Using PageRank to Uncover Patterns in Search Behavior Induced by the Bit Flip Operator
Understanding how the complex interactions of the problem-algorithm combination lead to an algorithm's search performance is arguably one of the most important open questions in metaheuristic algorithm theory. Examination of the fitness landscape does not provide all of the information needed to understand a metaheuristic algorithm's search behavior. We introduce an extension to the fitness landscape, which we call a search behavior diagram, that models a metaheuristic algorithm's expected search behavior across an entire fitness landscape. We then show that analyzing a search behavior diagram can produce insights into the nature of metaheuristic algorithm search behavior on problems from binary optimization, including one interesting insight about the relationship between the distribution of optima and adaptive search behavior.
Green, Thomas M. and Andersen, Timothy L.. (2020). "Using PageRank to Uncover Patterns in Search Behavior Induced by the Bit Flip Operator". GECCO '20: Proceedings of the 2020 Genetic and Evolutionary Computation Conference Companion, 1866-1871. https://doi.org/10.1145/3377929.3398140