Effective and Robust Adversarial Training against Data and Label Corruptions

arXiv:2405.04191v1 Announce Type: new
Abstract: Corruptions due to data perturbations and label noise are prevalent in the datasets from unreliable sources, which poses significant threats to model training. Despite existing efforts in developing robust models, current learning methods commonly overlook the possible co-existence of both corruptions, limiting the effectiveness and practicability of the model.
In this paper, we develop an Effective and Robust Adversarial Training (ERAT) framework to simultaneously handle two types of corruption (i.e., data and label) without prior …

abstract adversarial adversarial training arxiv cs.cv cs.lg current data datasets noise robust robust models threats training type