PRTools contents

PRTools manual

featselm

FEATSELM

Feature selection map

  [W,R] = FEATSELM(A,CRIT,METHOD,K,T,PAR1,...)

Input
 A Training dataset
 CRIT Name of criterion: 'in-in', 'maha-s', 'NN' or others  (see FEATEVAL) or an untrained classifier V (default: 'NN')
 METHOD 'forward' : selection by featself (default)
-  'float' : selection by featselp
-  'backward': selection by featselb
-  'b&b' : branch and bound selection by featselo
-  'ind' : individual
-  'lr' : plus-l-takeaway-r selection by featsellr
-  'sparse' : use sparse untrained classifier CRIT
 K Desired number of features (default: K = 0, return optimal set)
 T Tuning set to be used in FEATEVAL (optional)
 PAR1,.. Optional parameters:
L,R : for 'lr' (default: L = 1, R = 0)

Output
 W Feature selection mapping
 R Matrix with step by step results

Description

Computation of a mapping W selecting K features. This routines offers a  central interface to all other feature selection methods. W can be used  for selecting features in a dataset B using B*W.

See also

mappings, datasets, feateval, featselo, featselb, featseli, featselp, featself, featsellr,

PRTools contents

PRTools manual