Trajectory-Based Convergence Acceleration of Evolutionary Algorithms

Evolutionary algorithms are heuristic, nature-inspired search methods based on the concept of evolution and survival of the fittest. While they have proven to be effective across a variety of problems they are often inefficient as they do not use information generated during the search and could therefore require extensive computer resources to converge. To address this issue this paper proposes a new method for evolutionary convergence acceleration which is inspired by the method of successive-over-relation for the solution of linear equations sets. The main concept is to determine the direction in which the population centroid has shifted between successive generations, which suggests a favourable direction towards an optimum. The population of solutions is then propagated along that direction to accelerate its convergence. The proposed algorithm is flexible and can be applied to a variety of evolutionary algorithms. An extensive performance analysis based on representative test functions shows the effectiveness of the proposed algorithm.