search

Roche / BalancedLossNLP

Other
119 stars 23 forks source link

where are partical_cross_entropy and cross_entropy in util_loss.py #5

Open CNwangbin opened 1 year ago

CNwangbin commented 1 year ago

I can't fina them, please help help me. https://github.com/Roche/BalancedLossNLP/blob/main/PubMed/util_loss.py#:~:text=partial_cross_entropy

https://github.com/Roche/BalancedLossNLP/blob/main/PubMed/util_loss.py#:~:text=self.cls_criterion%20%3D%20cross_entropy

blessu commented 1 year ago

Thanks for your question.

As indicated in the paper, we implemented similar solutions as in ECCV'20 paper, and focused on those only based on binary_cross_entropy. Thus, the implementation of other approaches (like partical_cross_entropy or cross_entropy) was removed. The variables were retained for scalability concerns, sorry for misleading you.

Having that said, if you would explore further on partical_cross_entropy or cross_entropy approaches, I would recommend you to read the ECCV'20 paper, and check their implementation in https://github.com/wutong16/DistributionBalancedLoss/blob/master/mllt/models/losses/cross_entropy_loss.py

CNwangbin commented 1 year ago

Thanks for your reply, I got it a few hours later.     ------------------ Original ------------------ From: @.>; Date:  Wed, Jul 20, 2022 10:05 AM To: @.>; Cc: @.>; @.>; Subject:  Re: [Roche/BalancedLossNLP] where are partical_cross_entropy and cross_entropy in util_loss.py (Issue #5)

 

Thanks for your question.

As indicated in the paper, we implemented similar solutions as in ECCV'20 paper, and focused on those only based on binary_cross_entropy. Thus, the implementation of other approaches (like partical_cross_entropy or cross_entropy) was removed. The variables were retained for scalability concerns, sorry for misleading you.

Having that said, if you would explore further on partical_cross_entropy or cross_entropy approaches, I would recommend you to read the ECCV'20 paper, and check their implementation in https://github.com/wutong16/DistributionBalancedLoss/blob/master/mllt/models/losses/cross_entropy_loss.py

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>