A Unified Framework for Metaheuristics

Over the past decades, a multitude of new search heuristics, often called “metaheuristics” have been proposed, many of them inspired by principles observed in nature. What distinguishes them from random search is primarily that they maintain some sort of memory of the information gathered during the search so far, and that they use this information to select the location where the search space should be tested next. Based on this observation, we propose a general unified framework which is depicted in Fig. 1: A memory is used to construct one or more new solutions which are then evaluated and used to update the memory, after which the cycle repeats.