索引超过矩阵尺寸

Question

X= [P(1,:,:);
    P(2,:,:);
    P(3,:,:)];
y= P(4:end,:);
indTrain = randperm(4798);
indTrain = indTrain(1:3838);
trainX= X(indTrain,:);
trainy = y(indTrain); 
indTest = 3839:4798;
indTest(indTrain) = []; 
testX = X(indTest,:);
testy = y(indTest);

It shows error in trainX= X(indTrain,:); saying

Index exceeds matrix dimensions

can anyone pls clarify ? thanks.

by the way I am having a 4x4798 data which my first 3 rows serve as predictors, and last row (4th row) is my response. how would i correctly split the data into the first 3838 columns as my training set and remaining as testing set.

Thanks..!!

Answer 1

Fix your error

To fix the indexing error you need to select the column indices of X , rather than the row indices:

trainX = X(:, indTrain );  

Some words of advice

It seems like your P matrix is 4-by-4798 and it is two dimensional. Therefore, writing P(1,:,:) does select the first row, but it gives the impression as if P is three dimensional because of the extra : at the end. Don't do that . It's bad habit and makes your code harder to read/understand/debug.

X = P(1:3,:); % select all three rows at once
y = P(4,:); % no need for 4:end here - again, gives wrong impression as if you expect more than a single label per x.

Moreover, I do not understand what you are trying to accomplish with indTest(indTrain)=[] ? Are you trying to ascertain that the train and test set are mutually exclusive?
This line will most likely cause an error since the size of your test set is 960 and indTrain contains 1:3838 (randomly permuted) so you will get "index exceeds..." error again.
You already defined your indTrain and indTest as mutually exclusive, no need for another operation. If you want to be extra careful you can use setdiff

indTest = setdiff( indTest, indTrain );