Alexander P. Lyapin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Alexander P. Lyapin has a total of 32 co-authors. Here are the top co-authors by number of shared publications:
- Marina S. Apanovich
Professor V.F. Voino-Yasenetsky Krasnoyarsk State Medical University
7 shared publications - Konstantin V. Shadrin
Professor V.F. Voino-Yasenetsky Krasnoyarsk State Medical University, Krasnoyarsk, Russia
5 shared publications - Evgeny N. Mikhalkin
Siberian Federal University, Krasnoyarsk, Russia
3 shared publications - Svetlana S. Akhtamova
Lesosibirskij Pedagogical Institute–Branch of Siberian Federal University, 662544 Lesosibirsk, Russia
3 shared publications - Alexey Alexandrovich Kytmanov
MIREA—Russian Technological University, Moscow, Russia
2 shared publications - Evgeny Agafonov
Siberian Federal University
2 shared publications
Latest Publications
- Open Access
- 15 October 2026
- Published by MDPI AG in Computation
- Vol. 14 (10) , 253
Algorithm for the Cauchy Problem for Difference Equations with Variable Coefficients and Initial Data on a Parallelepiped
We consider the Cauchy problem for a scalar linear three-dimensional difference equation with variable coefficients and a fixed, finite rectangular stencil in a semi-infinite parallelepiped. Prescribed data occupy the complement of the unknown interior region, including the lower layers and lateral strips. We extend a layer-by-layer construction for constant-coefficient equations to coefficient functions evaluated at the lattice nodes of the corresponding function values. A symbolic block Toeplitz template is assembled once and evaluated column-wise at the coordinates of the unknowns on each layer. The resulting numerical matrices are generally not block Toeplitz. Precomputed coefficient arrays supply the values needed to assemble the right-hand sides. We verify that the assembled systems coincide with the original difference equation. If every layer matrix up to the prescribed target layer is nonsingular, the algorithm uniquely determines all unknown values on these layers, including the value at the target point. A worked example illustrates the coefficient arrays, matrix assembly, and successive solutions; exact rational verification confirms the displayed results. Storage requirements and dense-solver operation counts are also discussed.
- 15 October 2026
- Published by MDPI AG in Computation
- Vol. 14 (10) , 253
Open AccessAlgorithm for the Cauchy Problem for Difference Equations with Variable Coefficients and Initial Data on a Parallelepiped
We consider the Cauchy problem for a scalar linear three-dimensional difference equation with variable coefficients and a fixed, finite rectangular stencil in a semi-infinite parallelepiped. Prescribed data occupy the complement of the unknown interior region, including the lower layers and lateral strips. We extend a layer-by-layer construction for constant-coefficient equations to coefficient functions evaluated at the lattice nodes of the corresponding function values. A symbolic block Toeplitz template is assembled once and evaluated column-wise at the coordinates of the unknowns on each layer. The resulting numerical matrices are generally not block Toeplitz. Precomputed coefficient arrays supply the values needed to assemble the right-hand sides. We verify that the assembled systems coincide with the original difference equation. If every layer matrix up to the prescribed target layer is nonsingular, the algorithm uniquely determines all unknown values on these layers, including the value at the target point. A worked example illustrates the coefficient arrays, matrix assembly, and successive solutions; exact rational verification confirms the displayed results. Storage requirements and dense-solver operation counts are also discussed.
