MoroccoAI Webinar - Yassine Abbahaddou & Sofiane Ennadir - Adversarial Robustness of GNNs

MoroccoAI is a non-profit association led by AI scientists and researchers to promote AI education and excellence in Research in Morocco and abroad.
Subscribe and support us by sharing with friends and colleagues.
In this webinar, Yassine and Sofiane will delve into a theoretical approach to understanding Graph Neural Networks’ adversarial robustness, aiming to derive defense methodologies with provable robustness.
They will introduce an upper-bound on the expected adversarial robustness of Graph Convolutional Networks (GCNs) when subjected to both structural and node-feature based adversarial attacks, along with a probabilistic method to estimate the expected robustness. This method enables us to assess the effectiveness of GCORN on various datasets.
Their recent paper “Bounding the Expected Robustness of Graph Neural Networks Subject to Node Feature Attacks” was accepted in the Twelfth International Conference on Learning Representations (ICLR 2024).
🔗 Link to the full paper: openreview.net...
An abstract of the talk can be found on www.morocco.ai home page.
Follow us on LinkedIn at / morocco-ai