3
$\begingroup$

Some texts seem to list matrix factorization as a method for collaborative filtering, and more specifically categorize them as a "model-based approach" (e.g. here and here), while others seem to treat them differently (e.g. see here where the presenter discusses three distinct solutions, content-based, collaborative, and latent-factor-based. Similar implied distinctions are madehere or e.g. here)

This is perhaps a matter of semantics and etymology, but what are the merits for each argument? That is, from what perspective is MF a CF method ("model-based or not) or vice-versa?

$\endgroup$
2
  • $\begingroup$ What “arguments” you have in mind? $\endgroup$
    – Tim
    May 29, 2020 at 18:44
  • $\begingroup$ Thanks @Tim I don't actually know if there are mathematical arguments or interpretations of MF that would group MF as a model-based CF method. If there are none (perhaps because the term CF never had a sufficiently rigorous definition), the explanation may be historical context. So by "argument" here I mean what arguments would an author use to group MF under CF or not. $\endgroup$
    – Josh
    May 29, 2020 at 18:55

1 Answer 1

3
$\begingroup$

Matrix factorization is a kind of collaborative filtering,

collaborative filtering is a method of making automatic predictions (filtering) about the interests of a user by collecting preferences or taste information from many users (collaborating).

Matrix factorization is based on a model (it learns representation of the data vs just finding similar users using some kind of distance metric) that factorizes the matrix of user-item preferences $R$ into two latent (unobstructed) variable matrices $P$ and $Q$:

$$ R \approx PQ $$

So all the descriptions are correct, they are just of different granularity.

$\endgroup$
2
  • $\begingroup$ +1 - Thanks, although I'm not sure I would go as far as to say that all the descriptions are correct, since some of the sources above seem to list user-item matrix factorization approaches as a standalone entity outside of CF. It could be an oversight of the authors, or an unintended interpretation of their text / presentation. $\endgroup$
    – Josh
    May 30, 2020 at 0:31
  • $\begingroup$ @Josh I didn’t watch the YouTube video, but the rest of the sources seem to be inline with what I wrote. $\endgroup$
    – Tim
    May 30, 2020 at 7:06

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.