Adaptive Quantization Research Articles

A Method of Adaptive Quantization for the Coefficients of a Discrete Wavelet Transformation

A Method of Adaptive Quantization for the Coefficients of a Discrete Wavelet Transformation

An Energy-and-Area-Efficient CNN Accelerator for Universal Powers-of-Two Quantization

CNN model computation on edge devices is tightly restricted to the limited resource and power budgets, which motivates the low-bit quantization technology to compress CNN models into 4-bit or lower format to reduce the model size and increase hardware efficiency. Most current low-bit quantization methods use uniform quantization that maps weight and activation values onto evenly-distributed levels, which usually results in accuracy loss due to distribution mismatch. Meanwhile, some non-uniform quantization methods propose specialized representation that can better match various distribution shapes but are usually difficult to be efficiently accelerated on hardware. In order to achieve low-bit quantization with high accuracy and hardware efficiency, this paper proposes Universal Power-of-Two (UPoT), a novel low-bit quantization method that represents values as the addition of multiple power-of-two values selected from a series of subsets. By updating the subset contents, UPoT can provide adaptive quantization levels for various distributions. For each CNN model layer, UPoT automatically searches for the optimized distribution that minimizes the quantization error. Moreover, we design an efficient accelerator system with specifically optimized power-of-two multipliers and requantization units. Evaluations show that the proposed architecture can provide high-performance CNN inference with reduced circuit area and energy, and outperforms several mainstream CNN accelerators with higher ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$8\times $ </tex-math></inline-formula> – <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$65\times $ </tex-math></inline-formula> ) area efficiency and ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$2\times $ </tex-math></inline-formula> – <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$19\times $ </tex-math></inline-formula> ) energy efficiency. Further experiments of 4/3/2-bit quantization on ResNet18/50, MobileNet_V2 and EfficientNet models show that our UPoT can achieve high model accuracy which greatly outperform other state-of-the-art low-bit quantization methods by 0.3%–6%. The results indicate that our approach provides a highly-efficient accelerator for low-bit CNN model quantization with low hardware overheads and good model accuracy.

An edge-located uniform pattern recovery mechanism using statistical feature-based optimal center pixel selection strategy for local binary pattern

The Local Binary Pattern (LBP) is a commonly used method for texture classification that performs well in terms of feature discrimination. However, (1) LBP can misclassify some important edge-located textures as non-uniform patterns with only one bin in the feature histogram, thus losing their discrimination capability. (2) When center pixel is contaminated by noise, a uniform pattern may be transformed into a non-uniform pattern, which can seriously affect the obtained LBP, thus degrading the classification results. To overcome these drawbacks, an edge-located uniform pattern recovery mechanism using statistical feature-based optimal center pixel selection strategy (SFB-OCPS) is proposed in this paper.To extract the correct edge pixels, we divide the whole texture image into 16 = 4 × 4 sub-images and propose an edge pixel selection strategy (EPSS) based on adaptive quantization with local threshold on each sub-image. Then 3 candidate center pixels constructed by statistical features of the local sampling neighborhood are generated for each edge-located center pixel. After the steps above, the SFB-OCPS strategy is introduced into the LBP-based algorithms. It is possible to recover some important edge-located non-uniform patterns to uniform patterns with an optimal center pixel selection, thus improving feature discrimination capability of the LBP-based algorithms.It should be emphasized that any LBP variants can introduce the proposed SFB-OCPS strategy to achieve the recovery of the edge-located uniform patterns. To validate the effectiveness of the proposed SFB-OCPS strategy, we introduce the SFB-OCPS strategy into the original LBP and 5 representative LBP-based algorithms. Experiments are conducted on 6 representative texture databases. Classification comparison reveals that the introduction of SFB-OCPS strategy can significantly improve the texture classification performance of LBP-based algorithms. Additionally, the noise-robustness of the proposed SFB-OCPS strategy is also verified through a series of experiments.

