[BibTeX] [RIS]
Towards Model-based Approximate Query Processing
Type of publication: Inproceedings
Citation: mk:AIDB2019
Booktitle: Working Notes of the 1st International Workshop on Applied AI for Database Systems and Applications (held in conjunction with VLDB 2019)
Year: 2019
Month: August
Location: Los Angeles, USA
URL: https://drive.google.com/file/d/1VAJGsf1vemhKl_IsLbznTv9TRwAcsaB3/view
Abstract: In this paper, we present a new approach to Approximate Query Processing (AQP) called Model-based AQP that leverages deep generative models learned over a dataset to answer SQL queries at interactive speeds. Different from classical AQP approaches, deep generative models allow us not only to compute approximate responses to ad-hoc queries even over rare sub-populations but additionally support a new class of queries called counterfactual queries enabling users to ask what-if queries. Furthermore, we think that deep generative models can not only be used for AQP in databases but also have other applications for problems such as Query Optimization as well as Data Cleaning.
Keywords: Approximate Query Processing, Databases, Deep Learning, SQL Queries, Sum-Product Networks
Authors Kulessa, Moritz
Hilprecht, Benjamin
Molina, Alejandro
Kersting, Kristian
Binnig, Carsten