Relational Database Distillation: From Structured Tables to Condensed Graph Data

2025-10-08

AI Summary

To address the excessive storage and training overhead of graph representation learning on large-scale relational databases (RDBs), this paper proposes the Relational Database Distillation (RDD) framework, which compresses multi-table RDBs into compact heterogeneous graphs while preserving predictive utility. Methodologically, RDD avoids costly bilevel optimization via a kernel ridge regression–guided objective; enhances cross-task generalization through pseudo-labeling; and integrates multimodal column encoding, primary-foreign key structural modeling, and feature-level distillation. Experiments on multiple real-world databases demonstrate that RDD reduces graph size by an average of 72% while maintaining model performance on classification and regression tasks comparable to that achieved on the original databases. The framework thus establishes a scalable new paradigm for efficient graph learning over relational data.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Relational databases (RDBs) underpin the majority of global data management systems, where information is structured into multiple interdependent tables. To effectively use the knowledge within RDBs for predictive tasks, recent advances leverage graph representation learning to capture complex inter-table relations as multi-hop dependencies. Despite achieving state-of-the-art performance, these methods remain hindered by the prohibitive storage overhead and excessive training time, due to the massive scale of the database and the computational burden of intensive message passing across interconnected tables. To alleviate these concerns, we propose and study the problem of Relational Database Distillation (RDD). Specifically, we aim to distill large-scale RDBs into compact heterogeneous graphs while retaining the predictive power (i.e., utility) required for training graph-based models. Multi-modal column information is preserved through node features, and primary-foreign key relations are encoded via heterogeneous edges, thereby maintaining both data fidelity and relational structure. To ensure adaptability across diverse downstream tasks without engaging the traditional, inefficient bi-level distillation framework, we further design a kernel ridge regression-guided objective with pseudo-labels, which produces quality features for the distilled graph. Extensive experiments on multiple real-world RDBs demonstrate that our solution substantially reduces the data size while maintaining competitive performance on classification and regression tasks, creating an effective pathway for scalable learning with RDBs.

Problem

• Distilling large relational databases into compact graphs
• Reducing storage overhead and training time for databases
• Maintaining predictive performance while compressing database structure

Innovation

• Distills large databases into compact heterogeneous graphs
• Encodes table relations via multi-modal node features
• Uses kernel ridge regression for efficient feature distillation