Fully Dynamic Algorithm for Constrained Submodular Optimization
Abstract
The task of maximizing a monotone submodular function under a cardinality constraint is at the core of many machine learning and data mining applications, including data summarization, sparse regression and coverage problems. We study this problem in the context of fully dynamic streams, where elements can be both inserted and removed.
Our main result is a randomized algorithm that maintains an efficient data structure with a poly-logarithmic ammortized update time and returns a (1/2 - \epsilon)-approximate solution.
We complement our theoretical analysis with an empirical study of the performance of our algorithm.
Our main result is a randomized algorithm that maintains an efficient data structure with a poly-logarithmic ammortized update time and returns a (1/2 - \epsilon)-approximate solution.
We complement our theoretical analysis with an empirical study of the performance of our algorithm.
Research Areas
