I will describe my problem in details. I have a set of $N$ vectors, each of them defines a logistic regresion model. From this $N$ vectors (models) I want to take $n$ which are the most unique.
I'm not interested in similar vectors (models) because they share same information, and they will produce similar predictions. All of my $N$ models have high predictive power and I want to create single clasifier from $n$ most independent in a given set.
It is useles to make single calsifier $c=(A+B)/2$ from model A (predictive power=75%), and model $B$ (predictive power=75%) when they are equal ($A=B$).
But if they differ in predictor variables (defined in my vectors) and still have predictive power at high level (75%) in case when $A\neq B$ then combined clasifier $c=(A+B)/2$ will perform much better.
Is there any algoritmic (R), or mathematical framework for finding most independent (diversified?) subset of vectors in a given set?