EDICON

Multi-edit and condense a training set

    J = EDICON(D,NSETS,NITERS,NTRIES)

Input
D	Distance matrix dataset
NSETS	Number of subsets for editing, or [] for no editing (default: 3)
NITERS	Number of iterations for editing (default: 5)
NTRIES	Number of tries for condensing, or [] for no condensing (dflt: 10)

Output
J	Indices of retained samples

Description

Returns the set of objects J such that the nearest neigbour gives zero error  on the remaining objects. If MODE = 0, multi-edit the dataset represented by  distance matrix D first. D can be computed from a dataset A by A*proxm(A).

Reference(s)

Devijver, P. and Kittler, J. "Pattern recognition", Prentice-Hall, 1982.

See also

dataset, knnc, proxm,