0

I'm trying to find out how to carry out the multiplication of the matrices produced after SVD implementation in LSI. I need this for my research. I want to carry out document clustering.

Rochetta
  • 1
  • 1

1 Answers1

0

From what I've read, once you decompose the term-document matrix with the SVD to create [U, S, V], you need to multiple the singular values, S, with the left factor matrix, V'. V' is your document-space matrix, which is scaled by the singular values. The cosine similarity of the rows of the resulting matrix should give you document similarities.

David Jurgens
  • 304
  • 1
  • 8