This repository has been archived by the owner on Oct 31, 2023. It is now read-only.
Fix bug in how noise scale is computed #2
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Per Equation 15 in [the paper](https://arxiv.org/pdf/2102.11673.pdf), the relation between the noise parameter sigma (`scale`) and the spectral norm of the Jacobian (`eta_max`) should involve division by `eta`, not multiplication. Simply put: to guarantee a lower FIL, you need to add noise with a larger scale. This corrects that mistake in the code.
facebook-github-bot
added
the
CLA Signed
cg563
approved these changes
Aug 9, 2021
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Per Equation 15 in the paper, the relation between the noise parameter sigma (
scale) and the spectral norm of the Jacobian (eta_max) should involve division byeta, not multiplication. Simply put: to guarantee a lower FIL, you need to add noise with a larger scale.This corrects that mistake in the code.