Leave-One-​Out-Cross-​Validation

Hi, I need to do a Leave-One-Out-Cross-Validation, but I dont even know how to start.

I read cvpartition etc.

Can someone help me please?

Thanks!

[John D'Errico]

Which one logical thing? Ask the question, rather than asking for someone to start a completely open consulting relationship on Answers. That is something that will never end. Ask a SPECIFIC question.

https://www.mathworks.com/matlabcentral/answers/6200-tutorial-how-to-ask-a-question-on-answers-and-get-a-fast-answer

[Adam Danz]

Start here. The first link is the syntax you'll use. The second link is an example using this method on data.

c = cvpartition(n,'Leaveout')

https://www.mathworks.com/help/stats/cvpartition.html#mw_d7a9b20a-ab53-41ab-9539-d700d8f6d067

Then let us know if you have any problems applying this to your data.

[Milena]

I did it like in the link, but my problem is that I dont have clue what to do with values..

[Adam Danz]

It sounds like you have general quesitons about cross validation rather than how to perform cross validation in MATLAB.

As the example in the documentation explains, this line below applies the leave-one-out partition to the original data, X, and takes the mean of the training observations for each repetition by using crossval. The training observations are 9 of the 10 elements of the vector X and the test data is the single remaining element that was left out. This is repeated until each data point is left out of the training set. Therefore, values has 10 element and each element is the mean of 9 observations in X.

values = crossval(@(Xtrain,Xtest)mean(Xtrain),X,'Partition',c)

[Milena]

Thank you Adam, somehow I didnt understand it (maybe because of language barriers) but I have watched some videos now and it clicked. :)