Incremental gravitational search algorithm for high-dimensional benchmark functions

摘要

In this study, gravitational search algorithm (GSA), which is a powerful optimization algorithm developed in recent years and based on physics, is improved by integrating the incremental social learning structure. In this improvement, new agents have been added to the GSA starting from the first population at certain steps, agent insertion on the maximum population number has been terminated, and the search has been continued until the desired function call is accomplished. This improved algorithm, which is a recent version of the GSA, is named as the incremental gravitational search algorithm (IGSA). The process of adding agent to the population has been performed with three different approaches. Results of the 30-dimensional test functions, which are solved by the GSA in the literature, are compared with the obtained results of IGSA, developed for each approach. Thereafter, the dimensions of the same test functions have been increased (50 and 100 dimensions) and resolved with IGSA, and the results are discussed.