Marta Makowska
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Marta Makowska has a total of 3 co-authors. Here are the top co-authors by number of shared publications:
- Aleksandr Cariow
Faculty of Computer Science and Information Technology, West Pomeranian University of Technology in Szczecin, Żołnierska 49, 71-210 Szczecin, Poland
2 shared publications - Janusz Piotr Paplinski
West Pomeranian University of Technology in Szczecin
2 shared publications - Paweł Strzelec
Faculty of Computer Science and Information Technology, West Pomeranian University of Technology, Żołnierska 49, 71-210 Szczecin, Poland
1 shared publication
Latest Publications
- Open Access
Digital Signal Processing (DSP)-Oriented Reduced-Complexity Algorithms for Calculating Matrix–Vector Products with Small-Order Toeplitz Matrices
Toeplitz matrix–vector products are used in many digital signal processing applications. Direct methods for calculating such products require multiplications and additions, where N denotes the order of the Toeplitz matrix. In the case of large matrices, this operation becomes especially time intensive. However, matrix–vector products with small-order Toeplitz matrices are of particular interest because small matrices often serve as kernels in modern digital signal processing algorithms. Perhaps reducing the number of arithmetic operations when calculating matrix–vector products in the case of small Toeplitz matrices gives less effect than of large ones, but this problem exists, and it needs to be solved. The traditional way to calculate such products is to use the fast Fourier transform algorithm. However, in the case of small-order matrices, it is advisable to use direct factorization of Toeplitz matrices, which leads to a reduction in arithmetic complexity. In this paper, we propose a set of reduced-complexity algorithms for calculating matrix–vector products with Toeplitz matrices of order . The main emphasis will be on reducing multiplicative complexity since multiplication in most cases is more time-consuming than addition. This paper also provides assessments of the implementation of the developed algorithms on FPGAs.Open AccessDigital Signal Processing (DSP)-Oriented Reduced-Complexity Algorithms for Calculating Matrix–Vector Products with Small-Order Toeplitz Matrices
Toeplitz matrix–vector products are used in many digital signal processing applications. Direct methods for calculating such products require multiplications and additions, where N denotes the order of the Toeplitz matrix. In the case of large matrices, this operation becomes especially time intensive. However, matrix–vector products with small-order Toeplitz matrices are of particular interest because small matrices often serve as kernels in modern digital signal processing algorithms. Perhaps reducing the number of arithmetic operations when calculating matrix–vector products in the case of small Toeplitz matrices gives less effect than of large ones, but this problem exists, and it needs to be solved. The traditional way to calculate such products is to use the fast Fourier transform algorithm. However, in the case of small-order matrices, it is advisable to use direct factorization of Toeplitz matrices, which leads to a reduction in arithmetic complexity. In this paper, we propose a set of reduced-complexity algorithms for calculating matrix–vector products with Toeplitz matrices of order . The main emphasis will be on reducing multiplicative complexity since multiplication in most cases is more time-consuming than addition. This paper also provides assessments of the implementation of the developed algorithms on FPGAs. - Open Access
- 6 August 2023
- Published by MDPI AG in Applied Sciences
- Vol. 13 (15) , 9004
VLSI-Friendly Filtering Algorithms for Deep Neural Networks
The paper introduces a range of efficient algorithmic solutions for implementing the fundamental filtering operation in convolutional layers of convolutional neural networks on fully parallel hardware. Specifically, these operations involve computing M inner products between neighbouring vectors generated by a sliding time window from the input data stream and an M-tap finite impulse response filter. By leveraging the factorisation of the Hankel matrix, we have successfully reduced the multiplicative complexity of the matrix-vector product calculation. This approach has been applied to develop fully parallel and resource-efficient algorithms for M values of 3, 5, 7, and 9. The fully parallel hardware implementation of our proposed algorithms achieves approximately a 30% reduction in embedded multipliers compared to the naive calculation methods.
- 6 August 2023
- Published by MDPI AG in Applied Sciences
- Vol. 13 (15) , 9004
Open AccessVLSI-Friendly Filtering Algorithms for Deep Neural Networks
The paper introduces a range of efficient algorithmic solutions for implementing the fundamental filtering operation in convolutional layers of convolutional neural networks on fully parallel hardware. Specifically, these operations involve computing M inner products between neighbouring vectors generated by a sliding time window from the input data stream and an M-tap finite impulse response filter. By leveraging the factorisation of the Hankel matrix, we have successfully reduced the multiplicative complexity of the matrix-vector product calculation. This approach has been applied to develop fully parallel and resource-efficient algorithms for M values of 3, 5, 7, and 9. The fully parallel hardware implementation of our proposed algorithms achieves approximately a 30% reduction in embedded multipliers compared to the naive calculation methods.