*answer* q2
 

Can anyone explain to me why the answer to this question is logistic regression?

Logistic regression models trained on different data-sets will all yield value in the range [0,1] given new test points. Averaging [0,1] will still result in [0,1], wouldn't that also make g_avg a valid logistic regression?

The only thing I can think of might be that, the average might end up with total probability integrated in the interval to be greater than 1...

Re: *answer* q2
 

The key point is whether the average can be written as a single logistic regression with the appropriate parameters. If it can, then since contains all logistic regression functions. If sometimes it cannot, then indeed may not be in . Can you think of two logistics regression functions whose average is not one?