Aleksandra Jovanović
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Aleksandra Jovanović has a total of 33 co-authors. Here are the top co-authors by number of shared publications:
- Zoran Perić
Faculty of Electronic Engineering, University of Niš, Aleksandra Medvedeva 4, 18000 Niš, Serbia
12 shared publications - Jelena Nikolić
University of Nis
8 shared publications - Stefan Tomić
Abu Dhabi Polytechnic
4 shared publications - Riccardo Zese
Dip. di Scienze Chimiche Farmaceutiche e Agrarie Università di Ferrara Via Luigi Borsari 46 44121 Ferrara Italy
3 shared publications - Sofija Perić
Faculty of Electronic Engineering, University of Niš, Aleksandra Medvedeva 4, 18000 Niš, Serbia
3 shared publications - Danijela Aleksić
Department of Mobile Network Nis, Telekom Srbija, Vozdova 11, 18000 Nis, Serbia
2 shared publications
Latest Publications
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 121961-121977
Bias-Aware FP8 Format for Efficient Weight Representation in Post-Training Quantization
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 121961-121977
Open AccessBias-Aware FP8 Format for Efficient Weight Representation in Post-Training Quantization
- 13 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
SQNR Analysis of Power-of-Two Level Quantizer and Integer 3 Quantizer
In this paper, we present a detailed comparative analysis of two 3-bit quantizer models suitable for low bit-width quantization: the Power-of-Two Level Quantizer (PoTLQ) and the Integer 3 Quantizer (INT3). PoTLQ is a symmetric, non-uniform quantizer whose positive representation levels double successively, enabling a compact and hardware-friendly implementation. INT3, on the other hand, is a symmetric, uniform quantizer with integer-valued representation levels. While both quantizers offer simplicity in design and ease of implementation, the robustness of their performance to changes in input data variance remains insufficiently explored. In this paper, we therefore conduct a theoretical and numerical analysis of the performance of both quantizer models by comparing their signal-to-quantization noise ratio (SQNR) behavior under varying input variances. The analysis provides insights into achieving a trade-off between non-uniform and uniform quantization in low bit-width scenarios, offering guidance on quantizer selection based on input variance dynamic.
- 13 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
SQNR Analysis of Power-of-Two Level Quantizer and Integer 3 Quantizer
In this paper, we present a detailed comparative analysis of two 3-bit quantizer models suitable for low bit-width quantization: the Power-of-Two Level Quantizer (PoTLQ) and the Integer 3 Quantizer (INT3). PoTLQ is a symmetric, non-uniform quantizer whose positive representation levels double successively, enabling a compact and hardware-friendly implementation. INT3, on the other hand, is a symmetric, uniform quantizer with integer-valued representation levels. While both quantizers offer simplicity in design and ease of implementation, the robustness of their performance to changes in input data variance remains insufficiently explored. In this paper, we therefore conduct a theoretical and numerical analysis of the performance of both quantizer models by comparing their signal-to-quantization noise ratio (SQNR) behavior under varying input variances. The analysis provides insights into achieving a trade-off between non-uniform and uniform quantization in low bit-width scenarios, offering guidance on quantizer selection based on input variance dynamic.
- 13 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Golomb – Rice Coding for Efficient Compression of 8-Bit Floating- Point Data
Recent advancements in neural network (NN) hardware and software have introduced 8-bit Floating-Point (FP8) format as promising candidate for reducing memory usage and computational overhead. This article investigates the potential for further compressing FP8) data by applying Golomb-Rice (GR) coding. In the FP8 format, the 8 bits are allocated to represent the sign, exponent and mantissa using natural binary encoding. To achieve efficient compression, the GR code, a variable-length, lossless coding scheme, is applied exclusively to the exponent bits. In this way, data storage savings could be achieved, which can be very significant when the large amount of data has to be stored, such as storing NNs.
- 13 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Golomb – Rice Coding for Efficient Compression of 8-Bit Floating- Point Data
Recent advancements in neural network (NN) hardware and software have introduced 8-bit Floating-Point (FP8) format as promising candidate for reducing memory usage and computational overhead. This article investigates the potential for further compressing FP8) data by applying Golomb-Rice (GR) coding. In the FP8 format, the 8 bits are allocated to represent the sign, exponent and mantissa using natural binary encoding. To achieve efficient compression, the GR code, a variable-length, lossless coding scheme, is applied exclusively to the exponent bits. In this way, data storage savings could be achieved, which can be very significant when the large amount of data has to be stored, such as storing NNs.
