15
103

Submodular Maximization Beyond Non-negativity: Guarantees, Fast Algorithms, and Applications

Abstract

It is generally believed that submodular functions -- and the more general class of γ\gamma-weakly submodular functions -- may only be optimized under the non-negativity assumption f(S)0f(S) \geq 0. In this paper, we show that once the function is expressed as the difference f=gcf = g - c, where gg is monotone, non-negative, and γ\gamma-weakly submodular and cc is non-negative modular, then strong approximation guarantees may be obtained. We present an algorithm for maximizing gcg - c under a kk-cardinality constraint which produces a random feasible set SS such that E[g(S)c(S)](1eγϵ)g(OPT)c(OPT)\mathbb{E} \left[ g(S) - c(S) \right] \geq (1 - e^{-\gamma} - \epsilon) g(OPT) - c(OPT), whose running time is O(nϵlog21ϵ)O (\frac{n}{\epsilon} \log^2 \frac{1}{\epsilon}), i.e., independent of kk. We extend these results to the unconstrained setting by describing an algorithm with the same approximation guarantees and faster O(nϵlog1ϵ)O(\frac{n}{\epsilon} \log\frac{1}{\epsilon}) runtime. The main techniques underlying our algorithms are two-fold: the use of a surrogate objective which varies the relative importance between gg and cc throughout the algorithm, and a geometric sweep over possible γ\gamma values. Our algorithmic guarantees are complemented by a hardness result showing that no polynomial-time algorithm which accesses gg through a value oracle can do better. We empirically demonstrate the success of our algorithms by applying them to experimental design on the Boston Housing dataset and directed vertex cover on the Email EU dataset.

View on arXiv
Comments on this paper