- Open Access
Hybrid Physics-Guided Neural Network for Vibration Sensor Nonlinearity Correction
Accelerometers built on micro-electromechanical systems (MEMS) play a critical role in structural monitoring and machinery diagnostics; however, their accuracy suffers from intrinsic nonlinearities—dead zones, hysteresis, saturation, and colored noise. Conventional physics-based correction methods are interpretable yet cannot capture complicated hysteretic behavior, while purely neural-network approaches generalize poorly and lack a physical foundation. This paper proposes a hybrid architecture that combines a residual convolutional neural network with a physics-guided low-pass filter prior, fused through an attention-gated mechanism. The CNN learns only the residual nonlinearity; the filter supplies a steady, band-limited baseline. We validate the model on two simulated scenarios—a noise-dominant track and a nonlinear-dominant track—across three random seeds. The resulting Hybrid LPF-CNN outperforms a standalone CNN by 15.2% and an LSTM by 33.5% on the severely nonlinear track, reaching a mean R2 of 0.9970 and an RMSE of 0.0179 g. On the noise-dominant track, it reaches R2 = 0.9407 and RMSE = 0.0800 g, surpassing both CNN and LSTM baselines. The model is also stable across seeds ( in R2) and gives a legible breakdown of the correction it applies. Our systematic architectural search revealed that a dual-encoder design with attention-gated fusion—where raw and filtered signals are processed separately and combined via a learnable spatial gate—provides the optimal balance between stability, accuracy, and interpretability. Even basic physics priors substantially improve the performance, stability, and interpretability of deep learning models for sensor error correction.Open AccessHybrid Physics-Guided Neural Network for Vibration Sensor Nonlinearity Correction
Accelerometers built on micro-electromechanical systems (MEMS) play a critical role in structural monitoring and machinery diagnostics; however, their accuracy suffers from intrinsic nonlinearities—dead zones, hysteresis, saturation, and colored noise. Conventional physics-based correction methods are interpretable yet cannot capture complicated hysteretic behavior, while purely neural-network approaches generalize poorly and lack a physical foundation. This paper proposes a hybrid architecture that combines a residual convolutional neural network with a physics-guided low-pass filter prior, fused through an attention-gated mechanism. The CNN learns only the residual nonlinearity; the filter supplies a steady, band-limited baseline. We validate the model on two simulated scenarios—a noise-dominant track and a nonlinear-dominant track—across three random seeds. The resulting Hybrid LPF-CNN outperforms a standalone CNN by 15.2% and an LSTM by 33.5% on the severely nonlinear track, reaching a mean R2 of 0.9970 and an RMSE of 0.0179 g. On the noise-dominant track, it reaches R2 = 0.9407 and RMSE = 0.0800 g, surpassing both CNN and LSTM baselines. The model is also stable across seeds ( in R2) and gives a legible breakdown of the correction it applies. Our systematic architectural search revealed that a dual-encoder design with attention-gated fusion—where raw and filtered signals are processed separately and combined via a learnable spatial gate—provides the optimal balance between stability, accuracy, and interpretability. Even basic physics priors substantially improve the performance, stability, and interpretability of deep learning models for sensor error correction. - 2 July 2026
- Published by Springer Nature
- p. 19-38
Discrete Generating Series and Linear Difference Equations
- 2 July 2026
- Published by Springer Nature
- p. 19-38
Discrete Generating Series and Linear Difference Equations
- 1 January 2026
- Published by Irkutsk State University in The Bulletin of Irkutsk State University. Series Mathematics
- Vol. 55 (55) , 159-169
Evgeniy Konstantinovich Leinartas (1954–2024)
The article is devoted to the scientific and pedagogical activity of Doctor of Physical and Mathematical Sciences, Professor of the Department of Function Theory at Siberian Federal University Evgeniy Konstantinovich Leinartas (1954–2024).
- 1 January 2026
- Published by Irkutsk State University in The Bulletin of Irkutsk State University. Series Mathematics
- Vol. 55 (55) , 159-169
Evgeniy Konstantinovich Leinartas (1954–2024)
The article is devoted to the scientific and pedagogical activity of Doctor of Physical and Mathematical Sciences, Professor of the Department of Function Theory at Siberian Federal University Evgeniy Konstantinovich Leinartas (1954–2024).
- Open Access
- 11 October 2025
- Published by MDPI AG in Mathematics
- Vol. 13 (20) , 3259
D-Finite Discrete Generating Series and Their Sections
This paper investigates D-finite discrete generating series and their sections. The concept of D-finiteness is extended to multidimensional discrete generating series and its equivalence to p-recursive sequences is rigorously established. It is further shown that sections of the D-finite series preserve D-finiteness, and that their generating functions satisfy systems of linear difference equations with polynomial coefficients. In the two-dimensional case, explicit difference relations are derived that connect section values with boundary data, while in higher dimensions general constructive methods are developed for obtaining such relations, including cases with variable coefficients. Several worked examples illustrate how the theory applies to solving difference equations and analyzing multidimensional recurrent sequences. The results provide a unified framework linking generating functions and recurrence relations, with applications in combinatorics, number theory, symbolic summation, and the theory of discrete recursive filters in signal processing.
- 11 October 2025
- Published by MDPI AG in Mathematics
- Vol. 13 (20) , 3259
Open AccessD-Finite Discrete Generating Series and Their Sections
This paper investigates D-finite discrete generating series and their sections. The concept of D-finiteness is extended to multidimensional discrete generating series and its equivalence to p-recursive sequences is rigorously established. It is further shown that sections of the D-finite series preserve D-finiteness, and that their generating functions satisfy systems of linear difference equations with polynomial coefficients. In the two-dimensional case, explicit difference relations are derived that connect section values with boundary data, while in higher dimensions general constructive methods are developed for obtaining such relations, including cases with variable coefficients. Several worked examples illustrate how the theory applies to solving difference equations and analyzing multidimensional recurrent sequences. The results provide a unified framework linking generating functions and recurrence relations, with applications in combinatorics, number theory, symbolic summation, and the theory of discrete recursive filters in signal processing.
- Open Access
- 16 June 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (12) , 6768
A Combined Method for Localizing Two Overlapping Acoustic Sources Based on Deep Learning
Deep learning approaches for multi-source sound localization face significant challenges, particularly the need for extensive training datasets encompassing diverse spatial configurations to achieve robust generalization. This requirement leads to substantial computational demands, which are further exacerbated when localizing overlapping sources in complex acoustic environments with reverberation and noise. In this paper, a new methodology is proposed for simultaneous localization of two overlapping sound sources in the time–frequency domain in a closed, reverberant environment with a spatial resolution of using a small-sized microphone array. The proposed methodology is based on the integration of the sound source separation method with a single-source sound localization model. A hybrid model was proposed to separate the sound source signals received by each microphone in the array. The model was built using a bidirectional long short-term memory (BLSTM) network and trained on a dataset using the ideal binary mask (IBM) as the training target. The modeling results show that the proposed localization methodology is efficient in determining the directions for two overlapping sources simultaneously, with an average localization accuracy of 86.1% for the test dataset containing short-term signals of 500 ms duration with different signal-to-signal ratio values.- 16 June 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (12) , 6768
Open AccessA Combined Method for Localizing Two Overlapping Acoustic Sources Based on Deep Learning
Deep learning approaches for multi-source sound localization face significant challenges, particularly the need for extensive training datasets encompassing diverse spatial configurations to achieve robust generalization. This requirement leads to substantial computational demands, which are further exacerbated when localizing overlapping sources in complex acoustic environments with reverberation and noise. In this paper, a new methodology is proposed for simultaneous localization of two overlapping sound sources in the time–frequency domain in a closed, reverberant environment with a spatial resolution of using a small-sized microphone array. The proposed methodology is based on the integration of the sound source separation method with a single-source sound localization model. A hybrid model was proposed to separate the sound source signals received by each microphone in the array. The model was built using a bidirectional long short-term memory (BLSTM) network and trained on a dataset using the ideal binary mask (IBM) as the training target. The modeling results show that the proposed localization methodology is efficient in determining the directions for two overlapping sources simultaneously, with an average localization accuracy of 86.1% for the test dataset containing short-term signals of 500 ms duration with different signal-to-signal ratio values. - 25 March 2025
- Published by Pleiades Publishing Ltd in Programming and Computer Software
- Vol. 51 (2) , 88-92
Algorithm for Finding Singular Points of a General Algebraic Hypersurface
In this paper, an algorithm for calculating the parameterization of the Horn–Kapranov \(A\)-discriminant set and singular points of an algebraic hypersurface, using the Maple computer algebra system, is proposed.- 25 March 2025
- Published by Pleiades Publishing Ltd in Programming and Computer Software
- Vol. 51 (2) , 88-92
Algorithm for Finding Singular Points of a General Algebraic Hypersurface
In this paper, an algorithm for calculating the parameterization of the Horn–Kapranov \(A\)-discriminant set and singular points of an algebraic hypersurface, using the Maple computer algebra system, is proposed.- Open Access
Selection of Support System to Provide Vibration Frequency and Stability of Beam Structure
The current engineering theories on bending vibrations and the stability of beam structures are based on solving eigenvalue problems through similarly formulated differential equations. Solving the eigenvalue problem for engineering calculations is particularly laborious, especially for non-classical supports, where factors like the stiffness of supports, axial forces, or temperature must be considered. In this case, the solution can be obtained only by numerical methods using specially created programs, which makes it difficult to select supports for a given planar beam structure in engineering practice. This work utilizes established solutions from eigenvalue problems in the theory of vibrations and stability of beams, incorporating factors such as axial forces, temperature, and support stiffness. This combined solution is applicable to beam structures of any type and cross-section, as it is determined solely by the selected support conditions (stiffness) and loading (axial force, temperature). Approximation of eigenvalue problem solutions through continuous functions allows the readers to use them for the analytical solution of the design problem of choosing a support system to ensure the frequency of vibrations and stability of the planar beam structure. At the same time, the known solutions given in the reference books on bending vibrations and stability become their particular solutions. This approach is applicable to solving problems of vibrations and loss of stability of various types (torsional, longitudinal, etc.), and is also applicable in other disciplines where solving problems for eigenvalues is required.