- 13 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Statistical Analysis of Synaptic Weights in Spiking Neural Network Trained on the DVS128 Gesture Dataset
The application of compression techniques, such as quantization and pruning, to reduce neural network weight size and complexity is gaining momentum due to the significant savings they provide in both hardware and software resources, particularly important for edge devices with limited capabilities. To fully leverage the advantages of these techniques, their design should account for the amplitude dynamics and the probability distribution of the parameters to be compressed. Therefore, it is essential to analyze the statistical characteristics of these parameters prior to compression. Although this type of analysis has been extensively conducted in the context of Artificial Neural Networks (ANNs), it has received limited attention in the case of Spiking Neural Networks (SNNs). Motivated by this gap, this paper examines the statistical properties of the weights of an SNN trained on the DVS128 Gesture dataset. The results of this analysis are valuable not only for enabling a well-founded selection and effective design of appropriate quantization models, but also for supporting the efficient application of pruning techniques, which could benefit from using synaptic weight histograms as valuable guiding information.
- 13 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Statistical Analysis of Synaptic Weights in Spiking Neural Network Trained on the DVS128 Gesture Dataset
The application of compression techniques, such as quantization and pruning, to reduce neural network weight size and complexity is gaining momentum due to the significant savings they provide in both hardware and software resources, particularly important for edge devices with limited capabilities. To fully leverage the advantages of these techniques, their design should account for the amplitude dynamics and the probability distribution of the parameters to be compressed. Therefore, it is essential to analyze the statistical characteristics of these parameters prior to compression. Although this type of analysis has been extensively conducted in the context of Artificial Neural Networks (ANNs), it has received limited attention in the case of Spiking Neural Networks (SNNs). Motivated by this gap, this paper examines the statistical properties of the weights of an SNN trained on the DVS128 Gesture dataset. The results of this analysis are valuable not only for enabling a well-founded selection and effective design of appropriate quantization models, but also for supporting the efficient application of pruning techniques, which could benefit from using synaptic weight histograms as valuable guiding information.
- Open Access
- 24 February 2025
- Published by Kaunas University of Technology (KTU) in Elektronika ir Elektrotechnika
- Vol. 31 (1) , 56-61
Performance Analysis of Two 8-Bit Floating-Point-based Piecewise Uniform Quantizers for a Laplacian Data Source
In this paper, we employ the analogy between the representation of the floating-point (FP) format and the representation level distribution of the piecewise uniform quantizer (PWUQ) to assess the performance of FP-based solutions more thoroughly. We present theoretical derivations to assess the performance of the FP format and the PWUQ determined by this format for input data from the Laplacian source. We also provide a performance comparison of two selected 8-bit FP-based PWUQs. Beyond the typical evaluation of the applied FP format, through the accuracy degradation caused by the application of both FP8 solutions in neural network compression, we also use objective quantization measures. This approach offers insights into the robustness of these 8-bit FP-based solutions with respect to changes in input variance, which can be important when the input variance changes. The results demonstrate that the allocation of bits to encode the exponent and mantissa in the FP8 format is important, as it can significantly impact overall performance.
- 24 February 2025
- Published by Kaunas University of Technology (KTU) in Elektronika ir Elektrotechnika
- Vol. 31 (1) , 56-61
Open AccessPerformance Analysis of Two 8-Bit Floating-Point-based Piecewise Uniform Quantizers for a Laplacian Data Source
In this paper, we employ the analogy between the representation of the floating-point (FP) format and the representation level distribution of the piecewise uniform quantizer (PWUQ) to assess the performance of FP-based solutions more thoroughly. We present theoretical derivations to assess the performance of the FP format and the PWUQ determined by this format for input data from the Laplacian source. We also provide a performance comparison of two selected 8-bit FP-based PWUQs. Beyond the typical evaluation of the applied FP format, through the accuracy degradation caused by the application of both FP8 solutions in neural network compression, we also use objective quantization measures. This approach offers insights into the robustness of these 8-bit FP-based solutions with respect to changes in input variance, which can be important when the input variance changes. The results demonstrate that the allocation of bits to encode the exponent and mantissa in the FP8 format is important, as it can significantly impact overall performance.
