I am currently revising all the material from this course and I am wondering how the choice of kernel affects the complexity of a SVM model? Do we simply treat the choice of kernel as a hyper parameter and try different ones while grid searching over the regularisation variable lambda?
Great question! I also have a related question, should the grid-search be done separately for each variable? Testing all the possible kernel-variable combinations.
Impact of choice of kernel on model complexity
Good afternoon,
I am currently revising all the material from this course and I am wondering how the choice of kernel affects the complexity of a SVM model? Do we simply treat the choice of kernel as a hyper parameter and try different ones while grid searching over the regularisation variable lambda?
1
Great question! I also have a related question, should the grid-search be done separately for each variable? Testing all the possible kernel-variable combinations.
Any advice on how to choose the best kernel?
Add comment