Visualization of Binary String Convergence by Sammon Mapping

Understanding the evolution of a complex genetic algorithm is a non-trivial problem, however, genetic-algorithm visualization is in its infancy. This paper reviews some of the current approaches and presents a new visualization approach based on Sammon mapping. Sammon mapping is a nonlinear mapping of a set of vectors in p-dimensional space to a set in r-dimensional space, where r < p. The mapping attempts to preserve in r-space the Euclidean inter-vector distances present in p-space. We demonstrate that a Sammon mapping to 2-space of binary chromosomes present in a higher-dimensional allele space during the execution of a genetic algorithm can indicate the presence of multiple solutions. Shortfalls of this approach are discussed along with possible solutions.