Index in position 2 exceeds array bounds using crossval

Dear all,

I got the following error when using crossval.

Error using crossval>evalFun (line 488)
The function '@(Xtr,Ytr,Xte)predict(fitlm(Xtr,Ytr(:,2)),Xte)' generated the following error:
Index in position 2 exceeds array bounds (must not exceed 1).

Error in crossval>getLossVal (line 525)
funResult = evalFun(funorStr,arg(1:end-1));

Error in crossval (line 424)
                [funResult,outarg] = getLossVal(i, nData, cvp, data, predfun);

This is my code. The X and Y are stored in the attached .mat file.

%% Cross validation

hpartition = cvpartition(81,'Holdout',0.2);
idxTrain = training(hpartition);
Xtr = X(idxTrain,:);
Ytr = Y(idxTrain,:);
idxNew = test(hpartition);
Xte = X(idxNew,:);

for i = 1:2
    for j = 1:3
        fcn{i,1} = @(Xtr, Ytr, Xte) predict(fitlm(Xtr,Ytr(:,i)), Xte);
        fcn{i,2} = @(Xtr, Ytr, Xte) predict(fitrsvm(Xtr,Ytr(:,i)), Xte);
        fcn{i,3} = @(Xtr, Ytr, Xte) predict(fitrgp(Xtr,Ytr(:,i)), Xte); 
        mse(i,j) = crossval('mse', X, Y(:,i),'Predfun',fcn{i,j}, 'kfold',10);
    end
end

Answer 1

[Nagasai Bharat]

Hi,

The problem in above code was that crossval does the train-test spilt and it takes in the Ytr as column vector but you did provide Ytr are a 2-D Matrix due to which we were getting that array bound error.The below code snippet should resolve and simplify your issue.

load data
% hpartition = cvpartition(81,'Holdout',0.2);
% idxTrain = training(hpartition);
% Xtr = X(idxTrain,:);
% Ytr = Y(idxTrain,:);
% % Ytr_1 = Y(idxTrain,1);
% % Ytr_2 = Y(idxTrain,2);
% idxNew = test(hpartition);
% Xte = X(idxNew,:);

for i=1:2
    for j = 1:3
        fcn{i,1} = @(Xtr, Ytr, Xte) predict(fitlm(Xtr,Ytr), Xte);
        fcn{i,2} = @(Xtr, Ytr, Xte) predict(fitrsvm(Xtr,Ytr), Xte);
        fcn{i,3} = @(Xtr, Ytr, Xte) predict(fitrgp(Xtr,Ytr), Xte); 
        mse(i,j) = crossval('mse', X, Y(:,i),'Predfun',fcn{i,j}, 'kfold',10);
    end

For more information you could follow crossval documentation.

[Tessa Kol]

Thank for your reply. I discovered indeed that issue was the 2D-matrix. I forgot to close this post after I resolved the issue myself. But thank you so much for providing the answer!