All issues

The paper provides a solution of a task of calculating the parameters of a Rician distributed signal on the basis of the maximum likelihood principle in limiting cases of large and small values of the signal-tonoise ratio. The analytical formulas are obtained for the solution of the maximum likelihood equations’ system for the required signal and noise parameters for both the one-parameter approximation, when only one parameter is being calculated on the assumption that the second one is known a-priori, and for the two-parameter task, when both parameters are a-priori unknown. The direct calculation of required signal and noise parameters by formulas allows escaping the necessity of time resource consuming numerical solving the nonlinear equations’ s system and thus optimizing the duration of computer processing of signals and images. There are presented the results of computer simulation of a task confirming the theoretical conclusions. The task is meaningful for the purposes of Rician data processing, in particular, magnetic-resonance visualization.

In this paper, a gravitational system is understood as a set of point bodies that interact according to Newton's law of attraction and have a negative value of the total energy. The question of the stability (nonstability) of a gravitational system of general position is discussed by direct computational experiment. A gravitational system of general position is a system in which the masses, initial positions, and velocities of bodies are chosen randomly from given ranges. A new method for the numerical solution of ordinary differential equations at large time intervals has been developed for the computational experiment. The proposed method allowed, on the one hand, to ensure the fulfillment of all conservation laws by a suitable correction of solutions, on the other hand, to use standard methods for the numerical solution of systems of differential equations of low approximation order. Within the framework of this method, the trajectory of a gravitational system in phase space is assembled from parts, the duration of each of which can be macroscopic. The constructed trajectory, generally speaking, is discontinuous, and the points of joining of individual pieces of the trajectory act as branch points. In connection with the latter circumstance, the proposed method, in part, can be attributed to the class of Monte Carlo methods. The general conclusion of a series of computational experiments has shown that gravitational systems of general position with a number of bodies of 3 or more, generally speaking, are unstable. In the framework of the proposed method, special cases of zero-equal angular momentum of a gravitational system with a number of bodies of 3 or more, as well as the problem of motion of two bodies, are specially considered. The case of numerical modeling of the dynamics of the solar system in time is considered separately. From the standpoint of computational experiments based on analytical methods, as well as direct numerical methods of high-order approximation (10 and higher), the stability of the solar system was previously demonstrated at an interval of five billion years or more. Due to the limitations on the available computational resources, the stability of the dynamics of the planets of the solar system within the framework of the proposed method was confirmed for a period of ten million years. With the help of a computational experiment, one of the possible scenarios for the disintegration of the solar systems is also considered.

The paper proposes and investigates a mathematical model of a distributed computing system of parallel interacting processes competing for the use of a limited number of copies of a structured software resource. In cases of unlimited and limited parallelism by the number of processors of a multiprocessor system, the problems of determining operational and exact values of the execution time of heterogeneous and identically distributed competing processes in a synchronous mode are solved, which ensures a linear order of execution of blocks of a structured software resource within each of the processes without delays. The obtained results can be used in a comparative analysis of mathematical relationships for calculating the implementation time of a set of parallel distributed interacting competing processes, a mathematical study of the efficiency and optimality of the organization of distributed computing, solving problems of constructing an optimal layout of blocks of an identically distributed system, finding the optimal number of processors that provide the directive execution time of given volumes of computations. The proposed models and methods open up new prospects for solving problems of optimal distribution of limited computing resources, synchronization of a set of interacting competing processes, minimization of system costs when executing parallel distributed processes.

