Latent Dirichlet Allocation - Iterations?

Soni

Hi RapidMiner team,

I'm rather new to LDA and am a bit confused to what exactly the iterations parameter refers to since the description says "Number of iterations for optimization". However, as far as I understood it, the operator uses Gibbs sampling where the number of iterations should be set as well. Is it set using the iterations parameter or has the parameter something to do with optimizing the hyperparameters (whereby the input field does not disappear if I uncheck the hyperparameter optimization)?

Thanks in advance!

MartinLiebig

Hi @Soni,

you are absultely right. its the number of sampling steps in Gibbs Sampling and is effectivly a hyper parameter of LDA.

Best,

Martin

MarlaBot

Hi @Soni - this is MarlaBot. I found these great videos on our RapidMiner Academy that you may find helpful:

Instructional Video: Applying a Model to categorize Documents (Viewing time: ~11m)

Instructional Video: Text Association Rules (Viewing time: ~10m)

Please LIKE my comment if it helps! 👇

MarlaBot

MartinLiebig

Hi @Soni,

you are absultely right. its the number of sampling steps in Gibbs Sampling and is effectivly a hyper parameter of LDA.

Best,

Martin

RickB

Hi @mschmitz, I am currently trying to analyze short text documents for a research project, and I was wondering if there is a way to do Gibbs Sampling (GSDMM) with RapidMiner? 

MartinLiebig

Hi @RickB,

only inside LDA. Not as a standalone thing, sorry.

Best,

Martin

RickB

Hi @mschmitz, 

Thank you for the quick response. I tried analyzing the data using LDA, but I did not see an option for Gibbs Sampling. Would you be able to direct me to where I can find that inside the LDA operator? 

Best,

Rich

MartinLiebig

Sorry? I don't understand the question. The LDA we use internally is from Mallet. The operator executes similar code to whats shown here:

http://mallet.cs.umass.edu/topics-devel.php

What would you expect?

Best,

Martin