Empirical analysis of a micro-evolutionary algorithm for numerical optimization

Abstract

This paper presents an empirical comparison of some evolutionary algorithms to solve numerical optimization problems. The aim of the paper is to test a micro-evolutionary algorithm called Elitist evolution, originally designed to work with small populations, on a set of diverse test problems (unimodal, multimodal, separable, non-separable, shifted, and rotated) with different dimensionalities. The comparison covers micro-evolutionary algorithms based on differential evolution and particle swarm optimization. The number of successful runs, the quality of results and the computational cost, measured by the number of evaluations required to reach the vicinity of the global optimum, are used as performance criteria. Furthermore, a comparison against a state-of-the-art algorithm is presented. The obtained results suggest that the Elitist evolution is very competitive as compared with other algorithms, especially in high-dimensional search spaces.

Micro-differential evolution: Diversity enhancement and a comparative study

“…Such algorithms work based on a set of individuals, where optimal size setting of this parameter is imperative for algorithm performance [1]. Large population size setting in such algorithms supports a higher diversity of the population, which recombination of its diverse members offers a higher opportunity to the optimizer to locate the global solution(s) [2]- [4]. Although this diversity enhancement technique offers a better exploration of problem landscape, yet admits more function evaluations and as a consequence, lower convergence rate to the possible solution [2].…”

“…The term micro-algorithm, denoted by µ-algorithm, refers to population-based algorithms with a small population size [4]. The micro-algorithms have been used in diverse applica-1.2 Objectives 3 tions, exceptionally due to their lighter hardware requirements and opportunity to operate in embedded systems with a memory saving approach [1].…”

Efficiently Finding the Optimum Number of Clusters in a Dataset with a New Hybrid Cellular Evolutionary Algorithm

“…As described in the state of the art, µEAs can be classified into two classes: (1) those that are modified versions of original EAs and (2) those specifically designed to work with small populations [40]. The most representative algorithms of the first class are µGA [24], µPSO [4], and µDE [34].…”

Micro-differential evolution with vectorized random mutation factor

“…A population-based algorithms with a small population size is called a micro-algorithm, some times denoted by µ-algorithm [4]. The micro-algorithms due to their low populations size and less hardware demand than their parent algorithms can be utilized to solve many real-time problems such as in wireless communication networks [3], [5], or vehicle navigation systems [6].…”

“…During stagnation, the population remains unconverged but divert and the optimization process does not progress. A large population size offers a more diversified pool of individuals whose recombination offers higher likelihood to locate the global solution [2]- [4]. Therefore, reducing the population size while raising the diversity of the population is a key point to achieve a faster convergence speed while maintaining a low risk of premature convergence or stagnation.…”