Learned Cardinalities: Estimating Correlated Joins with Deep Learning

• DOI: 10.48550/arxiv.1809.00677
• Publication Date: 2018-01-01
• Authors: Kipf, Andreas; Kipf, Thomas; Radke, Bernhard; Leis, Viktor; Boncz, Peter; Kemper, Alfons
• Keywords: FOS: Computer and information sciences; Computer Science - Databases; 0202 electrical engineering, electronic engineering, information engineering; Databases (cs.DB); 02 engineering and technology

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 significantly enhances the quality of cardinality estimation, which is the core problem in query optimization.<br/>CIDR 2019. https://github.com/andreaskipf/learnedcardinalities<br/>