Open AccessSelection of Support System to Provide Vibration Frequency and Stability of Beam Structure
The current engineering theories on bending vibrations and the stability of beam structures are based on solving eigenvalue problems through similarly formulated differential equations. Solving the eigenvalue problem for engineering calculations is particularly laborious, especially for non-classical supports, where factors like the stiffness of supports, axial forces, or temperature must be considered. In this case, the solution can be obtained only by numerical methods using specially created programs, which makes it difficult to select supports for a given planar beam structure in engineering practice. This work utilizes established solutions from eigenvalue problems in the theory of vibrations and stability of beams, incorporating factors such as axial forces, temperature, and support stiffness. This combined solution is applicable to beam structures of any type and cross-section, as it is determined solely by the selected support conditions (stiffness) and loading (axial force, temperature). Approximation of eigenvalue problem solutions through continuous functions allows the readers to use them for the analytical solution of the design problem of choosing a support system to ensure the frequency of vibrations and stability of the planar beam structure. At the same time, the known solutions given in the reference books on bending vibrations and stability become their particular solutions. This approach is applicable to solving problems of vibrations and loss of stability of various types (torsional, longitudinal, etc.), and is also applicable in other disciplines where solving problems for eigenvalues is required.
- 15 April 2024
- Published by The Russian Academy of Sciences in Программирование
- No. 2,p. 79-83
Applying computer algebra systems to study Chaundy-Bullard identities for the vector partition function with weight
An algorithm for obtaining the Chaundy-Bullard identity for a vector partition function with weight that uses computer algebra methods is proposed. To automate this process in Maple, an algorithm was developed and implemented that calculates the values of the vector partition function with weight by finding non-negative solutions of systems of linear Diophantine equations that are used to form the identities involved. The algorithm’s input data is represented by the set of integer vectors that form a pointed lattice cone and by some point from this cone, and the Chaundy-Bullard identity for the vector partition function with weight is its output. The code involved is stored in the depository and is ready-to-use. An example demonstrating the algorithm’s operation is given.
- 15 April 2024
- Published by The Russian Academy of Sciences in Программирование
- No. 2,p. 79-83
Applying computer algebra systems to study Chaundy-Bullard identities for the vector partition function with weight
An algorithm for obtaining the Chaundy-Bullard identity for a vector partition function with weight that uses computer algebra methods is proposed. To automate this process in Maple, an algorithm was developed and implemented that calculates the values of the vector partition function with weight by finding non-negative solutions of systems of linear Diophantine equations that are used to form the identities involved. The algorithm’s input data is represented by the set of integer vectors that form a pointed lattice cone and by some point from this cone, and the Chaundy-Bullard identity for the vector partition function with weight is its output. The code involved is stored in the depository and is ready-to-use. An example demonstrating the algorithm’s operation is given.
- 1 April 2024
- Published by Pleiades Publishing Ltd in Programming and Computer Software
- Vol. 50 (2) , 176-179
Applying Computer Algebra Systems to Study Chaundy-Bullard Identities for the Vector Partition Function with Weight
An algorithm for obtaining the Chaundy-Bullard identity for a vector partition function with weight that uses computer algebra methods is proposed. To automate this process in Maple, an algorithm was developed and implemented that calculates the values of the vector partition function with weight by finding non-negative solutions of systems of linear Diophantine equations that are used to form the identities involved. The algorithm’s input data is represented by the set of integer vectors that form a pointed lattice cone and by some point from this cone, and the Chaundy-Bullard identity for the vector partition function with weight is its output. The code involved is stored in the depository and is ready-to-use. An example demonstrating the algorithm’s operation is given.
- 1 April 2024
- Published by Pleiades Publishing Ltd in Programming and Computer Software
- Vol. 50 (2) , 176-179
Applying Computer Algebra Systems to Study Chaundy-Bullard Identities for the Vector Partition Function with Weight
An algorithm for obtaining the Chaundy-Bullard identity for a vector partition function with weight that uses computer algebra methods is proposed. To automate this process in Maple, an algorithm was developed and implemented that calculates the values of the vector partition function with weight by finding non-negative solutions of systems of linear Diophantine equations that are used to form the identities involved. The algorithm’s input data is represented by the set of integer vectors that form a pointed lattice cone and by some point from this cone, and the Chaundy-Bullard identity for the vector partition function with weight is its output. The code involved is stored in the depository and is ready-to-use. An example demonstrating the algorithm’s operation is given.