Timeline for Does the standard deviation of the folds of LOO cross-validation have any practical meaning in comparison/evaluation of classifiers?
Current License: CC BY-SA 3.0
4 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 16, 2014 at 15:47 | comment | added | cbeleites | @davips: decrease size according to the size/minority class may be described as leave-$n$-out strategy, which is an intuitive minor variation of $k$-fold that is perfectly sensible in this context. If you need further literature, don't hesitate to ask. And thanks for the flowers. | |
| May 15, 2014 at 19:22 | comment | added | dawid | BTW, your answers have been useful all over this site. | |
| May 15, 2014 at 19:09 | comment | added | dawid | Yes, I have lots of datasets to compare classifiers (actually, the goal is to compare active learning strategies), from tiny to huge ones. I am trying to define a reasonable methodology that uses adequate CVs for each dataset. As I have no references about this, I started thinking about LOO for small datasets and 10-fold for big datasets. A second option would be to decrease k according to the size/minority class of the datasets. Using 10-fold for all datasets would be like driving a truck on a terrain with farms and gardens. :) | |
| May 15, 2014 at 6:18 | history | answered | cbeleites | CC BY-SA 3.0 |