Robust Learning for Data Poisoning Attacks

                  Yunjuan Wang 1 Poorya Mianjy 1 Raman Arora 1

             Abstract                 in settings where an adversary can affect any part of the
                                  training data. Therefore, in this paper, we are interested in
   We investigate the robustness of stochastic ap-        quantifying the maximal adversarial noise that is tolerable
   proximation approaches against data poisoning         by SGD when training wide ...