Accelerating the Convergence of Evolutionary Algorithms by Trajectory Analysis

Evolutionary algorithms are heuristic nature-inspired search methods based on the concepts of adaptation and survival of the fittest. While they have proven to be effective across varied problems they are often inefficient, namely, they may be slow to converge and require a large number of function evaluations to yield a satisfactory solution. To address these issues this paper proposes a new algorithm to accelerate the EA convergence based on the trajectory traversed by the EA population during the search. Based on this trajectory a vector is derived which approximately points to an optimum and the population is then shifted along it to bring it closer to an optimum thereby accelerating convergence. A numerical performance evaluation shows that the proposed algorithm was effective across different optimization problems.