Leave one out Help

Hi,

I have a structure of size S=18x112x50. Here 18=different patients, [112 is (1:37)=class 1 and 38:112=class 0] and 50 = no. of iterations. Now i need to perform Leave one out method to find the accuracy of the feature. For eg: Train1=s1(1:17,:,:),Test1=(18,:,:), run classification algorithm (anything like SVM, LDA etc)get the accuracy, and in second iteration Train2=s1(2:18,:,:),Test2=(1,:,:) and so on as leave one out.

Answer 1

[Shashank Prasanna]

Sunil, if you have the Statistics Toolbox you can use 'crossval' to perform leave one out cross validation:

http://www.mathworks.com/help/stats/crossval.html

You can set 'leaveout' to be 1. The above link has examples on how to use the function.

[Sunil]

Hi, I am new to this classification problem. Any demo code for LOO on an example data (of size mxnxp) would be helpful.

[Shashank Prasanna]

I am afraid I don't understand what you mean by LOO, but as I already mentioned there are examples on how to cross validate classification right at the bottom of the page. You will have to scroll all the way down.

[sweet dm]

hi please tell me when we have small data of 60 instance can we do the loo as classification without using test part ??

[Bernhard Suhm]

You can, but neither the model nor its accuracy estimate will be very reliable with this small a dataset.