New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Gradient analysis #3
Comments
You can use another weight term to filter out the gradients you don't want to collect. For example, suppose the number of samples and the number of classes are 5 and 4 respectively. And the gt_label is [0, 1, 2, 3, 3] Similarly, if you want to collect gradients from only negative samples. So, the detailed step is:
|
It's clear now! But I'm not sure if my understanding is correct. Based on your description, I feed data into a fixed weight model, then collect the gradients of (positive/negative) samples in the last classifier layer by using label as class-wise weight, is it? |
I'm not sure what "using label as class-wise weight" means. |
I will try to reproduce the results, thx. |
Hi, @tztztztztz !
How to collect average L2 norm of gradient of weight? Could you provide the detail step for gen the Fig.1 in paper. (It is better if there are source code!)
Thank a lot!
The text was updated successfully, but these errors were encountered: