A computational comparison of branch and bound and outer approximation algorithms for 0–1 mixed integer nonlinear programs

In this paper we compare the performance of two codes for convex 0–1 mixed integer nonlinear programs on a number of test problems. The first code uses a branch and bound algorithm. The second code is a commercially available implementation of an outer approximation algorithm. The comparison demonstrates that both approaches are generally capable of solving the test problems. However, there are significant differences in the robustness of the two codes and their performance on different classes of problems.