Papers of the day   All papers

FAST IS BETTER THAN FREE: REVISITING ADVERSARIAL TRAINING

Comments

Ian Goodfellow: When I invented adversarial training as a defense against adversarial examples, I focused on making it as cheap and scalable as possible. Eric and collaborators have now upgraded the original cheap version to compete with newer, more expensive versions.

10 replies, 714 likes


Eric Wong: 1/ New paper on an old topic: turns out, FGSM works as well as PGD for adversarial training!* *Just avoid catastrophic overfitting, as seen in picture Paper: https://arxiv.org/abs/2001.03994 Code: https://github.com/locuslab/fast_adversarial Joint work with @_leslierice and @zicokolter to be at #ICLR2020 https://t.co/2EmwFaX7Qp

3 replies, 252 likes


Eric Wong: Tomorrow, @_leslierice and I will present our work on "Fast is better than free: revisiting adversarial training" during virtual @iclr_conf at 1PM and 4PM EDT! Repo: https://github.com/locuslab/fast_adversarial Paper: https://arxiv.org/abs/2001.03994 ICLR: https://iclr.cc/virtual/poster_BJx040EFvH.html Work with @zicokolter

0 replies, 25 likes


Machine Learning Department at CMU: Fast is better than free: Revisiting adversarial training by Eric Wong, Leslie Rice, and J. Zico Kolter. Paper: http://arxiv.org/abs/2001.0399 Code: https://mld.ai/5b3 #ICLR2020 #mldcmu #cmuai #adversarialtraining #deepneuralnetworks #FGSM #CIFAR10

1 replies, 11 likes


Content

Found on Jan 24 2020 at https://arxiv.org/pdf/2001.03994.pdf

PDF content of a computer science paper: FAST IS BETTER THAN FREE: REVISITING ADVERSARIAL TRAINING