|
Hi all, I am beginning to learn how to build a topic model using Latent Dirichlet Allocation, however I am still having lots of problem trying to understand the formula and the process. Does anyone here know any websites/tutorial with working example that illustrate how to calculate so that I can verify my understanding is correct? Thanks. Regards, Andy. |
|
Gregor Henrich's note on Parameter Estimation for Text Analysis, although a bit more general than just LDA, also works out the inference (Gibbs sampling) details for LDA, in reasonable detail (especially section 5). If you want to understand variational inference for LDA, you can take a look at the Topic Models survey paper by Blei & Lafferty. |
|
I really like the derivation by Bob Carpenter of the LDA Gibbs sampler. It also shows you how the bayesian naive bayes gibbs sampler works, so you can use those equations to build more complex models easily. |
|
I found the Original paper by Blei, Jordan and NG to be really good as a starting point, also, try reading Yee Why Teeh paper on hierarchical Topic Models as a followup |
