Skip to content

Conversation

sayakpaul
Copy link
Contributor

No description provided.

Copy link
Collaborator

@arjung arjung left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the PR!

Comment on lines 48 to 49
`third_party` directory contains research project contributions from the open source
community.
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think that's the intent of the third_party directory here. @DualityGap, any thoughts on this? My understanding is that this indicates that contributions here are built on top of other third_party implementations (https://github.com/microsoft/denoised-smoothing in this example). So, I'd suggest removing these 2 lines. As another counterexample, the gnn-survey directory was also a community contribution but that was a from-scratch implementation.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I see. In this case, you are right. I have reused some of the bits from the official implementation (with all the citations) and reimplemented bits that were absolutely necessary to make things work in TensorFlow and Keras. I will reword the phrasing then.



The implementations of Denoised Smoothing [5] are provided in the
`neural_clustering` folder on a strict "as is" basis, without warranties or
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Typo. Not neural_clustering but third_party/denoised_smoothing.


## [Denoised Smoothing: A Provable Defense for Pretrained Classifiers](third_party/denoised_smoothing)


Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Remove extra blank line.


[[5] H. Salman, M.Sun, C. Mitelut, G. Yang, A. Kapoor., Z. Kolter "Denoised Smoothing:
A Provable Defense for Pretrained Classifiers." NeurIPS 2020](https://arxiv.org/abs/2003.01908)

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Remove blank line.

@arjung arjung self-assigned this Sep 15, 2021
@DualityGap DualityGap merged commit 0ed33d6 into tensorflow:master Sep 28, 2021
copybara-service bot pushed a commit that referenced this pull request Sep 28, 2021
PiperOrigin-RevId: 399526290
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants