Random lines: a novel population set-based evolutionary global optimization algorithm

  • Authors:

  • İsmet Şahin

  • Affiliations:

  • Department of Materials Science and Engineering, University of Maryland, College Park, MD

  • Venue:
  • EuroGP'11 Proceedings of the 14th European conference on Genetic programming
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, we present a new population set-based evolutionary optimization algorithm which aims to find global minima of cost functions. This algorithm creates random lines passing through pairs of points (vectors) in population, fits a quadratic function based on three points on each line, and then applies the crossover operation to extrema of these quadratic functions, and lastly performs the selection operation. We refer to the points determining random lines as parent points and the extremum of a quadratic model as the descendant or mutated point under some conditions. In the crossover operation, some entries of a descendant vector are randomly replaced with the corresponding entries of one parent vector and some other entries of the descendant vector are replaced with the corresponding entries of the other parent vector based on the crossover constant. The above crossover and mutation operations make this algorithm robust and fast converging. One important property of this algorithm is that its robustness in general increases with increasing population size which may become useful when more processing units are available. This algorithm achieves comparable results with the well-known Differential Evolution (DE) algorithm over a wide range of cost functions.