PRTools Contents

PRTools User Guide

rfishercc

RFISHERCC

Fisher combining of randomly generated classifiers

     W = RFISHERCC(A,N)
     W = A*RFISHERCC(N)

Input
 A prdataset to be used for training, M objects, C classes
 N number of base classifiers to be generated  default: M/10, <= 100.

Output
 W trained classifier

Description

This routine generates a random set of N simple classifiers, based on  the 1-NN rule using a single, randomly selected object per class. The  confindences (see KNNC) for the total training set A (in total N*(C-1)  per object) are used to train a combiner using FISHERC.

Example(s)

 a = gendatb;
 figure; scatterd(a); 
 plotc(a*rfishercc)

 a = gendatm; 
 figure; scatterd(a); 
 plotc(a*rfishercc(2),'col')

 a = setprior(sonar,0); % make priors equal
 w1 = setname(rfishercc(10),'RFisher-10');
 w2 = setname(rfishercc(20),'RFisher-20');
 w3 = setname(rfishercc(40),'RFisher-40');
 randreset(1); % for reproducability
 e = cleval(a,{w1,w2,w3},[5,10,20,40,80],10);
 plote(e);

See also

mappings, datasets, knnc, fisherc,

PRTools Contents

PRTools User Guide

This file has been automatically generated. If badly readable, use the help-command in Matlab.