PhDOpen: Aleksander Mądry - Shared screen with speaker view
but the accuracy on this robustified dataset will be lower than accuracy on non-robust one, right?
Does this mean that when the robust models are wrong are they usually more "sensibly" wrong?
great lectures! do you know of any work being done in the context of "robust" reinforcement learning, especially exploring phenomena qualitatively different from the supervised setting?
great lectures, I learned a lot, thank you!