- Open Access
- 1 December 2024
- Published by Walter de Gruyter GmbH in Journal of Electrical Engineering
- Vol. 75 (6) , 425-434
Accuracy degradation aware bit rate allocation for layer-wise uniform quantization of weights in neural network
Motivated by the fact that optimizing quantization error fails to minimize the accuracy degradation caused by the quantization of Neural Network (NN) weights, in this paper, we highlight the need for a comprehensive analysis of post-training quantization, covering average bit rate, accuracy degradation and SQNR. One such analysis, for the layer-wise uniform quantization and its application is NN weight compression is presented in this paper. We introduce an additional aspect of Uniform Quantizer (UQ) choice, and we allow the selection of one of two UQs (2-bit or 3-bit UQs) for each layer of NN. However, choosing a particular bit rate allocation to minimize average bit rate and accuracy degradation requires finding a trade-off solution. To solve the issue underpinning our motivation, for a post-training layer-wise uniform quantization framework with two UQs, we propose an algorithm for generating bit rate allocation patterns and corresponding pairs of average bit rate and accuracy degradation. We analyze the output of this algorithm and determine a solution of the elaborated two-objective optimization problem. To confirm the validity of the results we also apply the Pareto dominance method. This work can be considered a solid foundation for further analyses of more complex and deeper NNs.
- 1 December 2024
- Published by Walter de Gruyter GmbH in Journal of Electrical Engineering
- Vol. 75 (6) , 425-434
Open AccessAccuracy degradation aware bit rate allocation for layer-wise uniform quantization of weights in neural network
Motivated by the fact that optimizing quantization error fails to minimize the accuracy degradation caused by the quantization of Neural Network (NN) weights, in this paper, we highlight the need for a comprehensive analysis of post-training quantization, covering average bit rate, accuracy degradation and SQNR. One such analysis, for the layer-wise uniform quantization and its application is NN weight compression is presented in this paper. We introduce an additional aspect of Uniform Quantizer (UQ) choice, and we allow the selection of one of two UQs (2-bit or 3-bit UQs) for each layer of NN. However, choosing a particular bit rate allocation to minimize average bit rate and accuracy degradation requires finding a trade-off solution. To solve the issue underpinning our motivation, for a post-training layer-wise uniform quantization framework with two UQs, we propose an algorithm for generating bit rate allocation patterns and corresponding pairs of average bit rate and accuracy degradation. We analyze the output of this algorithm and determine a solution of the elaborated two-objective optimization problem. To confirm the validity of the results we also apply the Pareto dominance method. This work can be considered a solid foundation for further analyses of more complex and deeper NNs.
- 1 January 2024
- Published by University Library in Kragujevac
- p. 47-50
Performance Analysis of Two-stage Uniform Quantization of Laplacian Data
Residual quantization decomposes the quantization process into smaller ones, achieving a very small quantization error with significantly reduced computational costs. In this paper, we explore the application of the residual concept in two-stage uniform quantization (TSUQ). TSUQ employs two uniform quantizers, one in the first stage to quantize the value of the input data using R1 bits and the other in the second stage to quantize the residual error of the previous stage using R2 bits. We analyze the TSUQ of Laplacian data, covering the following aspects: i) we determine the signal to quantization noise ratio (SQNR) for different (R1, R2) pairs, ii) we estimate the gain in SQNR achieved by quantization of residual error, and iii) we examine the robustness of SQNR to changes in the variance of the data being quantized. The conducted analysis can be useful in evaluating the performance of a particular TSUQ. Moreover, it can be useful in designing a TSUQ by enabling the determination of R1 and R2 that provide the highest SQNR.
- 1 January 2024
- Published by University Library in Kragujevac
- p. 47-50
Performance Analysis of Two-stage Uniform Quantization of Laplacian Data
Residual quantization decomposes the quantization process into smaller ones, achieving a very small quantization error with significantly reduced computational costs. In this paper, we explore the application of the residual concept in two-stage uniform quantization (TSUQ). TSUQ employs two uniform quantizers, one in the first stage to quantize the value of the input data using R1 bits and the other in the second stage to quantize the residual error of the previous stage using R2 bits. We analyze the TSUQ of Laplacian data, covering the following aspects: i) we determine the signal to quantization noise ratio (SQNR) for different (R1, R2) pairs, ii) we estimate the gain in SQNR achieved by quantization of residual error, and iii) we examine the robustness of SQNR to changes in the variance of the data being quantized. The conducted analysis can be useful in evaluating the performance of a particular TSUQ. Moreover, it can be useful in designing a TSUQ by enabling the determination of R1 and R2 that provide the highest SQNR.
