English

On the Complexity of Dynamic Submodular Maximization

Data Structures and Algorithms 2022-04-19 v2 Machine Learning

Abstract

We study dynamic algorithms for the problem of maximizing a monotone submodular function over a stream of nn insertions and deletions. We show that any algorithm that maintains a (0.5+ϵ)(0.5+\epsilon)-approximate solution under a cardinality constraint, for any constant ϵ>0\epsilon>0, must have an amortized query complexity that is polynomial\mathit{polynomial} in nn. Moreover, a linear amortized query complexity is needed in order to maintain a 0.5840.584-approximate solution. This is in sharp contrast with recent dynamic algorithms of [LMNF+20, Mon20] that achieve (0.5ϵ)(0.5-\epsilon)-approximation with a polylog(n)\mathsf{poly}\log(n) amortized query complexity. On the positive side, when the stream is insertion-only, we present efficient algorithms for the problem under a cardinality constraint and under a matroid constraint with approximation guarantee 11/eϵ1-1/e-\epsilon and amortized query complexities O(log(k/ϵ)/ϵ2)\smash{O(\log (k/\epsilon)/\epsilon^2)} and kO~(1/ϵ2)logn\smash{k^{\tilde{O}(1/\epsilon^2)}\log n}, respectively, where kk denotes the cardinality parameter or the rank of the matroid.

Keywords

Cite

@article{arxiv.2111.03198,
  title  = {On the Complexity of Dynamic Submodular Maximization},
  author = {Xi Chen and Binghui Peng},
  journal= {arXiv preprint arXiv:2111.03198},
  year   = {2022}
}
R2 v1 2026-06-24T07:27:02.828Z