The problem setup is as follows: There a $N$ items which have to be clustered into $K$ clusters. Each Items has a certain number of dependencies, where each item takes dependencies from a discrete set $\mathbb{A}$. Each item can take 1 to $dim(\mathbb{A})$ dependencies from this set.
Items have to be clustered such that the total number of unique dependencies of each cluster is kept to a minimum.
I tried creating a bit array of size $dim(\mathbb{A})$ for each item denoting the presence of a dependency by $1$ and absence by $0$. And then the hamming distance between these could be a similarity metric, but am confused what would be the best way to cluster such a set of bit arrays?
