检索结果-内蒙古大学图书馆

arithmetic coding-Based 5-Bit Weight Encoding and Hardware Decoder for CNN Inference in Edge Devices

IEEE ACCESS 2021年 9卷 166736-166749页

作者： Lee, Jong Hun Kong, Joonho Munir, Arslan Kyungpook Natl Univ Sch Elect & Elect Engn Daegu 41566 South Korea Kyungpook Natl Univ Sch Elect Engn Daegu 41566 South Korea Kansas State Univ Dept Comp Sci Manhattan KS 66506 USA

Convolutional neural networks (CNNs) have gained a huge attention for real-world artificial intelligence (AI) applications such as image classification and object detection. On the other hand, for better accuracy, the size of the CNNs' parameters (weights) has been increasing, which in turn makes it difficult to enable on-device CNN inferences in resource-constrained edge devices. Though weight pruning and 5-bit quantization methods have shown promising results, it is still challenging to deploy large CNN models in edge devices. In this paper, we propose an encoding and hardware-based decoding technique which can be applied to 5-bit quantized weight data for on-device CNN inferences in resource-constrained edge devices. Given 5-bit quantized weight data, we employ arithmetic coding with range scaling for loss-less weight compression, which is performed offline. When executing on-device inferences with underlying CNN accelerators, our hardware decoder enables a fast in-situ weight decompression with small latency overhead. According to our evaluation results with five widely used CNN models, our arithmetic coding-based encoding method applied to 5-bit quantized weights shows a better compression ratio by 9.6x while also reducing the memory data transfer energy consumption by 89.2%, on average, as compared to the case of uncompressed 32-bit floating-point weights. When applying our technique to pruned weights, we obtain better compression ratios by 57.5x-112.2x while reducing energy consumption by 98.3%-99.1% as compared to the case of 32-bit floating-point weights. In addition, by pipelining the weight decoding and transfer with the CNN execution, the latency overhead of our weight decoding with 16 decoding unit (DU) hardware is only 0.16%-5.48% and 0.16%-0.91% for non-pruned and pruned weights, respectively. Moreover, our proposed technique with 4-DU decoder hardware reduces system-level energy consumption by 1.1%-9.3%.

关键词： Convolutional neural networks arithmetic coding weight compression edge devices 5-bit quantization

来源：评论

学校读者我要写书评

暂无评论

Research on Practical Implementation of Binary arithmetic coding with Forbidden Symbol for Error Resilience

Research on Practical Implementation of Binary Arithmetic Co...

引用

International Conference on Frontiers in Computer Education (ICFCE 2011)

作者： Zhang, Wei Cen, Feng Zhu, Fanglai Tongji Univ Minist Educ Key Lab Embedded Syst & Serv Comp 1239 Siping Rd Shanghai 200092 Peoples R China Tongji Univ Dept Control Sci & Engn Shanghai 200092 Peoples R China

ISBN: (纸本)9783642275517

arithmetic coding is a widely applied compression tool with superior coding efficiency to other entropy coding methods. However, it suffers from the error resilience and complexity. In this paper, the integer implementation of binary arithmetic coding with forbidden symbol for error resilience is studied. coding redundancies for employing different quantization coefficients in probability representation and cost effective backtracking distance in bits for maximum a posteriori (MAP) decoding are studied in depth. We observe that the optimal quantization coefficients are independent of forbidden symbol and the probabilities of source and the cost effective backtracking distance is related to the source entropy and the given forbidden symbol probabilities. Such observations are also demonstrated by extensive experiments.

关键词： arithmetic coding Forbidden symbol Error resilience MAP

来源：评论

学校读者我要写书评

暂无评论

Order-preserving encryption schemes based on arithmetic coding and matrices

Order-preserving encryption schemes based on arithmetic codi...

引用

Federated Conference on Computer Science and Information Systems (FedCSIS)

作者： Krendelev, Sergey F. Yakovlev, Mikhail Usoltseva, Maria Novosibirsk State Univ Lab Modern Comp Technol Novosibirsk 630090 Russia Novosibirsk State Univ Dept Informat Technol Novosibirsk 630090 Russia

ISBN: (纸本)9788360810583

In this article we describe two alternative order-preserving encryption schemes. First scheme is based on arithmetic coding and the second scheme uses sequence of matrices for data encrypting. In the beginning of this paper we briefly describe previous related work published in recent time. Then we propose alternative variants of OPE and consider them in details. We examine drawbacks of these schemes and suggest possible ways of their improvement. Finally we present statistical results of implemented prototypes and discuss further work.

关键词： Educational institutions Encoding Encryption Generators Linear approximation Polynomials cryptography arithmetic coding Linear approximation Educational institutions dynamos Polynomial matrices ENCRYPTION SCHEMES Scheme

来源：评论

学校读者我要写书评

暂无评论

Neural Network-Based arithmetic coding for Inter Prediction Information in HEVC

Neural Network-Based Arithmetic Coding for Inter Prediction ...

