Title:Finding the Submodularity Hidden in Symmetric Difference

Abstract:A fundamental property of convex functions in continuous space is that the convexity is preserved under affine transformations. A set function $f$ on a finite set $V$ is {\em submodular} if $f(X) + f(Y) \geq f(X \cup Y) - f(X \cap Y)$ for any pair $X, Y \subseteq V$. The {\em symmetric difference transformation} ({\em SD-transformation}) of $f$ by a {\em canonical set} $S \subseteq V$ is a set function $g$ given by $g(X) = f(X \vartriangle S)$ for $X \subseteq V$, where $X \vartriangle S = (X \setminus S) \cup (S \setminus X)$ is the {\it symmetric difference} between $X$ and $S$. Despite that submodular functions and SD-transformations are regarded as counterparts of convex functions and affine transformations in finite discrete space, not all SD-transformations do not preserve the submodularity. Starting with a characterization of SD-stransformations that preserve the submodularity, this paper investigates the problem of discovering a canonical set $S$, given the SD-transformation $g$ of a submodular function $f$ by $S$, provided that $g(X)$ is given by an oracle. A submodular function $f$ on $V$ is said to be {\it strict} if $f(X) + f(Y) > f(X \cup Y) - f(X \cap Y)$ holds whenever both $X \setminus Y$ and $Y \setminus X$ are nonempty. We show that the problem is solvable by using ${\rm O}(|V|)$ oracle calls when $f$ is strictly submodular, although it requires exponentially many oracle calls in general.