Tweeted By @DeepMindAI
"Are labels required for improving adversarial robustness?" TL;DR: no!
— DeepMind (@DeepMindAI) September 16, 2019
With only 10% of CIFAR10 labels, UAT has almost no drop in robust accuracy. With additional unlabeled data, UAT obtains SOTA robust accuracy.
Paper: https://t.co/3HF362CcEK
Code: https://t.co/kApBTF1nJj pic.twitter.com/DToqzeMm8l