Computational Aspects of the Colorful Carathéodory Theorem

Let P1, . . . , Pd+1 ⊂ R be d-dimensional point sets such that the convex hull of each Pi contains the origin. We call the sets Pi color classes, and we think of the points in Pi as having color i. A colorful choice is a set with at most one point of each color. The colorful Carathéodory theorem guarantees the existence of a colorful choice whose convex hull contains the origin. So far, the computational complexity of finding such a colorful choice is unknown. We approach this problem from two directions. First, we consider approximation algorithms: an m-colorful choice is a set that contains at most m points from each color class. We show that for any fixed ε > 0, an dεde-colorful choice containing the origin in its convex hull can be found in polynomial time. This notion of approximation has not been studied before, and it is motivated through the applications of the colorful Carathéodory theorem in the literature. In the second part, we present a natural generalization of the colorful Carathéodory problem: in the Nearest Colorful Polytope problem (NCP), we are given sets P1, . . . , Pn ⊂ R that do not necessarily contain the origin in their convex hulls. The goal is to find a colorful choice whose convex hull minimizes the distance to the origin. We show that computing local optima for the NCP problem is PLS-complete, while computing a global optimum is NP-hard. 1998 ACM Subject Classification F.2.2 Nonnumerical Algorithms and Problems – Geometrical problems and computations