Two-Stage Submodular Maximization Under Knapsack Problem

Abstract

Two-stage submodular maximization problem under cardinality constraint has been widely studied in machine learning and combinatorial optimization. In this paper, we consider knapsack constraint. In this problem, we give $n$ articles and $m$ categories, and the goal is to select a subset of articles that can maximize the function $F(S)$ . Function $F(S)$ consists of $m$ monotone submodular functions $f_{j}, j=1,2, \ldots, m$ , and each $f_{j}$ measures the similarity of each article in category $j$ . We present a constant-approximation algorithm for this problem.