Clean dataset for small class size behavior of classifiers
[B,M,K,C,LABLIST,L,W] = CLEANDSET(A,N,U)
| A|| Dataset|
| N|| Minimum desired class size, default 1 |
| U|| Untrained fallback classifier, default ONEC |
| B|| Dataset with small and empty classes removed|
| M|| Number of objects in B |
| K|| Feature size of B |
| C|| Number of classes in B |
| LABLIST|| Label list of A |
| L|| Classes of A still availiable in B |
| W|| Trained fallback classifier|
This routine serves three purposes
- It summarises a number of statements in the training parts of a classifier in orer to make the source more readable.
- Removal of small classes.
- In case B does not contain at least two classes of the desired sample size, the fallback classifier U is trained by A and returned in W.
This routine takes facilitates the handling of imcomplete training sets, together with the support routines ONEC, ALLCLASS and, CLASSUSE.
datasets, mappings, onec, allclass, classuse,
|This file has been automatically generated. If badly readable, use the help-command in Matlab.|