Why does simhash work? I understand how to implement the hash algorithm, mechanically, from the many articles such as http://matpalm.com/resemblance/simhash/. But is there a simple intuitive explanation for why this particular procedure is so effective at capturing similarity?

  • 2
    $\begingroup$ Could you edit the question to clarify what problem you are asking about? The assumption in the question is that simhash (whatever it is) works, but works for what? The question talks about “capturing similarity,” but similarity of what? $\endgroup$ – Tsuyoshi Ito Oct 12 '14 at 1:51
  • 1
    $\begingroup$ Isn't the whole premise of simhash that (from the link) "similar items are hashed to similar hash values (by similar we mean the bitwise hamming distance between hash values)"? I believe "items" is generally documents, strings, etc. $\endgroup$ – xyz Oct 13 '14 at 8:18
  • $\begingroup$ Well, I do not know what simhash is. I am just trying to help you state your question in such a way that theoretical computer science can be useful. In the current form, it is unclear to me why you are asking this question on a website about theoretical computer science. $\endgroup$ – Tsuyoshi Ito Oct 13 '14 at 9:38

Browse other questions tagged or ask your own question.