Conventional image compression methods are inadequate for intelligent analysis, as they overemphasize pixel-level precision while neglecting semantic significance and the interaction among multiple tasks. This paper introduces a Taskonomy-Aware Multi-Task Compression framework comprising (1) inter-coherent task grouping, which organizes synergistic tasks into shared representations to improve multi-task accuracy and reduce encoding volume, and (2) a conditional entropy-based directed acyclic graph (DAG) that captures causal dependencies among grouped representations. By leveraging parent representations as contextual priors for child representations, the framework effectively utilizes cross-task information to improve entropy model accuracy. Experiments on diverse vision tasks, including Keypoint 2D, Depth Z-buffer, Semantic Segmentation, Surface Normal, Edge Texture, and Autoencoder, demonstrate significant bitrate-performance gains, validating the method’s capability to reduce system entropy uncertainty. These findings underscore the potential of leveraging representation disentanglement, synergy, and causal modeling to learn compact representations, which enable efficient multi-task compression in intelligent systems.

Traditional image codecs prioritize signal fidelity and human perception, often neglecting machine vision tasks. Deep learning approaches have shown promising coding performance by leveraging rich semantic embeddings that can be optimized for both human and machine vision. However, these compact embeddings struggle to represent low-level details like contours and textures, leading to imperfect reconstructions. Additionally, existing learning-based coding tools lack scalability. To address these challenges, this paper presents a content-adaptive diffusion model for scalable image compression. The method encodes accurate texture through a diffusion process, enhancing human perception while preserving important features for machine vision tasks. It employs a Markov palette diffusion model with commonly-used feature extractors and image generators, enabling efficient data compression. By utilizing collaborative texture-semantic feature extraction and pseudo-label generation, the approach accurately learns texture information. A content-adaptive Markov palette diffusion model is then applied to capture both low-level texture and high-level semantic knowledge in a scalable manner. This framework enables elegant compression ratio control by flexibly selecting intermediate diffusion states, eliminating the need for deep learning model re-training at different operating points. Extensive experiments demonstrate the effectiveness of the proposed framework in image reconstruction and downstream machine vision tasks such as object detection, segmentation, and facial landmark detection. It achieves superior perceptual quality scores compared to state-of-the-art methods.

Satisfiability, equivalence, and implication problems involving conjunctive queries are important and widely encountered problems in database management systems. These problems need to be efficiently and effectively solved. In this paper, we consider queries which are conjunctions of the inequalities of the form (X op C), (X op Y), and/or (X op Y+C), where X and Y are two attributes, C is a constant, and op /spl epsiv/ {, /spl ges/}. These types of inequalities are widely used in database systems, since the first type is a selection, the second type is a /spl theta/-join, and the third type is a very popular clause in a deductive database system. The satisfiability, equivalence, and implication problems in the integer domain (for attributes and constants) have been shown to be NP-hard. However, we show that these problems can be solved efficiently in the real domain. The incorporation of the real domain is significant, because the real domain is practically and widely used in a database. Necessary and sufficient conditions and algorithms are presented. A novel concept of the "module closure" and a set of sound and complete axioms with respect to the "module closure" are also proposed to infer all correct and necessary inequalities from a given query. The proposed axioms generalize Ullman's axioms (1989) where queries only consist of /spl theta/-joins.

Satisfiability, implication, and equivalence problems involving conjunctive inequalities are important and widely encountered database problems that need to be efficiently and effectively processed. In this article we consider two popular types of arithmetic inequalities, ( X op Y ) and ( X op C ), where X and Y are attributes, C is a constant of the domain or X , and op ∈{<, ≤, =, ≠, >, ≥). These inequalities are most frequently used in a database system, inasmuch as the former type of inequality represents a 0-join, and the latter is a selection. We study the satisfiability and implication problems under the integer domain and the real domain, as well as under two different operator sets ({<, ≤, =, ≥, >} and {<, ≤, =, ≠, ≥, >}). Our results show that solutions under different domains and/or different operator sets are quite different. Out of these eight cases, excluding two cases that had been shown to be NP-hard, we either report the first necessary and sufficient conditions for these problems as well as their efficient algorithms with complexity analysis (for four cases), or provide an improved algorithm (for two cases). These iff conditions and algorithms are essential to database designers, practitioners, and researchers. These algorithms have been implemented and an experimental study comparing the proposed algorithms and those previously known is conducted. Our experiments show that the proposed algorithms are more efficient than previously known algorithms even for small input. The C++ code can be obtained by an anonymous ftp from .

The IS-A relationship (the class-subclass relationship) is one of the most fundamental and important properties in an object-oriented (OO) language and OO database management system (OODBMS). Due to the popularity and domination of relational database management systems (RDBMSs) and the fundamental importance of inheritance, supporting IS-A relationships in a RDBMS becomes very desired, and is essential to adapt a RDBMS to more advanced applications. The authors present an extended SQL system called ESQL which facilitates IS-A relation hierarchies in a RDBMS. The proposed ESQL uses constraints to resolve data redundancy and updating abnormality problems that exist in current OO languages such as C++ and Smalltalk, and OODBMS such as GemStone, PostGres, O/sub 2/, Iris, and Orion. Features such as inheritance constraints, subrelation assertions, mappings, and automatic tuple placement into its most specific subrelation are distinct in ESQL. These features are missing in current RDBMSs, OODBMSs and OO languages.>