This paper presents an experimental study of the application of automatic annotation of text data in the question – answer format (QA pairs) under conditions of limited computing resources and data protection requirements. Unlike traditional approaches based on rigid rules or the use of external APIs, we propose using small language models with a small number of parameters that can function locally without a GPU on standard CPU systems. Two models were selected for testing — Gemma-3-4b and Qwen-2.5-3b (quantized 4-bit versions) — and a corpus of documents with a clear structure and a formally rigorous style of presentation was used as source material. An automatic annotation system was developed that implements the full cycle of QA dataset generation: automatic division of the source document into logically connected fragments, formation of “question – answer” pairs using the Gemma-3-4b model, preliminary verification of their correctness using Qwen-2.5-3b based on evidence span from the context and expert quality assessment. The results are exported in JSONL format. Performance evaluation covers the entire QA pair generation system, including fragment processing by the local language model, text preprocessing and postprocessing modules. Performance is measured by the time it takes to generate a single QA pair, the total throughput of the system, RAM usage, and CPU load, which allows for an objective assessment of the computational efficiency of the proposed approach when running on a CPU. An experiment on an extended sample of 12 documents showed that automatic annotation demonstrates stable performance when processing different types of documents, while manual annotation is characterized by significantly higher time costs and high variability. Depending on the type of document, the acceleration of annotation compared to the manual process ranges from 8 to 14 times. Quality analysis showed that most of the generated QA pairs have high semantic consistency with the original context, with only a limited proportion of data requiring expert correction or exception. Although full manual validation of the corpus (the “gold standard”) was not performed as part of this work, the combination of automatic evaluation and selective expert review allows us to consider the resulting quality level acceptable for preliminary automated annotation tasks. Overall, the results confirm the practical applicability of small language models for building autonomous and reproducible automatic text annotation systems under limited computational resources and provide a basis for further research in the field of effective training corpus preparation for natural language processing tasks.

The paper develops a new mathematical method of the joint signal and noise parameters determination at the Rice statistical distribution by method of moments based upon the analysis of data for the 1-st and the 3-rd raw moments of the random rician value. The explicit equations’ system have been obtained for required parameters of the signal and noise. In the limiting case of the small value of the signal-to-noise ratio the analytical formulas have been derived that allow calculating the required parameters without the necessity of solving the equations numerically. The technique having been elaborated in the paper ensures an efficient separation of the informative and noise components of the data to be analyzed without any a-priori restrictions, just based upon the processing of the results of the signal’s sampled measurements. The task is meaningful for the purposes of the rician data processing, in particular in the systems of magnetic-resonance visualization, in ultrasound visualization systems, at the optical signals’ analysis in range measuring systems, in radio location, etc. The results of the investigation have shown that the two parameter task solution of the proposed technique does not lead to the increase in demanded volume of computing resources compared with the one parameter task being solved in approximation that the second parameter of the task is known a-priori There are provided the results of the elaborated technique’s computer simulation. The results of the signal and noise parameters’ numerical calculation have confirmed the efficiency of the elaborated technique. There has been conducted the comparison of the accuracy of the sought-for parameters estimation by the technique having been developed in this paper and by the previously elaborated method of moments based upon processing the measured data for lower even moments of the signal to be analyzed.

This paper presents the results of experimental validation of some structural issues concerning the practical use of methods to overcome catastrophic forgetting of neural networks. A comparison of current effective methods like EWC (Elastic Weight Consolidation) and WVA (Weight Velocity Attenuation) is made and their advantages and disadvantages are considered. It is shown that EWC is better for tasks where full retention of learned skills is required on all the tasks in the training queue, while WVA is more suitable for sequential tasks with very limited computational resources, or when reuse of representations and acceleration of learning from task to task is required rather than exact retention of the skills. The attenuation of the WVA method must be applied to the optimization step, i. e. to the increments of neural network weights, rather than to the loss function gradient itself, and this is true for any gradient optimization method except the simplest stochastic gradient descent (SGD). The choice of the optimal weights attenuation function between the hyperbolic function and the exponent is considered. It is shown that hyperbolic attenuation is preferable because, despite comparable quality at optimal values of the hyperparameter of the WVA method, it is more robust to hyperparameter deviations from the optimal value (this hyperparameter in the WVA method provides a balance between preservation of old skills and learning a new skill). Empirical observations are presented that support the hypothesis that the optimal value of this hyperparameter does not depend on the number of tasks in the sequential learning queue. And, consequently, this hyperparameter can be picked up on a small number of tasks and used on longer sequences.

