Discrete Convex Analysis Survey of the Present Status.pdfVIP

Discrete Convex Analysis | Survey of the Present Status | Kazuo MUROTA Research Institute for Mathematical Sciences of Kyoto University Kyoto 606-8502, Japan murota@kurims.kyoto-u.ac.jp Abstract: The present knowledge ab out L-convex and M-convex functions are surveyed on the basis of [27]. Keywords: Discrete convexity, Submo dular function, Matroid Intro duction if p 2 D, and = +1 otherwise) is an L-convex function, i.e., if (i) D =6 ; , (ii) p;q 2 D ) p _ In the eld of nonlinear programming (in contin- q;p ^ q 2 D, and (iii) p 2 D ) p  1 2 D. uous variables) convex analysis [30, 31] plays a V A function f : Z ! Z [ f+1g with dom f =6 ; pivotal role b oth in theory and in practice. An is called M-convex if it satises analogous theory for discrete optimization (non- linear integer programming), called \discrete con- + (M-EXC) For x;y 2 dom f and u 2 supp (x y), vex analysis [24, 25], is develop ed for L-convex there exists v 2 supp (x y) such that and M-convex functions by adapting the ideas in convex analysis and generalizing the results in f(x) + f(y)  f(x  +  ) + f(y +   ) u v u v matroid theory. The L- and M-convex functions where, for any u 2 V ,  is t