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.

9th Biennial Conference on Innovative Data Systems Research, CIDR 2019
Centrum Wiskunde & Informatica, Amsterdam, The Netherlands

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