search

VietHoang1512 / GLOT

Global-Local Regularization Via Distributional Robustness (AISTATS 2023)
9 stars 1 forks source link

Code for multi-source domain adaptation and multi-source domain generalization #1

Closed maoshiyun closed 1 year ago

maoshiyun commented 1 year ago

Hello, can you provide the code for multi-source domain adaptation and multi-source domain generalization?

VietHoang1512 commented 1 year ago

Hi @maoshiyun, thank you for your interest in our work. Unfortunately I deleted my code on the server and was not able to get back the training file along with the bash script (models and data loader etc are already available in our public repository). However, I found that the training file is hosted in one of my private repositories. Will update them to this repo later this week.

VietHoang1512 commented 1 year ago

Hi @maoshiyun, I have uploaded the main python file. Please let me know if you have any trouble running our scripts.

maoshiyun commented 1 year ago

Thank you for your sharing. There is a line of code “from model_pacs import ModelAggregate” in the file “main_pacs.py”, But the function “ModelAggregate” is not found in the file “main_pacs.py”. Can you share the code of " ModelAggregate"? Thank you very much!

VietHoang1512 commented 1 year ago

I have updated the code in the newest commit, please reopen the issue if you encounter any problem

maoshiyun commented 1 year ago

Thank you for your answer! I have another question, how is the global regularization term used in domain generalization reflected in the code, i.e., can you provide the code for Eq. (13) in the paper? image

VietHoang1512 commented 1 year ago

Thank you for your question, each of the term W_d is implemented similar to the DA code and the SSL code.