-
Notifications
You must be signed in to change notification settings - Fork 0
overfitting
Tansu Dasli edited this page Sep 15, 2023
·
5 revisions
generally;
- more data,
- cross-validation,
- reducing variables (future selection & extraction),
- early stop for dl, neural networks
overfitting scaling multi-collinearity
regression | L1 (lasso), L2 (rigde) Regularization | standardize | eliminate high correlated features or
| use Ridge(Lasso)Regression which penalties high correlation
trees | ensembling (bagging, boosting, RF) | no-need | n/a
neural network, DL | early stop, drop-out | normalize | n/a