Abstract
Given a set of n points in a metric space, and a parameter k, the goal of “diversity maximization” is to pick a subset of size k with maximum “diversity”. Several measures have been proposed in theory and used in practice to model the notion of diversity. Diversity maximization comes up in many practical tasks such as data summarization, recommendation systems, and search.
In this talk, I will survey recent results on diversity maximization problems in both the offline and the composable coresets settings. Composable coresets are small subsets of the data with a composability property: given a collection of data sets, the union of the coresets for all data sets in the collection, should contain an approximately good solution with respect to union of the whole data. Composable coreset is a powerful tool that simultaneously leads to algorithms in several models of computation concerning massive data.