SQL4NN: Validation and expressive querying of models as data

TL;DR

This paper proposes using relational databases and SQL to perform validation and expressive querying of machine learning models treated as data, enabling complex analysis tasks on models and their associated data.

Contribution

It introduces a novel approach to model analysis by leveraging relational databases and SQL for querying models as data, bridging the gap between machine learning and database systems.

Findings

SQL can effectively support model validation tasks

Relational databases facilitate complex model analysis queries

Models can be treated as data for analysis within database systems

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.