search

cemoody / lda2vec

MIT License
3.15k stars 628 forks source link

Dimension of Topic Vector #68

Open aggarwalpiush opened 6 years ago

aggarwalpiush commented 6 years ago

According to LDA algorithm, dimension of topic vector (word probabilities) should be K X V. K is number of topics and V is vocabulary size (http://www.jmlr.org/papers/volume3/blei03a/blei03a.pdf). But here dimension is defined as K X W. W is word vector dimension or constant somewhere.

#Number of dimensions in a single word vector n_units = int(os.getenv('n_units', 300))

Kindly explain, how these dimension can also follow LDA algorithm.

Thanks in advance!!

fan8502 commented 4 years ago

@aggarwalpiush Hi, I'm wondering the problem, too. Did u get any solution? Thanks!