Faster implementation of LDA

Vote count:

0

Fellows,

My dataset comprises of around 3 Million documents containing 16k words, and my document term matrix is mostly sparse, the frequency is represented in binary form, 1 for present, 0 for absent.

I ran LDA in R using topic models package and lda package with both the inference methods: Gibbs Sampling and VEM for indocument document terms matrix involving 300 million documents, and 300 features only. It took 10 hours to return me the results.

I wanted to get suggestions for faster implementation of lda.

Thanks.

asked 36 secs ago

Basmah

210