Are you intending to contradict the contents of the article quoted by the question? Please be clear on this since the question is logically following the procedures of the article. So you're not actually arguing about a misapplication of the article's steps by the OP but rather that the premise of the procedure is incorrect. Given you admit that you are unfamiliar with IForest then well you should likely do that homework first as part of preparing an answer?

nice question and good investigation you did already. not sure why others did not upvote

This is another way of saying the set of variables whose distributions we are currently not interested in, but whose values we treat as constants Is it not the case that we are interested in the input variables' distributions but in the sense of how they affect the distribution of the output variable

yea, it is strange to see the basic "regression analysis" definition employ the term "independent variables" in this way since as the OP states they may be highly dependent [or even collinear]

@FrankHarrell I found this about how logistic regression works well with extremely unbalanced datasets: stats.stackexchange.com/a/6086/64552 . It says the intercept and threshold need to be adjusted. Do you have any further insights?

This is actually a more readily actionable list for day to day training than the accepted answer - which tends towards steps that would be needed when doing more serious attention to a more complicated network.

yes this is a great answer. If you're still "around" (>7 yrs later): I am going to ask a separate question about "In MLE, 𝜃̂ is assumed to be a fixed quantity that is unknown but able to be inferred, not a random variable"