FEATSELP

Trainable mapping for Pudil's floating feature selection

    [W,R] = FEATSELP(A,CRIT,K,T)
    [W,R] = A*FEATSELP([],CRIT,K,T)
    [W,R] = A*FEATSELP(CRIT,K,T)
    [W,R] = FEATSELP(A,CRIT,K,N)
    [W,R] = A*FEATSELP([],CRIT,K,N)
    [W,R] = A*FEATSELP(CRIT,K,N)

Input
A	Training dataset
CRIT	Name of the criterion or untrained mapping  (default: 'NN', 1-Nearest Neighbor error)
K	Number of features to select (default: K = 0, select optimal set)
T	Tuning dataset (optional)
N	Number of cross-validation folds (optional)

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

Description

Forward floating selection of K features using the dataset A. CRIT sets  the criterion used by the feature evaluation routine FEATEVAL. If the  dataset T is given, it is used as test set for FEATEVAL. Alternatively  a number of cross-validations N may be supplied. For K = 0, the optimal  feature set (maximum value of FEATEVAL) is returned. The result W can  be used for selecting features in a dataset B using B*W.  The selected features are stored in W.DATA and can be found by +W.

Note: this routine is highly time consuming.

In R the search is reported step by step

     R(:,1) : number of features
     R(:,2) : criterion value
     R(:,3) : added / deleted feature

See also

mappings, datasets, feateval, featselo, featselb, featseli, featsel, featself, featselm,