PhD Thesis: Evolutionary Computation Methods for Instance Generation in Optimisation Domains
Alejandro Marrero · ACM SIGEVOlution · 2024
The generation of instances of optimisation problems is a very common task in Computer Science. Traditionally, researchers apply statistical or pseudo-random methods to create instances used to validate their proposals: algorithms or operators. At the same time, some authors have proposed sets known as benchmarks so that new proposals can be evaluated in these instances, thus avoiding the task of generating instances. However, these sets are often characterised by (1) being designed to be hard to solve by off-the-shelf, state-of-the-art algorithms at the time of their creation and (2) by their low diversity, meaning the instances tend to share many similar characteristics.