Neuromorphic processor-oriented hybrid Q-format multiplication with adaptive quantization for tiny YOLO3

Deep neural networks (DNNs) have delivered unprecedented achievements in the modern Internet of Everything society, encompassing autonomous driving, expert diagnosis, unmanned supermarkets, etc. It continues to be challenging for researchers and engineers to develop a high-performance neuromorphic processor for deployment in edge devices or embedded hardware. DNNs’ superpower derives from their enormous and complex network architecture, which is computation-intensive, time-consuming, and energy-heavy. Due to the limited perceptual capacity of humans, accurate processing results from DNNs require a substantial amount of computing time, making them redundant in some applications. Utilizing adaptive quantization technology to compress the DNN model with sufficient accuracy is crucial for facilitating the deployment of neuromorphic processors in emerging edge applications. This study proposes a method to boost the development of neuromorphic processors by conducting fixed-point multiplication in a hybrid Q-format using an adaptive quantization technique on the convolution of tiny YOLO3. In particular, this work integrates the sign-bit check and bit roundoff techniques into the arithmetic of fixed-point multiplications to address overflow and roundoff issues within the convolution’s adding and multiplying operations. In addition, a hybrid Q-format multiplication module is developed to assess the proposed method from a hardware perspective. The experimental results prove that the hybrid multiplication with adaptive quantization on the tiny YOLO3’s weights and feature maps possesses a lower error rate than alternative fixed-point representation formats while sustaining the same object detection accuracy. Moreover, the fixed-point numbers represented by Q(6.9) have a suboptimal error rate, which can be utilized as an alternative representation form for the tiny YOLO3 algorithm-based neuromorphic processor design. In addition, the 8-bit hybrid Q-format multiplication module exhibits low power consumption and low latency in contrast to benchmark multipliers.

Towards perceptual image watermarking with robust texture measurement

Improving the perceptual quality of images is challenging for the robust watermarking method. To address this problem, most existing algorithms proposed to guide the watermark embedding based on just noticeable difference (JND) model, which mainly consists of corresponding contrast masking factor. However, existing metrics are mostly based on the intra-block texture energy, which cannot exactly describe the texture information and meet the needs of watermarking robustness. In this paper, we present a new texture metric, which incorporates the visual saliency based inter-block texture regularity with the robust discrete cosine transform (DCT) coefficients. The JND model is constructed as a modulation factor by using the proposed robust texture metric, which embeds the intra-block energy and inter-block regularity measurement to capture local and global features. Based on the proposed model, we introduce an adaptive quantization step for the watermarking framework. Both quantitative and qualitative results show that our scheme achieves better performance in comparison with several state-of-the-art schemes.

Adaptive Color Quantization Method with Multi-level Thresholding

In this study, a novel color quantization approach which automatically estimates the number of colors by multi-level thresholding based on the histogram is proposed. The method consists of three stages. First, red–green–blue is clustered by threshold values. Thus, the pixels are positioned in a cluster or sub-prism. Second, the color palette is produced by determining the centroids of the clusters. Finally, the pixels are reassigned to clusters based on their distance from each centroid. The average of the pixels included in each cluster also represents the color of that cluster. While conventional methods are user-dependent, the proposed algorithm automatically generates the number of colors by considering the pixels assigned to the clusters. Additionally, the multi-level thresholding approach is also a solution to the initialization problem, which is another important issue for quantization. Consequently, the experimental results of the method tested with various images show better performance than many frequently used quantization techniques.

Neuroadptive quantization tracking control with accelerate convergence rate for self‐restructuring systems

AbstractIn this article, we investigate the tracking control problem for a class of self‐restructuring systems with quantized input. The underlying system model is different from the one with fixed structure, and is able to reflect the impact arising from subsystem failure, system switching, and subsystem self‐expansion and so forth. Furthermore, the system is driven with quantized input. For such systems we develop a neural network‐based adaptive quantization control method with several attractive features including: (1) it is a less model‐dependent based control approach with which little information on the system model is required; (2) the quantized input does not require exact knowledge of quantization parameters; (3) the tracking error is ensured to be ultimately uniformly bounded and convergence rate of the tracking error is adjustable via the introduced rate function in the control algorithm, and the tracking error converges into a specific compact set. The benefits and feasibility of the proposed control method are also validated and confirmed by numerical simulations.