- Open Access
- 12 December 2022
- Published by Kaunas University of Technology (KTU) in Information Technology And Control
- Vol. 51 (4) , 625-637
Performance Analysis of a 2-bit Dual-Mode Uniform Scalar Quantizer for Laplacian Source
The main issue when dealing with the non-adaptive scalar quantizers is their sensitivity to variance-mismatch, the effect that occurs when the data variance differs from the one used for the quantizer design. In this paper, we consider the influence of that effect in low-rate (2-bit) uniform scalar quantization (USQ) of Laplacian source and also we propose adequate measure to suppress it. Particularly, the approach we propose represents the upgraded version of the previous approaches used to improve performance of the single quantizer. It is based on dual-mode quantization that combines two 2-bit USQs (with adequately chosen parameters) to process input data, selected by applying the special rule. Analysis conducted in theoretical domain has shown that the proposed approach is less sensitive to variance-mismatch, making the dual-mode USQ more efficient in terms of robustness than the single USQ. Also, a gain is achieved compared to other 2-bit quantizer solutions. Experimental results are also provided for quantization of weights of the multi-layer perceptron (MLP) neural network, where good matching with the theoretical results is observed. Due to these achievements, we believe that the solution we propose can be a good choice for compression of non-stationary data modeled by Laplacian distribution, such as neural network parameters.
- 12 December 2022
- Published by Kaunas University of Technology (KTU) in Information Technology And Control
- Vol. 51 (4) , 625-637
Open AccessPerformance Analysis of a 2-bit Dual-Mode Uniform Scalar Quantizer for Laplacian Source
The main issue when dealing with the non-adaptive scalar quantizers is their sensitivity to variance-mismatch, the effect that occurs when the data variance differs from the one used for the quantizer design. In this paper, we consider the influence of that effect in low-rate (2-bit) uniform scalar quantization (USQ) of Laplacian source and also we propose adequate measure to suppress it. Particularly, the approach we propose represents the upgraded version of the previous approaches used to improve performance of the single quantizer. It is based on dual-mode quantization that combines two 2-bit USQs (with adequately chosen parameters) to process input data, selected by applying the special rule. Analysis conducted in theoretical domain has shown that the proposed approach is less sensitive to variance-mismatch, making the dual-mode USQ more efficient in terms of robustness than the single USQ. Also, a gain is achieved compared to other 2-bit quantizer solutions. Experimental results are also provided for quantization of weights of the multi-layer perceptron (MLP) neural network, where good matching with the theoretical results is observed. Due to these achievements, we believe that the solution we propose can be a good choice for compression of non-stationary data modeled by Laplacian distribution, such as neural network parameters.
- Open Access
- 1 October 2022
- Published by MDPI AG in Mathematics
- Vol. 10 (19) , 3590
Two Interval Upper-Bound Q-Function Approximations with Applications
The Gaussian Q-function has considerable applications in numerous areas of science and engineering. However, the fact that a closed-form expression for this function does not exist encourages finding approximations or bounds of the Q-function. In this paper, we determine analytically two novel interval upper bound Q-function approximations and show that they could be used efficiently not only for the symbol error probability (SEP) estimation of transmission over Nakagami-m fading channels, but also for the average symbol error probability (ASEP) evaluation for two modulation formats. Specifically, we determine analytically the composition of the upper bound Q-function approximations specified at disjoint intervals of the input argument values so as to provide the highest accuracy within the intervals, by utilizing the selected one of two upper bound Q-function approximations. We show that a further increase of the accuracy, achieved in the case with two upper-bound approximations composing the interval approximation, can be obtained by forming a composite interval approximation of the Q-function that assumes another extra interval and by specifying the third form for the upper-bound Q-function approximation. The proposed analytical approach can be considered universal and widely applicable. The results presented in the paper indicate that the proposed Q-function approximations outperform in terms of accuracy other well-known approximations carefully chosen for comparison purposes. This approximation can be used in numerous theoretical communication problems based on the Q-function calculation. In this paper, we apply it to estimate the average bit error rate (ABER), when the transmission in a Nakagami-m fading channel is observed for the assumed binary phase-shift keying (BPSK) and differentially encoded quadrature phase-shift keying (DE-QPSK) modulation formats, as well as to design scalar quantization with equiprobable cells for variables from a Gaussian source.
