Task-Oriented Lossy Compression with Data, Perception, and Classification Constraints

doi:arXiv:2405.04144

	Task-Oriented Lossy Compression with Data, Perception, and Classification Constraints
	Wang, Yuhan 1; Wu, Youlong2 ; Ma, Shuai 3; Zhang, Ying-Jun Angela 1
	2025-01-20
发表期刊	ARXIV (IF:13.8[JCR-2023],13.1[5-Year])
ISSN	1558-0008
EISSN	1558-0008
卷号	PP 期号:99
发表状态	已发表
DOI	arXiv:2405.04144
摘要	By extracting task-relevant information while maximally compressing the input, the information bottleneck (IB) principle has provided a guideline for learning effective and robust representations of the target inference. However, extending the idea to the multi-task learning scenario with joint consideration of generative tasks and traditional reconstruction tasks remains unexplored. This paper addresses this gap by reconsidering the lossy compression problem with diverse constraints on data reconstruction, perceptual quality, and classification accuracy. Firstly, we study two ternary relationships, namely, the rate-distortion-classification (RDC) and rate-perception-classification (RPC). For both RDC and RPC functions, we derive the closed-form expressions of the optimal rate for binary and Gaussian sources. These new results complement the IB principle and provide insights into effectively extracting task-oriented information to fulfill diverse objectives. Secondly, unlike prior research demonstrating a tradeoff between classification and perception in signal restoration problems, we prove that such a tradeoff does not exist in the RPC function and reveal that the source noise plays a decisive role in the classification-perception tradeoff. Finally, we implement a deep-learning-based image compression framework, incorporating multiple tasks related to distortion, perception, and classification. The experimental results coincide with the theoretical analysis and verify the effectiveness of our generalized IB in balancing various task objectives.
关键词	Information bottleneck lossy compression task-oriented communication rate-distortion theory perceptual quality
URL	查看原文
收录类别	PPRN.PPRN ; EI
语种	英语
资助项目	National Natural Science Foundation of China (NSFC)[62471270] ; Guangdong Basic and Applied Basic Research Foundation[2024A1515030028] ; General Research Fund from Research Grants Council of Hong Kong["14202421","14214122","14202723"] ; Area of Excellence Scheme from Research Grants Council of Hong Kong[AoE/E-601/22-R] ; NSFC/RGC Collaborative Research Scheme from Research Grants Council of Hong Kong[CRS_HKUST603/22]
WOS类目	Computer Science, Information Systems ; Mathematics
WOS记录号	PPRN:89076763
出版者	Institute of Electrical and Electronics Engineers Inc.
EI入藏号	20251618253719
EI主题词	Signal distortion
EI分类号	701.1 Electricity: Basic Concepts and Phenomena ; 716.1 Information Theory and Signal Processing ; 1106.3.1 Image Processing
原始文献类型	Article in Press
来源库	IEEE
文献类型	期刊论文
条目标识符	https://kms.shanghaitech.edu.cn/handle/2MSLDSTB/503591
专题	信息科学与技术学院信息科学与技术学院_PI研究组_吴幼龙组
通讯作者	Wu, Youlong
作者单位	1.Chinese Univ Hong Kong, Dept Informat Engn, Hong Kong, Peoples R China 2.ShanghaiTech Univ, Sch Informat Sci & Technol, Shanghai 201210, Peoples R China 3.Peng Cheng Lab, Shenzhen 518055, Peoples R China
通讯作者单位	信息科学与技术学院
推荐引用方式 GB/T 7714	Wang, Yuhan,Wu, Youlong,Ma, Shuai,et al. Task-Oriented Lossy Compression with Data, Perception, and Classification Constraints[J]. ARXIV,2025,PP(99).
APA	Wang, Yuhan,Wu, Youlong,Ma, Shuai,&Zhang, Ying-Jun Angela.(2025).Task-Oriented Lossy Compression with Data, Perception, and Classification Constraints.ARXIV,PP(99).
MLA	Wang, Yuhan,et al."Task-Oriented Lossy Compression with Data, Perception, and Classification Constraints".ARXIV PP.99(2025).