The article is devoted to solving ill-posed problems of mathematical physics for elliptic and parabolic equations, such as the Cauchy problem for the Helmholtz equation and the retrospective Cauchy problem for the heat equation with constant coefficients. These problems are reduced to problems of convex optimization in Hilbert space. The gradients of the corresponding functionals are calculated approximately by solving two well-posed problems. A new method is proposed for solving the optimization problems under study, it is component-by-component descent in the basis of eigenfunctions of a self-adjoint operator associated with the problem. If it was possible to calculate the gradient exactly, this method would give an arbitrarily exact solution of the problem, depending on the number of considered elements of the basis. In real cases, the inaccuracy of calculations leads to a violation of monotonicity, which requires the use of restarts and limits the achievable quality. The paper presents the results of experiments confirming the effectiveness of the constructed method. It is determined that the new approach is superior to approaches based on the use of gradient optimization methods: it allows to achieve better quality of solution with significantly less computational resources. It is assumed that the constructed method can be generalized to other problems.

This paper proposes an architectural solution for organizing a convolutional neural network (CNN) oriented towards hardware implementation on edge devices under limited resources. To this goal, an approach to compressing spectrograms to a given size (28 × 28) is proposed using discretization, monoconversion, windowed Fourier transform, and two-dimensional interpolation. A balanced convolution procedure is developed based on compact convolutional filters, the size of which provides the balance between computational complexity and accuracy required for edge devices. An algorithm that enables convolution operations and calculation of the error function gradient in the convolutional layer in a single cycle ensuring increased performance in both inference and training modes of the CNN is proposed. The tradeoff between network trainability and its resistance to overfitting is optimized by applying the Dropout regularization method with a dropout coefficient of 0.5 for the fully connected layer.

The effectiveness of the proposed solution was demonstrated using the example of recognizing audio spectrograms of car and airplane engine sounds. The CNN was trained on a balanced dataset consisting of 7160 audio recordings. The trained network demonstrated high recognition accuracy (95%), low loss values (< 0.2), and balanced precision/recall/F-metric, demonstrating the effectiveness of the developed CNN model.

The work is devoted to the description of the author’s formal statements of the clustering problem for a given number of clusters, algorithms for their solution, as well as the results of using this toolkit in medicine.

The solution of the formulated problems by exact algorithms of implementations of even relatively low dimensions before proving optimality is impossible in a finite time due to their belonging to the NP class.

In this regard, we have proposed a hybrid algorithm that combines the advantages of precise methods based on clustering in paired distances at the initial stage with the speed of methods for solving simplified problems of splitting by cluster centers at the final stage. In the development of this direction, a sequential hybrid clustering algorithm using random search in the paradigm of swarm intelligence has been developed. The article describes it and presents the results of calculations of applied clustering problems.

To determine the effectiveness of the developed tools for optimal clustering of multidimensional objects according to a variety of heterogeneous indicators, a number of computational experiments were performed using data sets including socio-demographic, clinical anamnestic, electroencephalographic and psychometric data on the cognitive status of patients of the cardiology clinic. An experimental proof of the effectiveness of using local search algorithms in the paradigm of swarm intelligence within the framework of a hybrid algorithm for solving optimal clustering problems has been obtained.

The results of the calculations indicate the actual resolution of the main problem of using the discrete optimization apparatus — limiting the available dimensions of task implementations. We have shown that this problem is eliminated while maintaining an acceptable proximity of the clustering results to the optimal ones. The applied significance of the obtained clustering results is also due to the fact that the developed optimal clustering toolkit is supplemented by an assessment of the stability of the formed clusters, which allows for known factors (the presence of stenosis or older age) to additionally identify those patients whose cognitive resources are insufficient to overcome the influence of surgical anesthesia, as a result of which there is a unidirectional effect of postoperative deterioration of complex visual-motor reaction, attention and memory. This effect indicates the possibility of differentiating the classification of patients using the proposed tools.