I am reading Mining Significant Graph Patterns by Leap Search (Yan et al., 2008), and I am unclear on how their technique translates to the unlabeled setting, since $p$ and $q$ (the frequency functions for positive and negative examples, respectively) are omnipresent.

On page 436 however, the authors clearly state that "In the following presentation, we are going to use the second setting (Figure 3) to illustrate the main idea. Nevertheless, the proposed technique can also be applied to the 1st [unlabeled] setting."

  • $\begingroup$ If I understand it correctly from skimming the paper, $p$ and $q$ are frequencies of graphs in a sample, whereas the labels you mention are classifying information independent from that. I have not seen actual uses of the class labels. $\endgroup$
    – Raphael
    Mar 7, 2012 at 10:18
  • $\begingroup$ @Raphael: the labels are positive and negative, and $p$ and $q$ are the frequencies in the positive and negative graphs, respectively, so unless I am missing something they are not independent at all. $\endgroup$
    – mitchus
    Mar 7, 2012 at 15:55
  • $\begingroup$ I meant independent of the class labels. $\endgroup$
    – Raphael
    Mar 7, 2012 at 16:13


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.