Exploring Adversarial Machine Learning - 2_evasion_assessment

I am trying to solve the first assessment. Although I am able to generate an adversarial image, the autograder says ‘bad label’. I am not sure what is expected to get a full score.

Some information or nudges would be helpful. Thanks!

Also, I did not know that the assessment would be part of the 30 hour limit. Is it possible to get an extension before I run out?

Hi @shreyassriram99,

I added eight hours to your course.

Good luck,
Tom

Thanks Tom, appreciate that!

Can I get some information on the ‘bad label’ error from autograder for evasion assessment? I’ve tried high prob, low L2 loss and other ways to increase evasion. All attempts seem to result in ‘bad label’. Any information or nudges around this will help me improve the attack. Thanks!

Sorry, I am not familiar with that error. Maybe @jwyatt can chime in here.

Hey team,

I want to follow up with the same question, according to the feedback provided by the notebook the image was recognized well as a laptop, but the evaluation shows “bad image”, could you please provide more explanation about the error? I have made more perturbations but without success.

Hello! I’m experiencing a similar issue with my Adversarial Machine Learning Certification, could I get some extra time for my exercises?

Hi @es5888

I addded eight hours to your lab.

Best of luck,
Tom

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.