Timezone: »

 
Poster
Smoothed Embeddings for Certified Few-Shot Learning
Mikhail Pautov · Olesya Kuznetsova · Nurislam Tursynbek · Aleksandr Petiushko · Ivan Oseledets

Thu Dec 01 02:00 PM -- 04:00 PM (PST) @ Hall J #417
in Poster Session 6 »
Randomized smoothing is considered to be the state-of-the-art provable defense against adversarial perturbations. However, it heavily exploits the fact that classifiers map input objects to class probabilities and do not focus on the ones that learn a metric space in which classification is performed by computing distances to embeddings of class prototypes. In this work, we extend randomized smoothing to few-shot learning models that map inputs to normalized embeddings. We provide analysis of the Lipschitz continuity of such models and derive a robustness certificate against $\ell_2$-bounded perturbations that may be useful in few-shot learning scenarios. Our theoretical results are confirmed by experiments on different datasets.
Keywords: Few-Shot Learning · randomized smoothing · certified robustness
[ Paper]  [ Poster]  [ OpenReview] 

Author Information

Mikhail Pautov (Skolkovo Institute of Science and Technology)
Olesya Kuznetsova
Nurislam Tursynbek (Skoltech)
Aleksandr Petiushko (Lomonosov Moscow State University)
Ivan Oseledets (Skolkovo Institute of Science and Technology)

More from the Same Authors