Hi,
I overlooked the term “by default” in the question.
Now, looking at the given answer and the documentation, could you clarify here?
I understand that the bootstrap resampling is performed in any case for both the samples and the features.
The default setting just say that we resample the whole set of samples and the whole set of features, with replacement and without replacement respectively.
bootstrap_features=False by default so there is no bootstrap on the features here.
So bagging will use bootstrapping for the sample (resampling with replacement). For the feature, it will use all the original feature (even if there is shuffling involved).
Since we are interested in checking that one understood that there is no bootstrapping on the feature, we should reformulate the answer to not have any ambiguity regarding the additional shuffling.
That’s right! However, from the documentation, we understand that:
Regarding samples, we will, by default, draw all samples with replacement.
Regarding features, we will by default, draw all features without replacement. It would just mean using all features.
I can still randomly draw samples without replacement over features until I draw all features, can’t I? If so, one of the other options is still valid, anyway really ambiguous.
If you guys are just interested in the term bootstrapping, the question should definitely be reformulated, explicitly asking about bootstrapping, not over-complicating things. Alternatively, the question could just be about what is bootstrapping.
Yes and in this case, you obtain the original set of features that is just shuffled.
Therefore, the last option in the MCQ is valid.
Hi,
Even after reading this discussion I do not understand the logic of the expected answer
During this courses, I got confused countless time because of the multiple choice questions. Could you at least give partial point?
