The KPCA Procedure

LRAPPROXIMATION Statement

  • LRAPPROXIMATION <options>;

The LRAPPROXIMATION statement specifies the parameters of the low-rank approximation method.

You can specify the following options:

CCRITERION=number

specifies a double value for the convergence criterion in k-means clustering. By default, CCRITERION=0.02.

CLUSMETHOD=FC | KMPP | RANDOM

specifies the method of seed initialization in k-means clustering to generate the centroids. You can specify one of the following seed initialization options:

FC

stands for fast clustering and spreads out the initial seeds in a heuristic way.

KMPP / <RANDOMSEED=number>

spreads out the initial seeds by introducing a weighted probability distribution in seed selection. For more information, see Arthur and Vassilvitskii (2007).

RANDOM / <RANDOMSEED=number NRESTART=number>

selects initial seeds by pure randomization.

Note that if CLUSMETHOD=KMPP or RANDOM, you can specify an integer value for the RANDOMSEED= option. If you do not specify a value for the RANDOMSEED= option, the seed is generated by reading the time of day from the computer’s clock. If you want reproducible results, specify an integer for number. Also, if CLUSMETHOD=RANDOM, you can specify the number of restarts by using the NRESTART= option, whose value should be an integer. By default, NRESTART=5. Also, by default, CLUSMETHOD=KMPP.

MAXCLUS=number

specifies the maximum number of clusters to use in k-means clustering. By default, MAXCLUS=100.

MAXITER=number

specifies the maximum number of iterations to use in k-means clustering. By default, MAXITER=50.

Last updated: November 11, 2020