SQL4NN: Validation and expressive querying of models as data

2025-02-20

AI Summary

This work addresses the absence of a unified, verifiable, data-centric analytical paradigm for machine learning models. We propose, for the first time, treating trained models as *first-class data objects* within the relational database paradigm. Methodologically, we design a declarative query framework built upon SQL extensions, integrating model structural metadata, behavioral interfaces, and interpretability modules, while deeply embedding into standard relational query engines. Our core contribution is enabling unified, composable, and formally verifiable SQL expressions for critical model-assurance tasks—including fairness auditing, robustness verification, and data provenance tracing—thereby overcoming the longstanding analytical fragmentation between models and data. Experimental evaluation demonstrates that our framework achieves both high efficiency and formal verifiability across diverse diagnostic tasks, establishing a foundation for rigorous, database-backed ML model analysis.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Data Mining & Knowledge Management: Representing, Reasoning, and Using Provenance, Trust
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Web data provenance, reliability, and authenticity

Abstract

We consider machine learning models, learned from data, to be an important, intensional, kind of data in themselves. As such, various analysis tasks on models can be thought of as queries over this intensional data, often combined with extensional data such as data for training or validation. We demonstrate that relational database systems and SQL can actually be well suited for many such tasks.

Problem

• Machine learning models as data
• Querying models with SQL
• Combining intensional and extensional data

Innovation

• SQL for model validation
• Querying machine learning models
• Relational databases for data analysis