LABNN

Find indices of nearest neighbors

    [LABELS,NND] = LABNN(TEST,TRAIN,K,PROC)
    [LABELS,NND] = LABNN(TEST,CLASSF,K,PROC)
    [LABELS,NND] = LABNN(TEST,K,PROC)

Input
TEST	Dataset with M objects, doubles or PRTools dataset
TRAIN	Dataset with N objects, doubles or PRTools dataset
CLASSF-	A classifier trained by KNNC. Its training set will be used.
K	Neighbor number to be found
PROC	String indicating procedure to be used in the search, see below.

Output
LABELS	Array with size MxK of indices pointing for each element of
TEST	to the K nearest neighbors in TRAIN.
NND	Vector of length M with nearest neighbor distances.

Description

In PROC one of the following routines can be defined

• 'prtools' exact computation using PRTools
• 'ann' approximate computation, needs the ANN Matlab Wrapper
• 'kdtree' using Matlab's KNNSEARCH with kdtree
• 'exhaustive' using Matlab's KNNSEARCH with kdtree (default)

In case TRAIN is omitted LABELS point to the nearest neighbors of TEST to itself by a leave-one-out procedure. This is defined only for PROC is 'prtools' only.

See also

datasets, mappings, prmem, proxm, knnsearch,