Learning Models over Relational Databases

ABSTRACT. In this talk I will overview recent results on learning classification and regression models over training datasets defined by feature extraction queries over relational databases. I will show that the complexity of this task can be connected with known notions of widths that measure the complexity of relational queries, such as the fractional hypertree width. This complexity can be much lower than that of the state of the art approach that first materialises the training dataset. Recent joint work with collaborators from industry shows that this approach can speed up real analytical workloads by several orders of magnitude over state-of-the-art systems. I will also highlight on-going work on linear algebra over databases and point out exciting directions for future work. This work is based on long-standing collaboration with Maximilian Schleich and Jakub Zavodny and more recent collaboration with Mahmoud Abo-Khamis, Hung Ngo, and XuanLong Nguyen.

Authors

• Dan Olteanu Oxford University

Keywords

No keywords are indexed for this paper.