I was wondering, what if we know the subspace generated F from the data instances, but we cannot constitute the origin space E that can be in higher dimension, and can easily lead us to the true join density distribution.
In other words, the basis vector of the F space is already constructed but there is no way that we can add certain features to that basis towards constituting the basis of E.
Some of you can give as an answer that i had to select the right features, i partly agree simply because defining the supposed right metrics is never deterministic and differ from one person to another or from certain circumstances to other circumstances.
I am hoping to get from you a sort of relief kind of answer.
Thank you so much.