A Deep Learning Blueprint for Relational Databases
Abstract
We introduce a modular neural message-passing scheme that closely follows the formal model of relational databases, effectively enabling end-to-end deep learning directly from database storages. We experiment with several instantiations of the scheme, including notably the use of cross-attention modules to capture the referential constraints of the relational model. We address the issues of efficient learning data representation and loading, salient to the database setting, and compare against representative models from a number of related fields, demonstrating favorable initial results.
Cite
Text
Zahradník et al. "A Deep Learning Blueprint for Relational Databases." NeurIPS 2023 Workshops: TRL, 2023.Markdown
[Zahradník et al. "A Deep Learning Blueprint for Relational Databases." NeurIPS 2023 Workshops: TRL, 2023.](https://mlanthology.org/neuripsw/2023/zahradnik2023neuripsw-deep/)BibTeX
@inproceedings{zahradnik2023neuripsw-deep,
title = {{A Deep Learning Blueprint for Relational Databases}},
author = {Zahradník, Lukáš and Neumann, Jan and Šír, Gustav},
booktitle = {NeurIPS 2023 Workshops: TRL},
year = {2023},
url = {https://mlanthology.org/neuripsw/2023/zahradnik2023neuripsw-deep/}
}