Come up with a better cleanup strategy

I really don't like this solution still, need to examine the need for the clean_up() function. There is a strong possibility for stale similarity data to reside in the table. So the easiest option is to insert into the similarity table all values regardless of how dissimilar and then call the clean_up() function to delete the dissimilar items.

If you don't do that, you have the possibility of having a once very similar relationship but over time became dissimilar. These are never cleaned up. Wonder if there is a way to use a timestamp to expire the relationships over time. Most similarity handlers will use an INSERT () ON DUPLICATE KEY UPDATE clauses, so it could update the timestamp stuff. Thus allowing a relationship to expire.