Question 4: Embedded methods combine the best of both worlds, filter and wrapper methods. Embedded methods are: (Check all that apply)
Can anyone help explain why each answer should/should not be selected? I didn’t feel it was clearly covered in the lecture, neither was the grading feedback helpful…
if we summary all methods it should be clear from that …
Filter methods - we are using feature correlation, thus eliminate redundancy, selected subset of “best” features is served to the model.
Wrapper methods - we are using model in loop as a mesure to select best features, starting with one feature or with all features (forward/backward selection) - time consuming process.
Embedded methods - we are using L1, L2 regularization for feature selection and feature importance is another method, both are connected to the model.
So we can see that Embedded methods are definitely faster than Wrapper methods and more efficient than Filter methods.
Hi @tomstryja , I am confused as to what is the difference here between being “faster” and being “more efficient” - Could you explain?