search

danielzuegner / nettack

Implementation of the paper "Adversarial Attacks on Neural Networks for Graph Data".
https://www.cs.cit.tum.de/daml/forschung/nettack/
MIT License
218 stars 56 forks source link

Will the default execution of the code exactly reproduce the example image? #9

Closed suyeecav closed 5 years ago

suyeecav commented 5 years ago

If we strictly follow the setup in the code, will it produce exact same output as shown in the example image?

I tried to run your demo code without any modification (but converted it to python file for convenience in running). However, the attack is not very effective and the correct class is shown to be class 1 instead of class 5 shown in the example output (node id is 0 and seed is 15). Since I did not modify anything, I am trying to figure out the possible reason.

danielzuegner commented 5 years ago

Hi. I've just noticed that by default the demo notebook has perturb_structure=False, which is not the setting we used in the paper and for generating the example figure. I've changed that now -- can you try again with the updated settings? You will most likely not get the exact same results since we are not setting the tensorflow seed, but it should be pretty close. Let me know if it works so that I can close this issue then.

Thanks!

suyeecav commented 5 years ago

Hi. I've just noticed that by default the demo notebook has perturb_structure=False, which is not the setting we used in the paper and for generating the example figure. I've changed that now -- can you try again with the updated settings? You will most likely not get the exact same results since we are not setting the tensorflow seed, but it should be pretty close. Let me know if it works so that I can close this issue then.

Thanks!

Thanks for the response. It works perfectly now, at least misclassification can be easily reached.