Date: November 08, 2018 16:15–17:00
Place: Exactum C122 (Kumpula)
Speaker: Michael Mathioudakis
Speaker: Michael Mathioudakis (University of Helsinki)
Title: Coresets for k-Means clustering
Abstract:
Coresets are compact representations of data, accompanied with a guarantee that models trained on them have competitive performance with models trained on all the data. In this talk, I will discuss state-of-the-art algorithms on coresets for k-Means by Bachem et.al. [1], presented earlier this year at the KDD conference.
[1] Bachem, Olivier, Mario Lucic, and Andreas Krause. “Scalable k-Means Clustering via Lightweight Coresets.” International Conference on Knowledge Discovery and Data Mining (KDD). 2018.