Genotypic-Space Mapping: Population Visualization for Genetic Algorithms

This paper presents one proposed method for representing the population data of Genetic Algorithms (GAs). Typical population data from GAs are large high-dimensional sets of binary, decimal, real or string, state values. This makes their representation by two or three spatial dimensions somewhat difficult. Several attempts at population visualization have been made but have failed to efficiently solve this high dimensional to 2/3 dimensional space mapping problem. The use of the proposed "Genotypic-Space Mapping" method is put forward as a solution to this problem. It provides a unique linear mapping of a high-dimensional population string to a pair of x,y and/or z co- ordinates, thus enabling each population to be displayed as a scatter-plot in two or three dimensional space.