Let's say I have seven vectors (1 x n) e.g.,A, B, C, D, E, F, and G, and I randomly pick one e.g., F.
I have calculated the similarity matrix using cosine similarity and I picked D as the most dissimilar vector to F.
Now I need to pick the most dissimilar vector from the remaining set (A, B, C, E). Does it make sense if I accumulate F and D: New Vector = F + D and calculate cosine similarities between new Vector and A, B, C, and E, and pick the minimum?
I want to order these vector so that they have the most coverage on their features or columns?