I would like to clarify the meaning of

in question 2 of the final exam, if possible...

Suppose the polynomial model

which is applied to all

*X* in a data (sub)set

*D*. There are

data subsets, indexed by

*d*. Under this model,

is considered to be:

or

where

?

In other words, we average the "functional form" of the function, or we average the parameters of the function over the training subsets?

Thanks in advance.