Indexed by:
Abstract:
In the paper, we study Regularized Submodular Maximization (RegularizedSM) problem over a down-closed family of sets by applying the Lyapunov method. The Regularized Submodular Maximization can be viewed as a generalization of the Submodular Maximization since it adds an extra linear regular term (possibly negative) to the objective so that it may no longer be non-negative. For Regularized Non-monotone Submodular Maximization (RegularizedNSM), we systematically design an algorithm framework in two phases. With a proper choice of coefficients in the framework, a (1/e,γ-γ/e-1γ-1) -approximation algorithm is obtained in the continuous-time phase, where γ∈ [ 0, 1 ) ∪ (1, + ∞) is a parameter reflecting the relative dominance of the positive and negative parts of the linear optimal value. In the second phase, we make the algorithm implemented by discretization with almost the same approximation guarantee and O(n3ϵ) time complexity. Moreover, the Lyapunov method can also be applied in Regularized Monotone Submodular Maximization (RegularizedMSM) with (1 - 1 / e, 1 ) -approximation performance, which coincides with the state-of-the-art result given by Feldman [9]. This observation implies that the algorithm framework designed by the Lyapunov method can unify some of the existing approximation algorithms. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.
Keyword:
Reprint Author's Address:
Email:
Source :
ISSN: 0302-9743
Year: 2024
Volume: 14423 LNCS
Page: 118-143
Language: English
Cited Count:
WoS CC Cited Count: 0
SCOPUS Cited Count:
ESI Highly Cited Papers on the List: 0 Unfold All
WanFang Cited Count:
Chinese Cited Count:
30 Days PV: 2
Affiliated Colleges: