what is the go-to tool for LDA these days, which is easy to run and gets good results? back at the day it was mallet with the awesome hyperparametr optimization by @hannawallach and @dmimno , is that still the case? or do we have friendlier/jvm-less options now?
-
Does it work well also for smaller document collections? (I am not familiar with the spectral method, does it solve the actual objective, the variational one, or other?)
Yes, we tried for both small document collections (UCI datasets) and larger ones. This is the paper (short document example at end of appendix) https://arxiv.org/pdf/1204.6703.pdf … and my talkhttps://www.youtube.com/watch?v=mINfveR9VVE …
-
-
Spectral methods for
#LDA does not solve the variational objective, instead it solves the method of moments. In terms of topic quality, we tried coherence measures which are better aligned with human interpretability rather than variational approximation of perplexity. -
This sounds great, thanks! I will look into it.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.