0
$\begingroup$

I refer to the Wikipedia and other tutorials on topic modeling, which said

although PLSA is a generative model of the documents in the collection it is estimated on, it is not a generative model of new documents.

while another tutorial page 15 that illustrates pLSA and LDA in geometric way said

The pLSI model allows a document to possess a distribution over topics that was seen in the training data, thus placing new documents at particular points within the topic simplex.

Could I and should I use a trained pLSA model to generate topic distributions on unseen (new) documents?

$\endgroup$
1
$\begingroup$

pLSA learns the topics from the training set(fitting the model). pLSA doesn't predict new topics from unseen data. it just uses the topics learned during training and returns the document-topic distribution for your unseen test set(transform).

$\endgroup$
1
  • $\begingroup$ The Wiki says that LDA has overcome the shortcoming of pLSA which is not a generative model for new document. So does it mean that LDA can learn a new generative model on a new document? $\endgroup$ – Zelong Oct 25 '16 at 8:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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