Skip to content
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

Use of k=hwc for selecting beta #40

Open
KellyWuuu opened this issue Nov 16, 2021 · 2 comments
Open

Use of k=hwc for selecting beta #40

KellyWuuu opened this issue Nov 16, 2021 · 2 comments

Comments

@KellyWuuu
Copy link

Hi! The paper is really great, and thank you for sharing the code here.

I have a quick question about how you calculate k. In the paper, it is said that k = hwc, where h is height, w is width and c is channels. However, when I referred to Appendix C for VGG-16, it seems that k needs to equal 1. Would you mind explaining the calculation of k and its use?

Thanks a lot!

@berleon
Copy link
Member

berleon commented Nov 17, 2021

Hi! Thank you for your interest in our paper. Also in the Appendix, k = hwc. In the Appendix C, there is only the Figure 10. How do you got the impression that k=1?

@KellyWuuu
Copy link
Author

Thank you for your reply!
Maybe I'm making a wrong connection. So, I'm looking at the third column of Figure 10, and the top and bottom heatmaps looked the same. Since the bottleneck is inserted at the same layer, conv4_1, I'm getting a sense that beta = 10 = 10/k.

Using k = hwc makes k an extremely large number; will this make beta too small?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants