Hi,
For me there is a reversal in causality expressed in the questions of quizz M2.03 :
Instead of
«Fitting a model with a high bias:
a) causes underfitting?
b) causes overfitting? »
It should be reversed :
«Fitting a model with a high bias:
a) is caused by underfitting?
b) is caused by overfitting? »
(same for question 2)
Maybe we should not use the term “cause” in this question. What about:
Fitting a model with a high bias:
a) is a case of underfitting?
b) is a case of overfitting?
or alternatively if we want to keep the “cause”:
Fitting a model with a high bias:
a) causes an underfitted model?
b) causes an overfitted model?
The terms “underfitted” / “overfitted” feel grammatically wrong but they appear in Wikipedia:
I’d rather use “is a case of underfitting” (or “corresponds to underfitting”). “causes” feels weird in either direction …
I think that Olivier’s suggestion is good.
However, to me, one of the problems is that the bias can be either at fit time, or at prediction time. It is legit to talk about biased predictions.
Still, Olivier’s second suggestion seems good to me.
The first solution sounds a little bit better to a novice like me. But the second one is also working.
OK I made the changes here: https://gitlab.inria.fr/learninglab/mooc-scikit-learn/mooc-scikit-learn-coordination/-/commit/cf40799058bd729aa5455acc6ebe49c64298a830
@lfarhi @MarieCollin Could you update FUN.
This is also corrected in FUN now