# how to create leave one out cross validation in matlab?

itPublisher 分享于

%说明：下面是我自己写的matlab代码，其实matlab有自带的交叉验证代码crossvalind， 见Chunhou Zheng师兄的Metasample Based Sparse Representation for Tumor提

I am still confused with my code. I tried to implement leave one out cross validation in matlab for classification. so in here . I take out one data from training become testing data. I already make a code in matlab. but Iam not sure it's correct because the result is wrong. can someone help me to correct it?? thank you very much.

this is my code :

``````clc
[C,F] = train('D:\fp\',...
'D:\tp\');

for i=size(F,1)
testVal = i;
trainingSet = setdiff(1:numel(C), testVal); % use the rest for training

Ctrain = C(trainingSet,:);
Ftrain = F(trainingSet,:);
test= F(testVal,:);
svmStruct = svmtrain(Ftrain,Ctrain,'showplot',true,'Kernel_Function','rbf');
result_class(i)= svmclassify(svmStruct,test,'showplot',true);
ax(i)=result_class;
i=i+1;
end
``````
matlab cross-validation
|
this question
edited Mar 16 '13 at 16:58 Parag S. Chandakkar 5,583 1 16 46 asked Mar 16 '13 at 15:55 user2157806 139 2 6 15

## marked as duplicate by Shai, Charles Menguy, Stony, nsgulliver, Mia Clarke Mar 17 '13 at 11:12

|

This is what I usually use to create leave one out cross-validation.

``````[Train, Test] = crossvalind('LeaveMOut', N, M)
``````

Here, `N` will be the number of total samples you have in your training+testing set. `M=1` in your case. You can put this in a for loop.

Also, you can use random number generation to perform leave-one out crossvalidation without using predefined function.

|
answered Mar 16 '13 at 16:57 Parag S. Chandakkar 5,583 1 16 46

|

×
• 登录
• 注册

×