This paper contributes to the goal of finding an efficient compression solution for post-training quantization from the perspective of support region choice under the framework of low-bit uniform quantization. The decision to give preference to uniform quantization comes from the fact that support region choice is the most sensitive in the uniform quantization of nonuniform sources (e.g., Laplacian sources). Therefore, in this paper, we analyse in detail how the choice of the support region influences the performance of two-bits uniform quantization, measured with signal to quantization noise ratio (SQNR), and the accuracy of the compressed neural network (NN) model. We provide experimental and theoretical results for a few significant cases of two-bits uniform quantizer design, where we assume that Laplacian source models the distribution of weights in our fully connected NN. We opt for Laplacian distribution since it models well weights of NNs. Specifically, we analyse whether it is possible to apply the simplest uniform quantization in trained NN model weight representation with a bit rate of R=2 bit/sample while preserving the accuracy of the model to a great extent. Also, our goal is to determine whether the choice of the key parameter of two-bits uniform quantizer (support region threshold) equally reflects on both, SQNR and accuracy. Moreover, we extend our analysis to the application of layer-wise two-bits uniform quantization in order to examine whether it is possible to achieve an additional improvement of the accuracy of our NN model for the MNIST dataset. We believe that the detailed analysis of post-training quantization described and conducted in this paper is very useful for all further research studies of this very current topic, especially due to the fact that the problem regarding post-training quantization is addressed from a particularly important perspective of choosing the support region.

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.

Motivated by the fact that uniform quantization is not suitable for signals having non-uniform probability density functions (pdfs), as the Laplacian pdf is, in this paper we have divided the support region of the quantizer into two disjunctive regions and utilized the simplest uniform quantization with equal bit-rates within both regions. In particular, we assumed a narrow central granular region (CGR) covering the peak of the Laplacian pdf and a wider peripheral granular region (PGR) where the pdf is predominantly tailed. We performed optimization of the widths of CGR and PGR via distortion optimization per border–clipping threshold scaling ratio which resulted in an iterative formula enabling the parametrization of our piecewise uniform quantizer (PWUQ). For medium and high bit-rates, we demonstrated the convenience of our PWUQ over the uniform quantizer, paying special attention to the case where 99.99% of the signal amplitudes belong to the support region or clipping region. We believe that the resulting formulas for PWUQ design and performance assessment are greatly beneficial in neural networks where weights and activations are typically modelled by the Laplacian distribution, and where uniform quantization is commonly used to decrease memory footprint.

In this paper, we consider the opportunities and constraints, which rest on quantization as a guiding principle for data representation and compression. In particular, we propose a novel model of Symmetric Quantile Quantizer (SQQ) and we describe in detail its parameterization. We suggest a simple method for offline precalculation of its parameters and we examine the inevitable loss of information introduced by SQQ, as an important part of bit optimization task at the traditional network level, which can be globally mapped out in many contemporary solutions. Our anticipation is that such precalculated values can be leveraged in deterministic quantization process. We highlight that this notice heavily relies on the fact that the values of interest are distributed according to the Laplacian distribution, which we consider in the paper. The basic difference of our SQQ and the previously established asymptotically optimal quantizer model, that is, Scalar Companding Quantizer (SCQ), is reflected in the fact that, in SCQ model, both decision thresholds and representation levels are determined in accordance with the specified compressor function, whereas in our SQQ model, a precedence of SCQ model for the straightforward decision thresholds calculation is used, while the representation levels are optimally determined for the specified decision thresholds and assumed Laplacian distribution. As a result, our SQQ outperforms SCQ in terms of signal-to-quantization noise ratio (SQNR). As stated in this paper, there are numerous indications to make us believe that appropriate quantizer parameterization will move us closer to an optimization in the amount of the transferred data in bits, which is strongly dependent on the amount of SQNR.

With increased network downsizing and cost minimization in deployment of neural network (NN) models, the utilization of edge computing takes a significant place in modern artificial intelligence today. To bridge the memory constraints of less-capable edge systems, a plethora of quantizer models and quantization techniques are proposed for NN compression with the goal of enabling the fitting of the quantized NN (QNN) on the edge device and guaranteeing a high extent of accuracy preservation. NN compression by means of post-training quantization has attracted a lot of research attention, where the efficiency of uniform quantizers (UQs) has been promoted and heavily exploited. In this paper, we propose two novel non-uniform quantizers (NUQs) that prudently utilize one of the two properties of the simplest UQ. Although having the same quantization rule for specifying the support region, both NUQs have a different starting setting in terms of cell width, compared to a standard UQ. The first quantizer, named the simplest power-of-two quantizer (SPTQ), defines the width of cells that are multiplied by the power of two. As it is the case in the simplest UQ design, the representation levels of SPTQ are midpoints of the quantization cells. The second quantizer, named the modified SPTQ (MSPTQ), is a more competitive quantizer model, representing an enhanced version of SPTQ in which the quantizer decision thresholds are centered between the nearest representation levels, similar to the UQ design. These properties make the novel NUQs relatively simple. Unlike UQ, the quantization cells of MSPTQ are not of equal widths and the representation levels are not midpoints of the quantization cells. In this paper, we describe the design procedure of SPTQ and MSPTQ and we perform their optimization for the assumed Laplacian source. Afterwards, we perform post-training quantization by implementing SPTQ and MSPTQ, study the viability of QNN accuracy and show the implementation benefits over the case where UQ of an equal number of quantization cells is utilized in QNN for the same classification task. We believe that both NUQs are particularly substantial for memory-constrained environments, where simple and acceptably accurate solutions are of crucial importance.

Analysis and Design of Robust Quasilogarithmic Quantizer for the Purpose of Traffic Optimization