Zobrazeno 1 - 1
of 1
pro vyhledávání: '"Athavale, Anagha"'
Autor:
Athavale, Anagha, Bartocci, Ezio, Christakis, Maria, Maffei, Matteo, Nickovic, Dejan, Weissenbacher, Georg
We present the first automated verification technique for confidence-based 2-safety properties, such as global robustness and global fairness, in deep neural networks (DNNs). Our approach combines self-composition to leverage existing reachability an
Externí odkaz:
http://arxiv.org/abs/2405.14400