Parameterized analysis of bio-inspired computing

Abstract

The parameterized analysis of bio-inspired computing provides a new way of gaining additional insights into the working behavior of popular approaches such as evolutionary algorithms and ant colony optimization. We give an overview of two important approaches in this area. The area of parameterized runtime analysis studies the runtime of bio-inspired computing with respect to different parameters of the given problem instance and builds on the success of rigorous runtime analysis of bio-inspired computing in the last 20 years. The feature-based analysis of algorithms for a given optimization problem uses statistical methods to figure out which features of a given problem instance lead to a good or bad performance of the algorithm under consideration. It often uses an evolutionary algorithm for evolving problem instances that exhibit performance differences between a given set of solvers and can be used for effective algorithm selection.