a combination of slow and stick-slip movements. Since the amplitude and the frequency of the oscillations are known, this fact enables proposing an experimental facility for prediction of the shell buckling with non-destructive techniques. The following requirement is set as a safety factor: maximum load combinations must not cause displacements exceeding specified limits. Based on the results of the experimental measurements a formula is obtained to estimate safety against buckling (safety factor) of the structure.

Corrosion damage to metals and alloys is one of the main problems of strength and durability of metal structures and products operated in contact with chemically aggressive environments. Recently, there has been a growing interest in computer modeling of the evolution of corrosion damage, especially pitting corrosion, for a deeper understanding of the corrosion process, its impact on the morphology, physical and chemical properties of the surface and mechanical strength of the material. This is mainly due to the complexity of analytical and high cost of experimental in situ studies of real corrosion processes. However, the computing power of modern computers allows you to calculate corrosion with high accuracy only on relatively small areas of the surface. Therefore, the development of new mathematical models that allow calculating large areas for predicting the evolution of corrosion damage to metals is currently an urgent problem.

In this paper, the evolution of the corrosion front in the interaction of a polycrystalline metal surface with a liquid aggressive medium was studied using a computer model based on a cellular automat. A distinctive feature of the model is the specification of the solid body structure in the form of Voronoi polygons used for modeling polycrystalline alloys. Corrosion destruction was performed by setting the probability function of the transition between cells of the cellular automaton. It was taken into account that the corrosion strength of the grains varies due to crystallographic anisotropy. It is shown that this leads to the formation of a rough phase boundary during the corrosion process. Reducing the concentration of active particles in a solution of an aggressive medium during a chemical reaction leads to corrosion attenuation in a finite number of calculation iterations. It is established that the final morphology of the phase boundary has a fractal structure with a dimension of 1.323 ± 0.002 close to the dimension of the gradient percolation front, which is in good agreement with the fractal dimension of the etching front of a polycrystalline aluminum-magnesium alloy AlMg6 with a concentrated solution of hydrochloric acid. It is shown that corrosion of a polycrystalline metal in a liquid aggressive medium is a new example of a topochemical process, the kinetics of which is described by the Kolmogorov–Johnson– Meil–Avrami theory.

The repressilator is the first genetic regulatory network in synthetic biology, which was artificially constructed in 2000. It is a closed network of three genetic elements $lacI$, $\lambda cI$ and $tetR$, which have a natural origin, but are not found in nature in such a combination. The promoter of each of the three genes controls the next cistron via the negative feedback, suppressing the expression of the neighboring gene. In our previous paper [Bratsun et al., 2018], we proposed a mathematical model of a delayed repressillator and studied its properties within the framework of a deterministic description. We assume that delay can be both natural, i.e. arises during the transcription / translation of genes due to the multistage nature of these processes, and artificial, i.e. specially to be introduced into the work of the regulatory network using gene engineering technologies. In this work, we apply the stochastic description of dynamic processes in a delayed repressilator, which is an important addition to deterministic analysis due to the small number of molecules involved in gene regulation. The stochastic study is carried out numerically using the Gillespie algorithm, which is modified for time delay systems. We present the description of the algorithm, its software implementation, and the results of benchmark simulations for a onegene delayed autorepressor. When studying the behavior of a repressilator, we show that a stochastic description in a number of cases gives new information about the behavior of a system, which does not reduce to deterministic dynamics even when averaged over a large number of realizations. We show that in the subcritical range of parameters, where deterministic analysis predicts the absolute stability of the system, quasi-regular oscillations may be excited due to the nonlinear interaction of noise and delay. Earlier, we have discovered within the framework of the deterministic description, that there exists a long-lived transient regime, which is represented in the phase space by a slow manifold. This mode reflects the process of long-term synchronization of protein pulsations in the work of the repressilator genes. In this work, we show that the transition to the cooperative mode of gene operation occurs a two order of magnitude faster, when the effect of the intrinsic noise is taken into account. We have obtained the probability distribution of moment when the phase trajectory leaves the slow manifold and have determined the most probable time for such a transition. The influence of the intrinsic noise of chemical reactions on the dynamic properties of the repressilator is discussed.

Migration has a significant impact on the shaping of the demographic structure of the territories population, the state of regional and local labour markets. As a rule, rapid change in the working-age population of any territory due to migration processes results in an imbalance in supply and demand on labour markets and a change in the demographic structure of the population. Migration is also to a large extent a reflection of socio-economic processes taking place in the society. Hence, the issues related to the study of migration factors, the direction, intensity and structure of migration flows, and the prediction of their magnitude are becoming topical issues these days.

