[Paper Reading] Adversarial training descends without descent: Finding actual descent directions based on Danskin’s Theorem

Please visit the link to open the slide: View PDF

Main References

• Fabian Latorre, Igor Krawczuk, Leello Tadesse Dadi, Thomas Pethick, and Volkan Cevher. Finding Actual Descent Directions for Adversarial Training. ICLR 2023.
• Aleksander Madry, Aleksandar Makelov, Ludwig Schmidt, Dimitris Tsipras, and Adrian Vladu Towards Deep Learning Models Resistant to Adversarial Attacks. ICLR 2018.