Automatically Configuring Algorithms for Scaling Performance

Automated algorithm configurators have been shown to be very effective for finding good configurations of high performance algorithms for a broad range of computationally hard problems. As we show in this work, the standard protocol for using these configurators is not always effective. We propose a simple and computationally inexpensive modification to this protocol and apply it to state-of-the-art solvers for two prominent problems, TSP and computer Go playing, where the standard protocol is unable or unlikely to yield performance improvements, and one problem, mixed integer programming, where the standard protocol is known to be effective. We show that our new protocol is able to find configurations between 5% and 75% better than the standard protocol within the same time budget.