SDNORM Add a normalization to pipeline returning soft outputs

 Normalization of data set (features sum to one)
   OUT=SDNORM(DATA)

 Normalization added to pipeline PIN
   POUT=SDNORM(PIN)

 INPUT
    DATA    Data set to be normalized
    PIN     Input pipeline returning 2 or more soft outputs

 OUTPUT
    OUT     Normalized data set
    POUT    Output pipeline with added normalization step

 DESCRIPTION
 SDNORM normalizes data sets so that the features sum to one. If a trained
 pipeline is givem SDNORM adds a normalization step to it.  This is useful
 to normalize model output returning probability densities into a
 discriminant providing posteriors.

 EXAMPLE
 Add normalization step so that Parzen returns posterior instead of density
 p=sdparzen(a); pn=sdnorm(-p);
 pr=sdreject(pn,a,'reject',0.1); % reject area of Parzen low confidence

 SEE ALSO
 SDREJECT

sdnorm is referenced in examples:

• kb18: How to protect a trained discriminant against outliers?