| PRTools contents |
KCENTRES
[LAB,J,DM] = KCENTRES(D,K,N)
| Input | |
| D | Distance matrix between, e.g. M objects (may be a dataset) |
| K | Number of center objects to be found (optional; default: 1) |
| N | Number of trials starting from a random initialization (optional; default: 1) |
| Output | |
| LAB | Integer labels: each object is assigned to its nearest center |
| J | Indices of the center objects |
| DM | A list of distances corresponding to J: for each center in J the maximum distance of the objects assigned to this center. |
Finds K center objects from a symmetric distance matrix D. The center objects are chosen from all M objects such that the maximum of the distances over all objects to the nearest center is minimized. For K > 1, the results depend on a random initialization. The procedure is repeated N times and the best result is returned.
If N = 0, initialisation is not random, but done by a systematic selection based on a greedy approach.
hclust, kmeans, emclust, modeseek,
| PRTools contents |