On the relativity in the assessment of blind optimization algorithms and the problem-algorithm coevolution
Carlos D. Toledo-Suárez, Manuel Valenzuela-Rendón, Hugo Terashima‐Marín, Eduardo Uresti-Charre · 2007
Considering as an optimization problem the one of knowing what is hard for a blind optimization algorithm, the usefulness of absolute algorithm-independent hardness measures is called into question, establishing as a working hypothesis the relativity in the assessment of blind search. The results of the implementation of an incremental coevolutionary algorithm for coevolving populations of tunings of a simple genetic algorithm and simulated annealing, random search and 20-bit problems are presented, showing how these results are related to two popular views of hardness for genetic search: deception and rugged fitness landscapes.