Smart algorithm in wireless networks for video streaming based on adaptive quantization

SummaryNowadays, smart multimedia network services have become crucial in the healthcare system. The network parameters of Quality of Service (QoS) are widely affecting the efficiency and accuracy of multimedia streaming in wireless environments. This paper proposes an adaptation framework model, which makes a relation between the QP (quantization parameter) in H.264 and H.265 codecs and the QoS of 5G wireless technology. Besides, the effect of QP and packet loss have been studied because of their impact on video streaming. Packet loss of 5G wireless network characteristic is emulated to determine the impact of QP on the received video quality using objective and subjective quality metrics such as PSNR (peak signal to noise ratio), SSIM (structure similarity), and DMOS (differential mean opinion score). In this research, a Testbed is implemented to stream the encoded video from the server to the end users. The application model framework has automatically evaluated the QoE (Quality of Experience). Accordingly, the model detects the defect of network packet loss and selects the optimum QP value to enhance the QoE by the end‐users. The application has been tested on low and high video motions with full high definition (HD) resolution (1920 × 1080) which were taken from ( https://www.xiph.org/downloads/). Test results based on the objective and subjective quality measurements indicate that an optimal QP = 35 and QP = 30 have been chosen for low and high motion respectively to satisfy user QoE requirements.

Sigma Delta Quantization for Images

AbstractIn signal quantization, it is well‐known that introducing adaptivity to quantization schemes can improve their stability and accuracy in quantizing bandlimited signals. However, adaptive quantization has only been designed for one‐dimensional signals. The contribution of this paper is two‐fold: (i) we propose the first family of two‐dimensional adaptive quantization schemes that maintain the same mathematical and practical merits as their one‐dimensional counterparts, and (ii) we show that both the traditional 1‐dimensional and the new 2‐dimensional quantization schemes can effectively quantize signals with jump discontinuities, which immediately enable the usage of adaptive quantization on images. Under mild conditions, we show that by using adaptivity, the proposed method is able to reduce the quantization error of images from the presently best to the much smaller , where s is the number of jump discontinuities in the image and P () is the total number of samples. This ‐fold error reduction is achieved via applying a total variation norm regularized decoder, whose formulation is inspired by the mathematical super‐resolution theory in the field of compressed sensing. Compared to the super‐resolution setting, our error reduction is achieved without requiring adjacent spikes/discontinuities to be well‐separated, which ensures its broad scope of application.We numerically demonstrate the efficacy of the new scheme on medical and natural images. We observe that for images with small pixel intensity values, the new method can significantly increase image quality over the state‐of‐the‐art method. © 2022 Wiley Periodicals, Inc.

Performance Analysis and Bit Allocation of Cell-Free Massive MIMO Network With Variable-Resolution ADCs

This paper concentrates on cell-free massive multiple-input and multiple-output (MIMO) network with variable-resolution analog-to-digital converters (ADCs). In such an architecture, all ADCs equipping at any access point (AP) can use arbitrary bit resolution to realize adaptive quantization and reduce power consumption. Under this circumstance, we first introduce a quantization-aware channel estimator based on linear minimum mean-square error (LMMSE) theory. On this basis, intra-AP and inter-AP bit allocation problems are investigated to maximize channel estimation quality subject to the total number of quantization bits. By leveraging the statistical characteristics of the estimated channels and estimation errors, we then derive the theoretical expressions of the achievable uplink spectral efficiency (SE) for maximal ratio combining (MRC) and minimum mean-square error (MMSE) combining, respectively. Furthermore, to maximize the sum SE under the constraint of total ADC quantization bits, we also investigate intra-AP and inter-AP bit allocation problems for both single-user and multi-user scenarios. Finally, simulation results confirm that our theoretical analyses are correct and accurate. In addition, we resort to numerical results to achieve some new insights and verify the advantages and conclusions pertinent to the proposed bit allocation techniques.

