natanielruiz

3 ms·

1.
by natanielruiz 7y ago
For now, we show that if StarGAN is trained on images augmented by adversarial attack it does become more resistant but not completely resistant to attacks.
2.
by natanielruiz 7y ago
I posted this on a small Facebook group with some of my friends. Those people commenting were some of my best friends.
3.
Disrupting Deepfakes: Adversarial Attacks on Image Translation Networks (Code)(github.com)
46 points by natanielruiz 7y ago|8 comments
4.
by natanielruiz 7y ago
I’m also confused when you say that it has better guarantees. Which guarantees?
5.
by natanielruiz 7y ago
Our "random parameters" baseline is actually the random search that you refer to, while the "random search" that we use, is the actual derivative-free optimization technique of random search. Reviewers asked for a compar
6.
by natanielruiz 7y ago
Maybe we're talking about different algorithms. Do you agree that we are talking about random search as outlined in this Wikipedia page ( https://en.wikipedia.org/wiki/Random_search )?
7.
by natanielruiz 7y ago
No problem! I'm here to help clarify any doubts. I believe in this specific case, there was a local optimum that random search (with the parameters we selected, which we actually tuned as well) was not able to escape. In most cases I t
8.
by natanielruiz 7y ago
https://towardsdatascience.com/learning-to-simulate-c53d8b39...
9.
by natanielruiz 7y ago
Hi there! I'm one of the authors of the work. I can assure you the numbers are not incorrect: random search did not achieve good results in the semantic segmentation experiment. We had similar skepticism with reviewers at ICLR initiall
10.
Learning to Simulate(towardsdatascience.com)
67 points by natanielruiz 7y ago|13 comments