Kevin Roth

  • e-mail
  • room CAB F 42.1
  • address Universitätstrasse 6, 8092 Zürich, Switzerland


My research focuses on adversarial deep learning, more specifically Generative Adversarial Networks (GANs) and robustness to adversarial attacks. I am currently pursuing a joint PhD in the Data Analytics group under the supervision of Prof. Thomas Hofmann at ETH and Dr. Sebastian Nowozin at Microsoft Research. My work on “Stabilizing Training of Generative Adversarial Networks through Regularization” has recently been published at NIPS'17 . Prior to that I was developing a model of interdependent neural networks together with an algorithm that allows to localize the most influential nodes responsible for broadcasting information in the brain. I completed my bachelors and masters in theoretical physics at ETH with main focus on quantum field theory and statistical physics.

My publications: Google Scholar

Youtube interview: Stabilizing Training of GANs
Blog posts: www.inference.vc, www.microsoft.com