1 paper · 1 filter
Houpu Yao, Malcolm Regan, Yezhou Yang +1
We demonstrate in this paper that a generative model can be designed to perform classification tasks under challenging settings, including adversarial attacks and input distributio…