The Discrimination Power of Molecular Identification Numbers Revisited

Abstract

In this paper, we explore the discrimination power of existing molecular ID numbers as the potential of these measures has not yet been investigated on a large scale. First, we find that many ID numbers are computationally insufficient and, hence, can not be calculated on large sets of graphs. Second, we also determine the discrimination power of recently developed eigenvalue-based indices which possess polynomial time complexity. Particularly we find that some of these measures outperform specific molecular ID numbers in terms of their discrimination power.