Efficient Configuration of Optimization Algorithms

We propose a set of capping methods to speed-up the automatic configuration of optimization algorithms. First, we build a performance envelope based on previous executions of known configurations, which defines the minimum required performance for new configurations. Then, we use the performance envelope to evaluate new configurations, stopping poor performers early. We propose different methods to aggregate previous executions into a performance envelope, and evaluate them on several configuration scenarios. The proposed methods produce solutions of the same or better quality as configuring without capping, but reduce the effort required to configure optimization algorithms.