Non-Negative Networks Against Adversarial Attacks

William Fleshman, Edward Raff, Jared Sylvester +2

Adversarial attacks against neural networks are a problem of considerable importance, for which effective defenses are not yet readily available. We make progress toward this probl…