K-Nearest Neighbor Classifier for dissimilarity datasets
[W,K] = KNNDC(D,K)
This is the KNN classifier on given dissimilarities. No dissimilarity space nor embedding is applied.
The classifiers optimizes K in the KNN rule for the given dissimilarity dataset D. It should be square. If not K=1 will be used. For executing the classifier on a testset S, formally just K is needed and the classification matrix C with class confidences is computed by TESTKD. KNNDC thereby supplies a routine that may be used in a similar way by routines like CROSSVALD for given dissimilarity datasets.
The classification matrix C can be used for finding labels by LABELD or for error estimation by TESTC or TESTD.