Talk by Prof. Michael Grossniklaus:
"Advances in Cardinality Estimation for Database Systems"
26 July 2023 - 13:15
Prof. Dr. Michael Grossniklaus, Full Professor at the University of Konstanz
Michael Grossniklaus obtained his diploma and doctorate at the Department of Computer Science of ETH Zurich. After post-docs at the Politecnico di Milano, Portland State University and the Technical University of Vienna, Michael started his research group at the Department of Computer and Information Science at the University of Konstanz in 2013, first as a Junior Professor and since 2017 as a Full Professor. Michael's research area is database and information systems, focusing on query optimization. Apart from relational database systems, Michael and his team are also interested in non-relational database systems, particularly graph database systems. Michael is the scientific co-lead of the "Kontaktstudium IMP", a state-wide blended-learning program to qualify instructors to teach computer science in various school types in Baden-Württemberg. As part of the Excellence Strategy of the University of Konstanz, Michael initiated the "Advanced Data and Information Literacy Track", which provides students from all disciplines with basic training in computer science and its application. Finally, Michael is a principal investigator in the Excellence Cluster "Center for the Advanced Study of Collective Behavior".
In database systems, cardinality estimation refers to estimating the number of results a given database query will return. Accurate cardinality estimation is a cornerstone of effective query optimization as it forms the basis for comparing and selecting the most efficient query execution plan among all enumerated candidates. In this talk, we present the results of recently completed and currently ongoing projects to improve the accuracy of cardinality estimation that we conduct at the Database and Information Systems Group at the University of Konstanz. First, we outline a novel cardinality estimation technique based on label probability propagation for subgraph matching queries in property graph databases. Next, we discuss our ongoing work on estimating the cardinality of single-column LIKE-predicates in SQL using a bespoke neural language model. Finally, we show how geometric deep learning can improve the sample efficiency or learned cardinality estimation of general SQL queries in relational databases.
Bldg. 50.34, Room -102