Point Cloud Soft Multicast for Untethered XR Users

3D point cloud data formats are used to express three-dimensional (3D) information using numerous points in a 3D space. A key challenge is the delivery of high-quality 3D point cloud for the users under a diverse channel quality and available bandwidth to share the same 3D space across multiple untethered extended reality (XR) users. The existing digital-based schemes suffer from two issues owing to the diversity: cliff and leveling-off effects. This paper proposes a novel soft multicasting scheme of point cloud data for untethered XR users. The key ideas of the proposed scheme are three-fold: 1) integration of graph signal processing and analog modulation to adaptively improve the 3D reconstruction quality according to the channel quality for all individual XR users, 2) integration of Givens rotation and non-uniform adaptive quantization to reduce metadata overhead for the graph Fourier transform, and 3) prioritized transmission of the metadata to realize adaptive quality improvement based on the bandwidth available for each XR user. This paper reveals that the proposed scheme prevents cliff and leveling-off effects even when the XR users experience different channel qualities. Furthermore, the proposed transmission exhibits better 3D reconstruction quality compared with the state-of-the-art graph-based delivery scheme in band-limited environments.

Toward Semantic Communications: Deep Learning-Based Image Semantic Coding

Semantic communications has received growing interest since it can remarkably reduce the amount of data to be transmitted without missing critical information. Most existing works explore the semantic encoding and transmission for text and apply techniques in Natural Language Processing (NLP) to interpret the meaning of the text. In this paper, we conceive the semantic communications for image data that is much more richer in semantics and bandwidth sensitive. We propose an reinforcement learning based adaptive semantic coding (RL-ASC) approach that encodes images beyond pixel level. Firstly, we define the semantic concept of image data that includes the category, spatial arrangement, and visual feature as the representation unit, and propose a convolutional semantic encoder to extract semantic concepts. Secondly, we propose the image reconstruction criterion that evolves from the traditional pixel similarity to semantic similarity and perceptual performance. Thirdly, we design a novel RL-based semantic bit allocation model, whose reward is the increase in rate-semantic-perceptual performance after encoding a certain semantic concept with adaptive quantization level. Thus, the task-related information is preserved and reconstructed properly while less important data is discarded. Finally, we propose the Generative Adversarial Nets (GANs) based semantic decoder that fuses both locally and globally features via an attention module. Experimental results demonstrate that the proposed RL-ASC is noise robust and could reconstruct visually pleasant and semantic consistent image in low bit rate condition.

Nonparametric Clustering of Mixed Data Using Modified Chi-Squared Tests.

We propose a non-parametric method to cluster mixed data containing both continuous and discrete random variables. The product space of the continuous and discrete sample space is transformed into a new product space based on adaptive quantization on the continuous part. Detection of cluster patterns on the product space is determined locally by using a weighted modified chi-squared test. Our algorithm does not require any user input since the number of clusters is determined automatically by data. Simulation studies and real data analysis results show that our proposed method outperforms the benchmark method, AutoClass, in various settings.

Computer-Aided Diagnosis Using Hybrid Technique for Fastened and Accurate Analysis of Tuberculosis Detection with Adaboost and Learning Vector Quantization

