To solve the problem of efficiently identifying certain structural characteristics in large sets of data, an algorithm is developed which enables, in any Euclidean space, the determination of a subset of a point set A which is small, but nonetheless meaningful for certain structural characteristics. This so-called set of Characteristic Vertices of A is determined as the solution of a non-linear optimization problem from the set of extreme points of the convex hull of A.
In order to illustrate that this approach enables the complexity of a pattern recognition problem to be reduced, an algorithm is developed to determine approximate n-symmetries of point sets in Euclidean spaces of any dimension. By assigning symmetry values, this algorithm is particularly suitable to enable the identification of incomplete or projectively distorted symmetries.
«To solve the problem of efficiently identifying certain structural characteristics in large sets of data, an algorithm is developed which enables, in any Euclidean space, the determination of a subset of a point set A which is small, but nonetheless meaningful for certain structural characteristics. This so-called set of Characteristic Vertices of A is determined as the solution of a non-linear optimization problem from the set of extreme points of the convex hull of A.
In order to illustrate t...
»