PRTools contents

PRTools manual

scalem

SCALEM

Compute scaling map

    W = SCALEM(A,T)

Input
 A Dataset
 T Type of scaling (optional; default: the class priors weighted mean of A is shifted to the origin)

Output
 W Scaling mapping

Description

Computes a scaling map W, whose type depends on the parameter T

 [], 'c-mean' - The mean of A is shifted to the origin. Class priors are taken  into account.
'mean'
  • The mean of A is shifted to the origin. This is computed for  the data as given in A, neglecting class priors.
'c-variance'
  • The mean of A is shifted to the origin and the average class  variances (within-scatter) are normalized. Class priors are  taken into account.
'variance'
  • The mean of A is shifted to the origin and the total variances  of all features are scaled to 1. This is computed for the  data as given in A, neglecting class priors.
'2-sigma'
  • For each feature f, the 2-sigma interval, [f-2*std(f),f+2*std(f)]  is rescaled to [0,1]. Values outside this domain are clipped.  Class priors are taken into account.
'domain'
  • The domain for all features in the dataset A is set to [0,1].  This is computed for the data as given in A, neglecting class priors.

The map W may be applied to a new dataset B by B*W.

See also

mapping, dataset,

PRTools contents

PRTools manual