BALLS5D

Generate a moderate non-Euclidean dissimilarity dataset.

    D = BALLS5D

This PRTools dataset has been generated by the DisTools command  GENBALLD([500 500 500 500],5,[0.02 0.04 0.06 0.08]) which generates the  given numbers of 5D balls with sizes [0.02 0.04 0.06 0.08]in a 5D hypercube with edge sizes 100. Balls do not overlap.  Dissimilarities are computed as the shortest distance between two points  on the surface of two balls. The intention is to study strong examples in  which non-Euclidean dissimilarities are informative.

Reference(s)

E. Pekalska, A. Harol, R.P.W. Duin, D. Spillman, and H. Bunke, Non-Euclidean or non-metric measures can be informative, in: D.-Y. Yeung et al., Proc. SSSPR2006 Lecture Notes in Comp. Sc., vol. 4109, Springer, Berlin, 2006, 871-880.

R.P.W. Duin, E. Pekalska, A. Harol, W.J. Lee, and H. Bunke, On Euclidean corrections for non-Euclidean dissimilarities, in: N. da Vitoria Lobo et al., Proc. SSSPR2008, Lecture Notes in Comp.Sc., vol. 5342, Springer, Berlin, 2008, 551-561.

J. Laub, V. Roth, J.M. Buhmann, K.R. Mueller, On the information and representation of non-euclidean pairwise data, Pattern Recognition, vol. 39, 2006, 1815-1826.

See also

prtools, datasets, prdisdata, balls3d, balls50d,