How to calculate the log likelihood of an LDA model in wowpal wabbit

Question

How to calculate the log likelihood of an LDA model in wowpal wabbit

I am a typical, regular, everyday R user. R is very useful lda.collapsed.gibbs.sampler

in lda

. The tha package uses a Gibbs convolved sampler that fits the Latent Dirichlet Distribution (LDA) model and returns the latent parameter point estimates using the state at the last iteration of the Gibbs sample.

This feature also has a nifty parameter compute.log.likelihood

that, when set, TRUE

will cause the sampler to compute a log word probability (with a constant factor) after each sweep over the variables. This is useful for evaluating convergence and comparing different LDA models (targeting different topics).

I am wondering if there is such an option in the vowpal_wabbit LDA model ?

+3

r lda vowpalwabbit

Marcin Kosiński 22 jul. 15 at 10:53

source to share

No one has answered this question yet

See similar questions:

0

Vowpal Wabbit LDA: model selection

or similar:

nine

The LDA model generates different themes every time I train on the same body

8

Modeling the theme. Assign document with top 2 topics as category label - sklearn Latent Dirichlet distribution

five

Latent Dirichlet Distribution (LDA) in Sparks

five

Topic modeling: how do I use my equipped LDA model to predict new topics for a new dataset in R?

five

What's the best way to get the optimal number of topics for an LDA model using Gensim?

4

Setting hyperparameters of LDA model in wowpal wabbit

2

In the LDA model, how are the multilinear parameters (theta) taken from the previous Dirichlet weight (alpha)?

1

How does LDA produce consistent results?

0

How to reproduce the results of modeling a theme with the LDA package in R

0

Theme stability in LDA models

How to calculate the log likelihood of an LDA model in wowpal wabbit

More articles: