The data set was partitioned into 10 subsets, one subsets was used as the testing set and the rest were used for training set. This process was repeated 10 times, with each of the subsets was used exactly once as the testing data. Then, to find the performace of the models, the average performance across all 10 trials was calculated.
Here's a pseudo code on doing this... the key is making sure you have your data divided in to 10 parts;
// k-fold
divide train data into 10 parts
for i = 1 to 10
train network using 9 parts
compute accuracy using 1 part
end for
compute average accuracy of the 10 runs
Here's a working example in MATLAB: see attached file. Let me know if it helps.
Thank you for your help Mr. Puneet Chhabra
i have another problem, how about if i have two parameter for example C and g (Support Vector Machines), how to find the best combining parameter C and g
best regards
That's something I haven't found out yet myself. In the past I used libSVM for SVM and it takes care of the cross-validation stuff but then again which combination is best is something that depends on your application domain and data. I think it's worth asking this as a separate question under SVM tag.. you might get some svm experts to answer that.
regards,
- Badges
- Science topic
- Similar topics
- Social Science
- Media
- Journalism