English

Streaming algorithms for Budgeted $k$-Submodular Maximization problem

Data Structures and Algorithms 2021-10-25 v3 Computer Science and Game Theory

Abstract

Stimulated by practical applications arising from viral marketing. This paper investigates a novel Budgeted kk-Submodular Maximization problem defined as follows: Given a finite set VV, a budget BB and a kk-submodular function f:(k+1)VR+f: (k+1)^V \mapsto \mathbb{R}_+, the problem asks to find a solution \s=(S1,S2,,Sk)\s=(S_1, S_2, \ldots, S_k), each element eVe \in V has a cost ci(e)c_i(e) to be put into ii-th set SiS_i, with the total cost of ss does not exceed BB so that f(\s)f(\s) is maximized. To address this problem, we propose two streaming algorithms that provide approximation guarantees for the problem. In particular, in the case of each element ee has the same cost for all ii-th sets, we propose a deterministic streaming algorithm which provides an approximation ratio of 14ϵ\frac{1}{4}-\epsilon when ff is monotone and 15ϵ\frac{1}{5}-\epsilon when ff is non-monotone. For the general case, we propose a random streaming algorithm that provides an approximation ratio of min{α2,(1α)k(1+β)kβ}ϵ\min\{\frac{\alpha}{2}, \frac{(1-\alpha)k}{(1+\beta)k-\beta} \}-\epsilon when ff is monotone and min{α2,(1α)k(1+2β)k2β}ϵ\min\{\frac{\alpha}{2}, \frac{(1-\alpha)k}{(1+2\beta)k-2\beta} \}-\epsilon when ff is non-monotone in expectation, where β=maxeV,i,j[k],ijci(e)cj(e)\beta=\max_{e\in V, i , j \in [k], i\neq j} \frac{c_i(e)}{c_j(e)} and ϵ,α\epsilon, \alpha are fixed inputs.

Keywords

Cite

@article{arxiv.2109.08863,
  title  = {Streaming algorithms for Budgeted $k$-Submodular Maximization problem},
  author = {Canh V. Pham and Quang C. Vu and Dung K. T. Ha and Tai T. Nguyen},
  journal= {arXiv preprint arXiv:2109.08863},
  year   = {2021}
}

Comments

There are some results of the article that need to be corrected