A New Search Strategy in Branch-and-Bound Algorithms

Abstract

The branch-and-bound method is used in many exact algorithms for optimization problems. In most case, those algorithms are implemented with the depth rst search. Other search strategies are rarely used because they require large storage areas. In this paper, we propose a new search strategy, the widening search. The widening search greedily nds a solution at rst, and then gradually widens the search space. The storage size required by the widening search is almost same as the depth rst search. We implemented the branch-and-bound algorithm for the maximum weight clique problem with the depth rst search and the widening search and compared their performances. Experimental results show that the solutions by the widening search are much better than the solutions by the depth rst search.