Yao-Yuan Yang

Probably not "all" are close to zero? Since bibtex dataset has a lot of labels, it is possible that many of the labels are simply not related to the feature,...

Have you train the model and have the loss converged?

Which cost function are you training the model with? If you trained RethingNet with Hamming loss (without the reweighting), it is possible with such result due to the nature of...

For training the RethinkNet, it is in fact training on a weighted binary crossentropy loss. You can check the implementation here. https://github.com/yangarbiter/multilabel-learn/blob/master/mlearn/models/rethinknet/rethinkNet.py If using the binary crossentropy alone without changing...

Sorry about this. Since this is a pretty old project, I don't remember the exact version of the packages. Could you try tensorflow 1.13.1 with keras 2.2.4? Sorry for the...

I am actually not very sure about the SOTA of multilabel classification right now. Maybe a place to start would be looking at papers that cited our paper (https://scholar.google.com/scholar?cites=7577675817034321822&as_sdt=2005&sciodt=0,5&hl=en).