search

murrayds / sci-mobility-emb

Embedding of scientific mobility across institutions, cities, regions, and countries
4 stars 0 forks source link

Rebuttal #137

Closed murrayds closed 3 years ago

murrayds commented 3 years ago

I began developing the rebuttal letter.

There are still tasks to do, mostly related to the mathematical justification, which @jisungyoon and @skojaku should address.

murrayds commented 3 years ago

I agree with your opinion to reviewer 3 @jisungyoon , but could we re-run the clustering figure with their suggestion? Just running the country-level vectors through UMAP to go from 300->30 dimensions? I really doubt anything would change, but it would better satisfy the reviewer.

jisungyoon commented 3 years ago

Yeah, I can try

jisungyoon commented 3 years ago

I tried result is awful, it highly depends on the hyperparameter choice of n_neighbor, min_dist, and results do not make sense. @murrayds

murrayds commented 3 years ago

Thanks @jisungyoon , I updated the rebuttal letter accordingly. The biggest todo is the math now.

jisungyoon commented 3 years ago

@murrayds @skojaku done with my pass! plz check and let me know what you think.!

skojaku commented 3 years ago

Opp, sorry! My latex formatter inserted/removed a lot of spaces. If you don't like it, I will recover it. I have just added text to "Some concerns on this section??" in line around 271

skojaku commented 3 years ago

@murrayds @jisungyoon I have added a paragraph regarding the neural embedding vs matrix factorization (at the end of the Embedding section in the Methods). I'd like to ask to fill some numbers in xxx. Feel free to fix the text as you want and also I will go over it again once you have done it.

murrayds commented 3 years ago

still needs to finish the Levy factorization text in the Supp, but looking good!

murrayds commented 3 years ago

This branch is way too big, so I'm closing it out. New rebuttal changes can go in their own.