TY - JOUR
T1 - Submodular maximization and its generalization through an intersection cut lens
AU - Xu, Liding
AU - Liberti, Leo
N1 - Publisher Copyright:
© Springer-Verlag GmbH Germany, part of Springer Nature and Mathematical Optimization Society 2024.
PY - 2024/1/1
Y1 - 2024/1/1
N2 - We study a mixed-integer set S:={(x,t)∈{0,1}n×R:f(x)≥t} arising in the submodular maximization problem, where f is a submodular function defined over {0,1}n. We use intersection cuts to tighten a polyhedral outer approximation of S. We construct a continuous extension F¯f of f, which is convex and defined over the entire space Rn. We show that the epigraph epi(F¯f) of F¯f is an S-free set, and characterize maximal S-free sets containing epi(F¯f). We propose a hybrid discrete Newton algorithm to compute an intersection cut efficiently and exactly. Our results are generalized to the hypograph or the superlevel set of a submodular-supermodular function over the Boolean hypercube, which is a model for discrete nonconvexity. A consequence of these results is intersection cuts for Boolean multilinear constraints. We evaluate our techniques on max cut, pseudo Boolean maximization, and Bayesian D-optimal design problems within a MIP solver.
AB - We study a mixed-integer set S:={(x,t)∈{0,1}n×R:f(x)≥t} arising in the submodular maximization problem, where f is a submodular function defined over {0,1}n. We use intersection cuts to tighten a polyhedral outer approximation of S. We construct a continuous extension F¯f of f, which is convex and defined over the entire space Rn. We show that the epigraph epi(F¯f) of F¯f is an S-free set, and characterize maximal S-free sets containing epi(F¯f). We propose a hybrid discrete Newton algorithm to compute an intersection cut efficiently and exactly. Our results are generalized to the hypograph or the superlevel set of a submodular-supermodular function over the Boolean hypercube, which is a model for discrete nonconvexity. A consequence of these results is intersection cuts for Boolean multilinear constraints. We evaluate our techniques on max cut, pseudo Boolean maximization, and Bayesian D-optimal design problems within a MIP solver.
KW - 90C10
KW - 90C26
KW - 90C57
KW - Boolean multilinear functions
KW - D-optimal design
KW - Intersection cuts
KW - MINLP
KW - Submodular maximization
KW - Submodular-supermodular functions
U2 - 10.1007/s10107-024-02059-2
DO - 10.1007/s10107-024-02059-2
M3 - Article
AN - SCOPUS:85185972972
SN - 0025-5610
JO - Mathematical Programming
JF - Mathematical Programming
ER -