Robustness and Security in ML Systems: Junfeng Yang

E6998 Robustness and Security in ML Systems

Spring 2021 -- Junfeng Yang

We put a tentative syllabus here to give you an idea what future may bring. This syllabus is subject to change as the course progresses.

# Day Date Topic Assignment Speakers
1 Tue Jan 12 Introduction Form reading group
2 Tue Jan 19 Deep learning Read Lecun-90c, AlexNet
3 Tue Jan 26 Adversarial ML (1) Read Intriguing properties of neural networks, FGSM attack
4 Tue Feb 2 Adversarial ML (2) Read PGD attack, Obfuscated gradients not useful
5 Tue Feb 9 Adversarial ML (3) Read Unrestricted attack, Blackbox attack
6 Tue Feb 16 Testing DL Read DeepXplore, VeriVis Guest: Kexin Pei
7 Tue Feb 23 Verifying DL (1) Read Reluplex, DeepSafe
8 Tue Mar 2 No class (Spring recess)
9 Tue Mar 9 Verifying DL (2) Read Reluval, Neurify Guest: Shiqi Wang
10 Tue Mar 16 Verifying DL (3) Read AI2, Abstract domain
11 Tue Mar 23 Robustness training Read Stability training, Adversarial logit training
12 Tue Mar 30 Robustness training (2) Read Metrics learning for robustness, Multitask learning for robustness Guest: Chengzhi Mao
13 Tue Apr 6 Robustness tradeoffs Read Robustness vs accuracy, Adversarial examples are features
14 Tue Apr 13 Mini-research conference Present and demo your final project