Discussion of Lecture 6

Xijia Liu

Department of Statistics, Umeå University

Lecture 6: Model Validation and Selection

What is KNN? What is special about KNN? Can you apply KNN for regression problem?
What is difference between model parameters and hyper-parameters?
How do you deteremine the values of hyper-parameters?
LOOCV V.S. KFCV V.S. Bootstrap
Why applying stepwise selection methods has the risk of local optimal solution trap?
What is the difference between ridge regression and LASSO?
What is the hyper-parameter in regularization methods? What is the its role?