- 1 October 2022
- Published by MDPI AG in Mathematics
- Vol. 10 (19) , 3590
Open AccessTwo Interval Upper-Bound Q-Function Approximations with Applications
The Gaussian Q-function has considerable applications in numerous areas of science and engineering. However, the fact that a closed-form expression for this function does not exist encourages finding approximations or bounds of the Q-function. In this paper, we determine analytically two novel interval upper bound Q-function approximations and show that they could be used efficiently not only for the symbol error probability (SEP) estimation of transmission over Nakagami-m fading channels, but also for the average symbol error probability (ASEP) evaluation for two modulation formats. Specifically, we determine analytically the composition of the upper bound Q-function approximations specified at disjoint intervals of the input argument values so as to provide the highest accuracy within the intervals, by utilizing the selected one of two upper bound Q-function approximations. We show that a further increase of the accuracy, achieved in the case with two upper-bound approximations composing the interval approximation, can be obtained by forming a composite interval approximation of the Q-function that assumes another extra interval and by specifying the third form for the upper-bound Q-function approximation. The proposed analytical approach can be considered universal and widely applicable. The results presented in the paper indicate that the proposed Q-function approximations outperform in terms of accuracy other well-known approximations carefully chosen for comparison purposes. This approximation can be used in numerous theoretical communication problems based on the Q-function calculation. In this paper, we apply it to estimate the average bit error rate (ABER), when the transmission in a Nakagami-m fading channel is observed for the assumed binary phase-shift keying (BPSK) and differentially encoded quadrature phase-shift keying (DE-QPSK) modulation formats, as well as to design scalar quantization with equiprobable cells for variables from a Gaussian source.
- Open Access
Whether the Support Region of Three-Bit Uniform Quantizer Has a Strong Impact on Post-Training Quantization for MNIST Dataset?
Driven by the need for the compression of weights in neural networks (NNs), which is especially beneficial for edge devices with a constrained resource, and by the need to utilize the simplest possible quantization model, in this paper, we study the performance of three-bit post-training uniform quantization. The goal is to put various choices of the key parameter of the quantizer in question (support region threshold) in one place and provide a detailed overview of this choice’s impact on the performance of post-training quantization for the MNIST dataset. Specifically, we analyze whether it is possible to preserve the accuracy of the two NN models (MLP and CNN) to a great extent with the very simple three-bit uniform quantizer, regardless of the choice of the key parameter. Moreover, our goal is to answer the question of whether it is of the utmost importance in post-training three-bit uniform quantization, as it is in quantization, to determine the optimal support region threshold value of the quantizer to achieve some predefined accuracy of the quantized neural network (QNN). The results show that the choice of the support region threshold value of the three-bit uniform quantizer does not have such a strong impact on the accuracy of the QNNs, which is not the case with two-bit uniform post-training quantization, when applied in MLP for the same classification task. Accordingly, one can anticipate that due to this special property, the post-training quantization model in question can be greatly exploited.
Open AccessWhether the Support Region of Three-Bit Uniform Quantizer Has a Strong Impact on Post-Training Quantization for MNIST Dataset?
Driven by the need for the compression of weights in neural networks (NNs), which is especially beneficial for edge devices with a constrained resource, and by the need to utilize the simplest possible quantization model, in this paper, we study the performance of three-bit post-training uniform quantization. The goal is to put various choices of the key parameter of the quantizer in question (support region threshold) in one place and provide a detailed overview of this choice’s impact on the performance of post-training quantization for the MNIST dataset. Specifically, we analyze whether it is possible to preserve the accuracy of the two NN models (MLP and CNN) to a great extent with the very simple three-bit uniform quantizer, regardless of the choice of the key parameter. Moreover, our goal is to answer the question of whether it is of the utmost importance in post-training three-bit uniform quantization, as it is in quantization, to determine the optimal support region threshold value of the quantizer to achieve some predefined accuracy of the quantized neural network (QNN). The results show that the choice of the support region threshold value of the three-bit uniform quantizer does not have such a strong impact on the accuracy of the QNNs, which is not the case with two-bit uniform post-training quantization, when applied in MLP for the same classification task. Accordingly, one can anticipate that due to this special property, the post-training quantization model in question can be greatly exploited.