search

vitutorial / VITutorial

This repository stores slides for a tutorial on variational inference for NLP audiences.
298 stars 55 forks source link

Illustration for mean field VI vs amortised mean field VI #25

Closed wilkeraziz closed 6 years ago

wilkeraziz commented 6 years ago

New illustration to better contrast mean field VI and amortised mean field VI.

In slide 40 ("amortised variational inference") we condition on x (which we defined earlier as a sequence x_1, ..., x_N, but I wonder whether it would help to explicitly write it as a sequence here. I used x to avoid clutter though.

In any case at this point I get questions like "what if N varies?" or "can K vary?", and it's way better to answer these questions than to prepare slides for it.

wilkeraziz commented 6 years ago

Alright, binary latent factors and example after LDA.

philschulz commented 6 years ago

Cool, thanks for the changes. I just had a look at the pdf. For some reason, slide 37 reappears after slide 40 (at least when I view the pdf in github). Can you check this and recompile if needed. Also, could you change the heading from "Example Model" to "Latent Factor Document Model" or something like this? Basically anything that is more specific. I'll happily merge it in then ;)

philschulz commented 6 years ago

Thanks!