-(Face clssification using Random Forest Technique),,,theri one problem occured during the use of Random-Forest MATLAB code,,,,,

1) what is "ytrain" and "ytest" and how to obtain it for say 20 image samples? 2) random forest would be applied on group of features (multiple images with different features) out of which it is going to extract the best feature.....right! so how exactly it is going o do so?any specified function or command for same in MATLAB? 3)refer to the piece of code given below:

       rfmodel = classRF_train(xtrain,ytrain,  300,20);
       rf_ypred = classRF_predict(xtest,rfmodel);
       rferr = cerror(rf_ypred, ytest'); % evaluate testing error
       fprintf('\nThe error from  Random Forest is %f',rferr)
       rf_record(repeat,1) = 1 - rferr;  
end
 rfCorrectRate = sum(rf_record)/nbrRepeat; 
 global zz
 zz=zz+1;
  correctRate = [rfCorrectRate];
    figure(2);subplot(2,4,zz);
  bar(correctRate,0.3);         %0.3
  set(gca, 'XTickLabel',{'RF '})

what is the role of nbrRepeat?what if the iterations are less? what do we mean by correct rate? every time correct rate shown would be 0.3.why?

Answer 1

[Ilya]

1. ytrain and ytest are likely arrays of true class labels (if you have a classification problem) or arrays of observed continuous response values (if you have a regression problem) for your training and test data. I don't know how you obtain them. Ask someone with expertise in image processing.

2. This classification algorithm is provided in Statistics Toolbox in MATLAB 9a (and later) as TreeBagger.

3. You should ask the author of this script.

[Swati]

whr to find tree bagger in MATLAB..?

[Walter Roberson]

TreeBagger is part of the Stats toolbox in sufficiently new MATLAB (R2009a and later)

http://www.mathworks.com/help/toolbox/stats/treebagger.html