引用

IEEE International Symposium on Circuits and Systems (IEEE ISCAS)

作者： Ma, Changyue Liu, Dong Peng, Xiulian Zha, Zheng-Jun Wu, Feng Univ Sci & Technol China CAS Key Lab Technol Geospatial Informat Proc & Ap Hefei 230027 Anhui Peoples R China Microsoft Res Asia Beijing 100080 Peoples R China

ISBN: (纸本)9781728103976

Entropy coding is a fundamental technique in video coding to remove the statistical redundancy in syntax elements. Currently, context-adaptive binary arithmetic coding (CABAC) is used as the entropy coding tool in HEVC. Considering that the manually designed binarization and context models are not flexible to estimate the probability of the syntax elements, we use neural networks to estimate the probability of the syntax elements, then the estimated probabilities together with the values of the syntax elements are fed into an arithmetic coding engine to fulfill entropy coding. In this paper, we focus on the syntax elements of inter prediction information that consists of merge flag, merge index, reference index, motion vector difference and motion vector prediction index in HEVC under low-delay P (LDP) setting. Compared with the previous work on neural network-based arithmetic coding for intra prediction modes and intra DC coefficients, there are three new characteristics in this paper. First, surrounding syntax elements are directly fed into the neural network without converting to reconstructed pixels. Second, unified neural networks are designed for different prediction block sizes. Finally, dependency among the syntax elements in current prediction unit is omitted to improve parallelism. Experimental results show that compared with HEVC, our proposed method achieves up to 0.5% and on average 0.3% BD-rate reduction in LDP configuration.

关键词： arithmetic coding CABAC HEVC inter prediction information neural network

来源：评论

学校读者我要写书评

暂无评论

MODIFIED DEcoding METRIC OF DISTRIBUTED arithmetic coding

MODIFIED DECODING METRIC OF DISTRIBUTED ARITHMETIC CODING

引用

IEEE International Conference on Image Processing (ICIP)

作者： Yang, Yanchao Qi, Mingwei Zhou, Junwei Cheng, Lee-Ming Wuhan Univ Technol Sch Comp Sci & Technol Wuhan Peoples R China City Univ Hong Kong Dept Elect Engn Hong Kong Peoples R China

ISBN: (数字)9781728163956

ISBN: (纸本)9781728163956

As an alternative implementation of Slepian-Wolf coding, distributed arithmetic coding is very competitive in short and medium block lengths. The existing distributed arithmetic coding decoder uses the maximum a posteriori metric and the M-algorithm to select the decoding sequence by considering the prior information. Since the prior probability distribution has been explored by the encoding process via the model stage and retained in the codeword, we propose a modified decoding metric ignoring the prior information. The reliability of the decoding paths is only determined by the correlation between the side information and the input source. Simulation results show that the modified metric can significantly reduce the decoding error.

关键词： Slepian-Wolf coding distributed arithmetic coding arithmetic coding distributed source coding

来源：评论

学校读者我要写书评

暂无评论

A flexible JPEG2000 image encryption based on arithmetic coding

A flexible JPEG2000 image encryption based on arithmetic cod...

引用

International Symposium on Intelligent Signal Processing and Communication Systems

作者： Ou, Yang Lee, Won-Young Rhee, Kyung Hyune Pukyong Natl Univ Dept Informat Security Pusan South Korea Pukyong Natl Univ Div Elect Comp & Telecommun Engn Busan 608737 South Korea

ISBN: (纸本)9781424414468

In this paper, we propose a flexible JPEG2000 image encryption based on arithmetic coding which combines both compression and encryption. Particularly, we randomly add a subinterval to the probability interval in each coding step during the iterative process of arithmetic coder. The range of this subinterval. can be flexibly adjusted depending on the properties of different applied environments. Furthermore, our approach supports backward compatibility so that an encryption-unaware format-compliant player can play the encrypted code stream without any crash. Notably, our algorithm achieves a very simple switch between the standard compression model and our joint model.

关键词： arithmetic coding joint compression and encryption MQ coder JPEG2000

来源：评论

学校读者我要写书评

暂无评论

Adaptive Path Profiling Using arithmetic coding 21

Adaptive Path Profiling Using Arithmetic Coding

引用

21st IEEE International Conference on Parallel and Distributed Systems ICPADS

作者： Chen, Gonglong Dong, Wei Zhejiang Univ Coll Comp Sci Hangzhou Zhejiang Peoples R China

ISBN: (纸本)9780769557854

Path profiling, which aims to trace a program's execution path, has been widely adopted in various areas such as record and replay, program optimizations, performance diagnosis, and etc. Many path profiling approaches have been proposed in the literature, including B.L. algorithm, and PAP. Unfortunately, both approaches suffer from large tracing overhead for representing long execution paths. In this paper, we propose AdapTracer, a path profiling approach based on arithmetic coding. There are two salient features in AdapTracer. First, it is space efficient by adopting a path profiling algorithm based on arithmetic coding. Second, it is adaptive by explicitly considering the execution frequency of each edge. We have implemented AdapTracer to profile Android applications. Our experimental evaluation uses modified JGF benchmarks to show AdapTracer's efficiency. Experimental results show that AdapTracer reduces the trace size by 44% on average and incurs execution overhead by 10% at most compared to PAP.