Background: The concept of tuberculosis diagnosis plays a significant role in the current world since, in accordance with the Global Tuberculosis (TB) Report in 2019, more than one million cases are reported per year in India. Various tests are available even then the chest X-ray is the most significant one, devoid of which the diagnosis will be incomplete. By the usage of computationally designed algorithms, various clinical, as well as diagnostic functions, were built in ancient poster anterior chest radiographs. The Digital image (X-ray) may be an essential medium for examining and annotating patient’s demographics coverage in the screening of TB via chest radiography. Results: Even though several medicines are available to cure TB, diagnosis with accuracy is a major challenge. So, we have introduced a fastened technique with the merged combination of Adaptive Boosting (AdaBoost) and learning vector quantization (LVQ) for determining TB in an easier way with the input chest X-ray image of a person with the aid of computer-aided diagnosis with greatest accuracy, precision, recall and F1 values. This finest technique got an accuracy of 94.73% when compared to the prior conventional methods used such as SVM and Convolutional Neural Network. Conclusions: Tuberculosis detection can be done in a meaningful way with the aid of MATLAB simulation using Computer Aided Diagnosis. The algorithms Adaboost and LVQ works best with the datasets for around 400 chest X-ray images for detecting the normal and abnormal images conditions for the detection of the disease for a patient suspected to have TB, in a fraction of seconds.

FLoRa: Sequential fuzzy extractor based physical layer key generation for LPWAN

The security of Low-Power Wide-Area Network (LPWAN) mainly relies on encryption for ensuring packet integrity and confidentiality. Unfortunately, the latest LPWAN specifications refrain from specifying how to distribute keys for encryption. In this paper, we tackle this problem via physical layer security, which exploits the channel characteristics to generate secret keys at the physical layer. In order to generate consistent keys from noisy feature sources and to achieve high reconciliation success rate, we propose FLoRa, a physical layer key generation system for LPWAN based on sequential fuzzy extractor. An adaptive multi-bit quantization algorithm is first proposed to generate the initial key, which accelerates the bit generation rate at the start-up procedure. We then design a novel fuzzy extractor by sequentially slicing the initial key, which improves the reconciliation success rate, as well as reduces the key reconciliation time. We implement FLoRa in a LoRaWAN based network prototype and evaluate it by conducting extensive indoor and outdoor experiments. Experimental results reveal that FLoRa is capable of generating consistent secret keys with high key generation performance in both static and dynamic network environments.

Photonics-Assisted Millimeter-Wave Communication System Based on Low-Bit Gaussian Mixture Model Adaptive Vector Quantization

The quantization technique with a low-bit resolution can significantly reduce the cost and power consumption of analog-to-digital converter (ADC). It will play an important role in energy conservation and cost reduction for the incoming B5G millimeter-wave (MMW) communication systems. In this paper, we propose and demonstrated experimentally a low-bit Gaussian mixture model (GMM) based non-uniform adaptive vector quantization (AVQ) scheme for the low-cost intensity modulated envelope detection photonics-assisted 28 GHz MMW communication system for the first time. The principles of GMM-based one-dimensional adaptive scalar quantization (ASQ) and multi-dimensional AVQ are first introduced, and then are used to realize the low-bit non-uniform adaptive quantization for reducing the ADC bit resolution of MMW receiver. Furthermore, the performance of traditional uniform quantization, the present <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">K</i> -means and proposed GMM-based non-uniform ASQ/AVQ schemes are evaluated and compared in detail. Utilizing the proposed GMM-based AVQ scheme, the ADC quantization resolution in our MMW receiver can be reduced from 5 bits of the traditional uniform quantization to as low as 2 bits, without noticeable performance penalty. Moreover, as compared with the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">K</i> -means-based quantization scheme, the MMW receiver enabled by GMM-based ASQ/AVQ scheme can save about half of the quantization time under similar performance. This is mainly because the clustering based on probability converges faster than the Euclidean distance, which significantly reduces the number of iterations required. Therefore, the GMM-based AVQ scheme is a promising solution to realize high performance ADCs with low-bit resolution for future MMW-enabled optical fiber wireless access networks.

Limited Coding-Length-Based Sliding-Mode Control With Adaptive Quantizer’s Parameter

In Niu and Ho (2014), an interesting issue was investigated on how an existing controller can still achieve the same control performance under the quantization case by designing the online adjusting strategy for the quantizer’s parameter. However, the above-mentioned work ignored the limitation of the coding length in the practical transmitting procedure such that the desirable performance may not be achieved due to the inevitable coding errors. Hence, this article will further consider the above-mentioned issue under the limited coding length. Different coding rules are designed for the quantized states and the adaptive quantizer’s parameters. The minimum bound of coding errors is explicitly given, which are dependent on the coding length and the coding range. Moreover, the relation between the coding length and the control performance is established, that is, the coding length has a direct effect on the convergent upper bound of the system state and the sliding domain around the specified sliding surface. Finally, an operational amplifier circuit system is provided to verify the proposed method.

