Verification of Backtracking and Branch and Bound Design Templates

Abstract

The Design and Analysis of Computer Algorithms is a must of Computer Curricula. It covers many topics that group around several core themes. These themes range from data structures to the complexity theory, but one very special theme is algorithmic design patterns, including greedy method, divide-and-conquer, dynamic programming, backtracking and branch-and-bound. Naturally, all the listed design patterns are taught, learned and comprehended by examples. But they can be semi-formalized as design templates, semi-specified by correctness conditions, and semi-formally verified by means of Floyd method. Moreover, this approach can lead to new insights and better comprehension of the design patterns, specification and verification methods. In this paper we demonstrate an utility of the approach by study of backtracking and branch-and-bound design patterns. In particular, we prove correctness of the suggested templates when the boundary condition is monotone, but the decision condition is anti-monotone on sets of "visited" vertices.