Biases in CNN Architectures
Florian and Wieland discuss how CNN architectures exhibit biases towards specific features, impacting robustness. Nicholas highlights the challenges in evaluating input representation papers, questioning their effectiveness in addressing adversarial robustness.In this clip
From this podcast
Machine Learning Street Talk (MLST)
#040 - Adversarial Examples (Dr. Nicholas Carlini, Dr. Wieland Brendel, Florian Tramèr)
Related Questions
What are adversarial attacks on machine learning models as discussed in the episode #040 - Adversarial Examples (Dr. Nicholas Carlini, Dr. Wieland Brendel, Florian Tramèr) and the clip Machine Learning Defense Strategies
How can we defend against adversarial attacks on machine learning models as discussed in the episode #52 - Unadversarial Examples (Hadi Salman, MIT) and the clip Neural Network Features
Are there biases in AI as discussed in the episode Yoshua Bengio: The Past, Present, and Future of Deep Learning and the clip General Inductive Biases?