PRTools contents

PRTools manual

gensubsets

GENSUBSETS

Generate sequence of nested training sets

     [L,R] = GENSUBSETS(NLAB,S)
     [L,R] = GENSUBSETS(A,S)

Input
 NLAB Column vector of numeric labels of some dataset A.
 NLAB = GETNLAB(A)
 A Dataset for which subsets are to be created
 S Array of growing subset sizes.
 S(K,J) should specify the size of training set K for class J with  numeric label J.

Output
 L Cell array of length SIZE(S,1)+1 containing a series of growing  sets of indices or datasets. Datasets can be reconstructed from  indices by A(L{K},:). The last element of L refers to the  original dataset A
 R Cell array of length SIZE(S,1)+1 containing a series of shrinking  sets of indices or datasets. Datasets can be reconstructed from  indices by A(R{K},:). The last element of R is empty.

Description

Learning curves of classifier performances should be based on a  consistent set of training sets, such that training set K1 is a subset of  training set K2 if K1 < K2. This routine generates such a set on the  basis of the numeric labels of the dataset A. L refers to the selected  objects and R to the deselected ones.

See also

datasets, gendat,

PRTools contents

PRTools manual