search

cora1021 / TreeLoss

0 stars 0 forks source link

Extreme Classification #3

Open mikeizbicki opened 2 years ago

mikeizbicki commented 2 years ago

At some point (probably after AISTATS submission), we should talk about extreme classification. I'm putting a bunch of links here so that I don't forget about them. There's some good background material here that you might want to check out if you have time, but don't worry about it if you don't.

https://icml.cc/Conferences/2020/ScheduleMultitrack?event=5719

http://manikvarma.org/downloads/XC/XMLRepository.html

https://www.microsoft.com/en-us/research/blog/everything-you-always-wanted-to-know-about-extreme-classification-but-were-afraid-to-ask/

mikeizbicki commented 2 years ago

Something else we should talk about (again after we're done with this paper) is the negative sampling loss used in word2vec. I think the same basic idea can be used to speed up our tree loss, and that there might be a cool application to fasttext where we can improve performance on low resource languages.