Instead of the common core being all the atoms which are in every core, could the core be more fluid? As in, every atom is in the core, but weighted by the number of pairs for which it's in the core. Then, we would consider the information about a 100% atom to be more significant than a 90% atom, but we wouldn't consider a 90% atom to be equivalent to a 0% atom (which is essentially what we do now).
This would require altering the overlay algorithm to overlay using all atoms, but weight them when minimizing the overlay score. This would also require a new definition for convergence being reached.
This would probably make computation take much longer.
Instead of the common core being all the atoms which are in every core, could the core be more fluid? As in, every atom is in the core, but weighted by the number of pairs for which it's in the core. Then, we would consider the information about a 100% atom to be more significant than a 90% atom, but we wouldn't consider a 90% atom to be equivalent to a 0% atom (which is essentially what we do now).
This would require altering the overlay algorithm to overlay using all atoms, but weight them when minimizing the overlay score. This would also require a new definition for convergence being reached.
This would probably make computation take much longer.