From my understanding, a lower covariance indicates a better and less noisy dataset, but the code here seems to require more noisy data in order to run? I'm very confused as to why the bound is set like this. Any explanation would be appreciated!