The problem of parameterization is often central to the effective deploymentof nature-inspired algorithms. However, finding the optimal set of parametervalues for a combination of problem instance and solution method is highlychallenging, and few concrete guidelines exist on how and when such tuning maybe performed. Previous work tends to either focus on a specific algorithm oruse benchmark problems, and both of these restrictions limit the applicabilityof any findings. Here, we examine a number of different algorithms, and studythem in a "problem agnostic" fashion (i.e., one that is not tied to specificinstances) by considering their performance on fitness landscapes with varyingcharacteristics. Using this approach, we make a number of observations on whichalgorithms may (or may not) benefit from tuning, and in which specificcircumstances.
展开▼