关键词： Path profiling arithmetic coding adaptive

来源：评论

学校读者我要写书评

暂无评论

arithmetic coding for countable alphabet sources with finite precision

引用

IEICE TRANSACTIONS ON FUNDAMENTALS OF ELECTRONICS COMMUNICATIONS AND COMPUTER SCIENCES 2001年第10期E84A卷 2576-2582页

作者： Nishiara, M Morita, H Univ Electrocommun Grad Sch Informat Syst Chofu Tokyo 1828585 Japan

An improved arithmetic coding which provides an encoder with finite calculation precision for source sequences over a countable alphabet is presented. Conventional arithmetic coding theoretically has infinite precision for real variables. However any algorithm implemented on a computer has finite precision. This implies that conventional arithmetic codes can only encode sequences over a finite alphabet. The improved arithmetic coding presented here has a computational complexity which is roughly proportional to the length of the source sequence for a given source.

关键词： arithmetic coding countable alphabet finite precision of calculation implementation of coding algorithm

来源：评论

学校读者我要写书评

暂无评论

arithmetic coding FOR DATA-COMPRESSION

引用

COMMUNICATIONS OF THE ACM 1987年第6期30卷 520-540页

作者： WITTEN, IH NEAL, RM CLEARY, JG Dept. of Computer Science The University of Calgary 2500 University Drive NW Calgary Canada T2N 1N4 Dept. of Computer Science The University of Calgary 2500 University Drive NW Calgary Canada T2N 1N4 Dept. of Computer Science The University of Calgary 2500 University Drive NW Calgary Canada T2N 1N4

This article presents information on the use of arithmetic coding for data compression. The article briefly reviews basic concepts of data compression and introduces the model-based approach that underlies most modern techniques. To many, data compression conjures up an assortment of ad hoc techniques such as conversion of spaces in text to tabs, creation of special codes for common words, or run-length coding of picture data. This contrasts with the more modern model-based paradigm for coding, where, from an input string of symbols and a model, an encoded string is produced that is usually a compressed version of the input. The decoder, which must have access to the same model, regenerates the exact input string from the encoded string. The model is a way of calculating in any given context, the distribution of probabilities for the next input symbol. It must be possible for the decoder to produce exactly the same probability distribution in the same context. Compression is achieved by transmitting the more probable symbols in fewer bits than the less probable ones.

关键词： Adaptive modeling arithmetic coding Huffman coding

来源：评论

学校读者我要写书评

暂无评论

An Integrated Approach for Lossless Image Compression Using CLAHE, Two-Channel Encoding and Adaptive arithmetic coding

引用

SN Computer Science 2024年第5期5卷 523页

作者： Kumar, P. R. Rajesh Prabhakar, M. School of Computer Science and Engineering REVA University Bangalore India

Lossless image compression techniques play a crucial role in preserving image quality while reducing storage space and transmission bandwidth. This paper proposes a novel hybrid integrated method for lossless image compression by combining Contrast Limited Adaptive Histogram Equalization (CLAHE), two-channel encoding, and adaptive arithmetic coding to achieve highly efficient compression without any loss of image information. The first step of the proposed approach involves applying Contrast Limited Adaptive Histogram Equalization (CLAHE) to enhance the local contrast of the image. This pre-processing step aids in reducing the entropy and increasing the redundancy in the image, creating a more favourable environment for subsequent compression algorithms. Next, the image is divided into two channels: one channel focuses on encoding essential structural information, while the other channel handles the finer details. This segregation leverages the inherent properties of images to improve compression efficiency. To achieve further compression gains, an adaptive arithmetic coding algorithm for encoding the data in each channel is utilized. Adaptive arithmetic coding adapts its probability model during the encoding process, leading to improved compression performance compared to traditional static coding methods. The proposed method offers significant potential in various applications, it is especially crucial in medical imaging, where large volumes of high-resolution images are generated during procedures such as MRI, CT scans, or digital pathology, transmitting high-quality images in resource-constrained environments, and facilitating image processing tasks requiring precise data preservation. CLAHE can be a valuable tool in medical imaging to enhance essential diagnostic information in medical images before compression. By improving contrast and visibility of structures, CLAHE may aid in achieving better compression efficiency and reduce the risk of introducing compres

关键词： arithmetic coding Bit rate CLAHE coding efficiency Compression ratio Lossless image coding Pre-processing Two-channel coding

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：