Mathematical tools are often used to analyze, predict migration processes and assess their consequences, allowing for essentially accurate modelling of migration processes for different territories on the basis of the available statistical data. In recent years, quite a number of scientific papers on modelling internal and external migration flows using mathematical methods have appeared both in Russia and in foreign countries in recent years. Consequently, there has been a need to systematize the currently most commonly used methods and tools applied in migration modelling to form a coherent picture of the main trends and research directions in this field.

The presented review considers the main approaches to migration modelling and the main components of migration modelling methodology, i. e. stages, methods, models and model classification. Their comparative analysis was also conducted and general recommendations on the choice of mathematical tools for modelling were developed. The review contains two sections: migration modelling methods and migration models. The first section describes the main methods used in the model development process — econometric, cellular automata, system-dynamic, probabilistic, balance, optimization and cluster analysis. Based on the analysis of modern domestic and foreign publications on migration, the most common classes of models — regression, agent-based, simulation, optimization, probabilistic, balance, dynamic and combined — were identified and described. The features, advantages and disadvantages of different types of migration process models were considered.

This work is devoted to the study of the problem of modeling and analyzing complex oscillatory modes, both regular and chaotic, in systems of interacting populations in the presence of random perturbations. As an initial conceptual deterministic model, a Volterra system of three differential equations is considered, which describes the dynamics of prey populations of two competing species and a predator. This model takes into account the following key biological factors: the natural increase in prey, their intraspecific and interspecific competition, the extinction of predators in the absence of prey, the rate of predation by predators, the growth of the predator population due to predation, and the intensity of intraspecific competition in the predator population. The growth rate of the second prey population is used as a bifurcation parameter. At a certain interval of variation of this parameter, the system demonstrates a wide variety of dynamic modes: equilibrium, oscillatory, and chaotic. An important feature of this model is multistability. In this paper, we focus on the study of the parametric zone of tristability, when a stable equilibrium and two limit cycles coexist in the system. Such birhythmicity in the presence of random perturbations generates new dynamic modes that have no analogues in the deterministic case. The aim of the paper is a detailed study of stochastic phenomena caused by random fluctuations in the growth rate of the second population of prey. As a mathematical model of such fluctuations, we consider white Gaussian noise. Using methods of direct numerical modeling of solutions of the corresponding system of stochastic differential equations, the following phenomena have been identified and described: unidirectional stochastic transitions from one cycle to another, trigger mode caused by transitions between cycles, noise-induced transitions from cycles to the equilibrium, corresponding to the extinction of the predator and the second prey population. The paper presents the results of the analysis of these phenomena using the Lyapunov exponents, and identifies the parametric conditions for transitions from order to chaos and from chaos to order. For the analytical study of such noise-induced multi-stage transitions, the technique of stochastic sensitivity functions and the method of confidence regions were applied. The paper shows how this mathematical apparatus allows predicting the intensity of noise, leading to qualitative transformations of the modes of stochastic population dynamics.

A method is proposed for analyzing the tremor of the earth’s surface, measured by means of space geodesy, in order to highlight the prognostic effects of seismicity activation. The method is illustrated by the example of a joint analysis of a set of synchronous time series of daily vertical displacements of the earth’s surface on the Japanese Islands for the time interval 2009–2023. The analysis is based on dividing the source data (1047 time series) into blocks (clusters of stations) and sequentially applying the principal component method. The station network is divided into clusters using the K-means method from the maximum pseudo-F-statistics criterion, and for Japan the optimal number of clusters was chosen to be 15. The Huang decomposition method into a sequence of independent empirical oscillation modes (EMD — Empirical Mode Decomposition) is applied to the time series of principal components from station blocks. To provide the stability of estimates of the waveforms of the EMD decomposition, averaging of 1000 independent additive realizations of white noise of limited amplitude was performed. Using the Cholesky decomposition of the covariance matrix of the waveforms of the first three EMD components in a sliding time window, indicators of abnormal tremor behavior were determined. By calculating the correlation function between the average indicators of anomalous behavior and the released seismic energy in the vicinity of the Japanese Islands, it was established that bursts in the measure of anomalous tremor behavior precede emissions of seismic energy. The purpose of the article is to clarify common hypotheses that movements of the earth’s crust recorded by space geodesy may contain predictive information. That displacements recorded by geodetic methods respond to the effects of earthquakes is widely known and has been demonstrated many times. But isolating geodetic effects that predict seismic events is much more challenging. In our paper, we propose one method for detecting predictive effects in space geodesy data.