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

mnist_perturbation notebook won't run #184

Open
tmerrittsmith opened this issue Feb 21, 2020 · 4 comments
Open

mnist_perturbation notebook won't run #184

tmerrittsmith opened this issue Feb 21, 2020 · 4 comments
Assignees

Comments

@tmerrittsmith
Copy link

A few issues i've found with running this:

  1. cell 5, methods[0] is called "input", but the outputs of the uploaded notebook show that it should be called 'random'. Is this correct?

  2. cell 9 returns an error on line 37: project() got an unexpected keyword argument 'input_is_postive_only' - this appears to be the same as project() got an unexpected keyword argument 'input_is_postive_only' #176

Happy to submit a pull request to fix these if that's helpful

@enryH
Copy link
Collaborator

enryH commented Feb 21, 2020

If you have done it already, just make an PR:)

@tmerrittsmith
Copy link
Author

ok will do - have run in to a few more issues so going to get to the bottom of those first

@tmerrittsmith
Copy link
Author

Hey @enryH, I don't think a PR is the right way to resolve this.

My original question was whether the 'input' perturbation method is the same as 'random'. I'm pretty sure it's not. In the git log for this notebook, @pseegerer added the 'random' method, but this was then removed in following commits. (added in e46f268 and removed by @albermax on a084b17

Can you take a look to confirm which commit is correct?

Reverting this line seems to get the notebook to run without problem now. However, the actual images of the perturbations don't look like they're working properly - my understanding is that this method should additively mask relevant regions of the image, but in my version of the notebook this masking seems to be happening at random (see attached snip) - not sure if this is related to the above overwrites or not?

Capture
)

@tmerrittsmith
Copy link
Author

I've run the notebook reverted to that previous commit (e46f268) and the perturbation looks right, so it must have been introduced since then.

@enryH enryH added the bug label Feb 25, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

3 participants