NeurIPS 2024poster0 citations

Gradient Methods for Online DR-Submodular Maximization with Stochastic Long-Term Constraints

Guanyu Nie, Vaneet Aggarwal, Christopher John Quinn

Abstract

In this paper, we consider the problem of online monotone DR-submodular maximization subject to long-term stochastic constraints. Specifically, at each round $t\in [T]$, after committing an action $\mathbf{x}_t$, a random reward $f_t(\mathbf{x}_t)$ and an unbiased gradient estimate of the point $\widetilde{\nabla}f_t(\mathbf{x}_t)$ (semi-bandit feedback) are revealed. Meanwhile, a budget of $g_t(\mathbf{x}_t)$, which is linear and stochastic, is consumed of its total allotted budget $B_T$. We propose a gradient ascent based algorithm that achieves $\frac{1}{2}$-regret of $\mathcal{O}(\sqrt{T})$ with $\mathcal{O}(T^{3/4})$ constraint violation with high probability. Moreover, when first-order full-information feedback is available, we propose an algorithm that achieves $(1-1/e)$-regret of $\mathcal{O}(\sqrt{T})$ with $\mathcal{O}(T^{3/4})$ constraint violation. These algorithms significantly improve over the state-of-the-art in terms of query complexity.

DR-submodularlong term constraintgradient ascent
BibTeX
@inproceedings{
nie2024gradient,
title={Gradient Methods for Online {DR}-Submodular Maximization with Stochastic Long-Term Constraints},
author={Guanyu Nie and Vaneet Aggarwal and Christopher John Quinn},
booktitle={The Thirty-eighth Annual Conference on Neural Information Processing Systems},
year={2024},
url={https://openreview.net/forum?id=PTxRRUEpHq}
}