An empirical approach to the measurement of interchromosomal distances in the genetic algorithm
Robert Collier, Mark Wineberg · 2012
Data visualizations, population diversity measurements, and cluster analyses are all invariably constructed from measures of distance or dissimilarity, and it is recognized that any measure of the distance between points should represent the manner and ease with which an algorithm or process can move from one point towards another. For the genetic algorithm, this traversal is largely accomplished by mutation and recombination, but in spite of this, measures like the Hamming distance and the edit distance are still used to assess the distance between population members. This represents a significant problem, because these measures were not designed with the genetic algorithm in mind and they do not consider how the genetic operators will actually traverse genotypic space. The need for distance measures to be accurate and representative cannot be overstated, but for the complex traversals of the genetic algorithm, it is exceedingly difficult to determine whether one measure is any more representative than another. To address this need, this paper will introduce an empirical approach to distance measurement, and since the resultant values are derived from actual traversals, the distance measured is guaranteed representative, and can be used as a baseline against which other measures can be evaluated.