Learned cardinalities: Estimating correlated joins with deep learning

Kipf A, Kipf T, Radke B, Leis V, Boncz P, Kemper A (2019)

Publication Type: Conference contribution

Publication year: 2019

Publisher: Conference on Innovative Data Systems Research (CIDR)

Conference Proceedings Title: CIDR 2019 - 9th Biennial Conference on Innovative Data Systems Research

Event location: Pacific Grove, CA US

Abstract

We describe a new deep learning approach to cardinality estimation. MSCN is a multi-set convolutional network, tailored to representing relational query plans, that employs set semantics to capture query features and true cardinalities. MSCN builds on sampling-based estimation, addressing its weaknesses when no sampled tuples qualify a predicate, and in capturing join-crossing correlations. Our evaluation of MSCN using a real-world dataset shows that deep learning signiicantly enhances the quality of cardinality estimation, which is the core problem in query optimization.

Authors with CRIS profile

Viktor Leis

Involved external institutions

University of Amsterdam NL Netherlands (NL) Technische Universität München (TUM) DE Germany (DE) Centrum Wiskunde & Informatica (CWI) NL Netherlands (NL)

How to cite

APA:

Kipf, A., Kipf, T., Radke, B., Leis, V., Boncz, P., & Kemper, A. (2019). Learned cardinalities: Estimating correlated joins with deep learning. In CIDR 2019 - 9th Biennial Conference on Innovative Data Systems Research. Pacific Grove, CA, US: Conference on Innovative Data Systems Research (CIDR).

MLA:

Kipf, Andreas, et al. "Learned cardinalities: Estimating correlated joins with deep learning." Proceedings of the 9th Biennial Conference on Innovative Data Systems Research, CIDR 2019, Pacific Grove, CA Conference on Innovative Data Systems Research (CIDR), 2019.

BibTeX: Download