machine learning - Which standard deviation of the cross-validation score? -
when doing cross-validation model selection, found there many ways quote "standard deviation" of cross-validation scores (here "score" means evaluation metric e.g. accuracy, auc, loss, etc.)
1) 1 way calculate standard deviation on mean of scores of k folds (= standard deviation of k folds / sqrt(k)).
2) second way calculate standard deviation of scores of k folds. example can found here:
http://scikit-learn.org/stable/auto_examples/svm/plot_svm_anova.html
3) way don't understand. seems calculate standard deviation of k folds / sqrt(n) n size of dataset...
http://scikit-learn.org/stable/auto_examples/exercises/plot_cv_diabetes.html
personally think 1) correct, care more standard error on sample mean (here = average score of k folds validation) rather standard deviation of sample. can explain way preferred?
Comments
Post a Comment