XIE CIHANG

Training robust neural networks via smooth activation functions