Extending time-to-target plots to multiple instances

Abstract

Time-to-target plots (ttt-plots) are a useful tool to characterize, evaluate, and compare the behavior of randomized heuristics for a given problem instance of some combinatorial optimization problem. Multiple time-to-target plots (mttt-plots) are their natural extension to sets of multiple instances. We show how to build an mttt-plot from the individual ttt-plots of each instance in the set and we illustrate several case studies to illustrate the applicability and usefulness of the new tool.