Neural Classification of H- and P-version Elements

This paper deals with a comparative performance of traditional elements and high order elements, making use of their formulation as vectors (or patterns) in a multi-dimensional space of proper attributes. The classification can be carried out with the help a self-organizing feature map of Kohonen with the patterns corresponding to the input space. The work makes use of the four attributes: its number of nodes, number of Lejendre terms, maximum degree of interpolation polynomials and number of degrees of freedom per node, though a more general characterization is also possible.