We leverage a lightweight regularizer to reduce forgetting in continual learning.
We leverage source-based auxiliary tasks to accelerate and enhance adaptation.
We study the generalizability of adversarial robustness to unseen domains.
We extend randomized smoothing to anisotropic distributions and optimize the anisotropic smoothing parameters per input point in randomized smoothing.
We characterizes the decision boundaries of neural networks using tropical geometry, and leverage this characterization into several applications.