rene_mobile’s Twitter Archive—№ 8,496

@mmirman If I interpret this correctly, there is provable behavior against some adversarial input attacks, but no good results in terms of adversarial training or even plain robustness wrt. noisy/untrained input?