search

murrayds / sci-mobility-emb

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

Firstdraft #92

Closed jisungyoon closed 4 years ago

jisungyoon commented 4 years ago

Revised mathematical part more precisely.

jisungyoon commented 4 years ago

Yeah, that's a good point. I removed the figure for now.

jisungyoon commented 4 years ago

Also, we can just factorize the matrix directly, and compare the result between node2vec and factorization results.

But, low priority

murrayds commented 4 years ago

Also, we can just factorize the matrix directly, and compare the result between node2vec and factorization results.

But, low priority

How long would that take however? it would be a big matrix, ~10,000 x 10,000, so the memory and computation requirements would be big.

jisungyoon commented 4 years ago

Also, we can just factorize the matrix directly, and compare the result between node2vec and factorization results. But, low priority

How long would that take however? it would be a big matrix, ~10,000 x 10,000, so the memory and computation requirements would be big.

In 5 min with this awesome technique. https://research.fb.com/blog/2014/09/fast-randomized-svd/

murrayds commented 4 years ago

Also, we can just factorize the matrix directly, and compare the result between node2vec and factorization results. But, low priority

How long would that take however? it would be a big matrix, ~10,000 x 10,000, so the memory and computation requirements would be big.

In 5 min with this awesome technique. https://research.fb.com/blog/2014/09/fast-randomized-svd/

interesting! I haven't seen this yet. Lets keep it low priority, but we can definitely try to run this at some point