This paper studies the class of polyhedrally tight functions in terms of the basic theorems on convex functions over <n, such as the Fenchel Duality Theorem, Separation Theorem etc. (Polyhedrally tight functions are those for which the inequalities y x ≤ f(y), y ∈ A in x ∈ A∗ with A,A∗ ⊆ <n can be satisfied as equalities for some vector x, not necessarily simultaneously.) It is shown, using results in [6], that the basic theorems hold for polyhedrally tight set functions provided the concerned functions can be extended to convex/concave functionals retaining certain essential features. These essential features carry over only if the functions are compatible in the sense that the normal cone structures of the associated polyhedra are related in a strong way.
[1]
A. Frank.
An Algorithm for Submodular Functions on Graphs
,
1982
.
[2]
H. Hirai.
Greedy Fans : A geometric approach to dual greedy algorithms
,
2005
.
[3]
Satoru Fujishige,et al.
Submodular functions and optimization
,
1991
.
[4]
藤重 悟.
Submodular functions and optimization
,
1991
.
[5]
Jack Edmonds,et al.
Submodular Functions, Matroids, and Certain Polyhedra
,
2001,
Combinatorial Optimization.
[6]
Hariharan Narayanan.
On set functions that can be extended to convex functionals
,
2005
.
[7]
László Lovász,et al.
Submodular functions and convexity
,
1982,
ISMP.