Tuesday Afternoon Contributed Talks

07 May 2019 |

Part of my series of notes from ICLR 2019 in New Orleans.

Slalom: Fast, Verifiable and Private Execution of Neural Networks in Trusted Hardware

slalom

secure matrix

Learning to Remember More with Less Memorization

Learning Robust Representations by Projecting Superficial Statistics Out

ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness

shape / texture

distortions