Adaptive Consensus Quantized Control for a Class of High-Order Nonlinear Multi-Agent Systems With Input Hysteresis and Full State Constraints

For a class of high-order nonlinear multi-agent systems with input hysteresis, an adaptive consensus output-feedback quantized control scheme with full state constraints is investigated. The major properties of the proposed control scheme are: 1) According to the different hysteresis input characteristics of each agent in the multi-agent system, a hysteresis quantization inverse compensator is designed to eliminate the influence of hysteresis characteristics on the system while ensuring that the quantized signal maintains the desired value. 2) A barrier Lyapunov function is introduced for the first time in the hysteretic multi-agent system. By constructing state constraint control strategy for the hysteretic multi-agent system, it ensures that all the states of the system are always maintained within a predetermined range. 3) The designed adaptive consensus output-feedback quantization control scheme allows the hysteretic system to have unknown parameters and unknown disturbance, and ensures that the input signal transmitted between agents is the quantization value, and the introduced quantizer is implemented under the condition that only its sector bound property is required. The stability analysis has proved that all signals of the closed-loop are semi-globally uniformly bounded. The StarSim hardware-in-the-loop simulation certificates the effectiveness of the proposed adaptive quantized control scheme.

Reconfigurable and hardware efficient adaptive quantization model-based accelerator for binarized neural network

Binarized neural networks (BNNs) architecture play a vital role in the development of deep learning accelerator for memory-constrained IoT devices. However, the cost-efficiency of the domain-specific accelerators still requires a simplified structure to enhance the computing performance of BNNs. We propose and present a reconfigurable BNN accelerator to improve the computing speed through channel amplitude and an adaptive spatial amplitude model. To reduce the redundant operations in the binarization of matrix multiplication, we introduce the channel amplitude model for BNN convolution functions. The adaptive spatial amplitude is implemented with the help of a matrix multiplication layer with 3 × 3 convolutions to get spatial information and increases the computation speed. For hardware implementation, the channel amplitude conversion is deployed through XNOR-popcount convolutions of each layer. The performance results of our proposed accelerator provide 2–10 × line buffer efficiency, 1.6–3.7 × processing frequency enhancement, and 2.35 × processing element reduction compared with existing baselines using VGG-16 based FPGA implementation.

Communication-Efficient Federated Learning with Adaptive Quantization

Federated learning (FL) has attracted tremendous attentions in recent years due to its privacy-preserving measures and great potential in some distributed but privacy-sensitive applications, such as finance and health. However, high communication overloads for transmitting high-dimensional networks and extra security masks remain a bottleneck of FL. This article proposes a communication-efficient FL framework with an Adaptive Quantized Gradient (AQG), which adaptively adjusts the quantization level based on a local gradient’s update to fully utilize the heterogeneity of local data distribution for reducing unnecessary transmissions. In addition, client dropout issues are taken into account and an Augmented AQG is developed, which could limit the dropout noise with an appropriate amplification mechanism for transmitted gradients. Theoretical analysis and experiment results show that the proposed AQG leads to 18% to 50% of additional transmission reduction as compared with existing popular methods, including Quantized Gradient Descent (QGD) and Lazily Aggregated Quantized (LAQ) gradient-based methods without deteriorating convergence properties. Experiments with heterogenous data distributions corroborate a more significant transmission reduction compared with independent identical data distributions. The proposed AQG is robust to a client dropping rate up to 90% empirically, and the Augmented AQG manages to further improve the FL system’s communication efficiency with the presence of moderate-scale client dropouts commonly seen in practical FL scenarios.