Timeline for Hyperparameter Tuning - What is possible in terms of accuracy gain?
Current License: CC BY-SA 3.0
2 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Mar 24, 2017 at 4:06 | comment | added | KevinKim | I have a question about introduction of feature transformations. I think if you use tree-based models, like xgboost, you don't have to do and variable scaling. Also, any functional transformation on the original set of features is also unnecessary, since tree-based models is able to capture almost any complex non-linear patterns. For example, if you squared a feature, this can be captured by a set of step functions. So I am wondering why for tree-based models we need to do any feature transformation. I think tune the hyper parameters in xgboost is effectively doing feature transformation | |
| Jun 7, 2015 at 16:15 | history | answered | Yannis Assael | CC BY-SA 3.0 |