CS 595C
Neural Network Verification
Spring 2022
Description:
In recent years Neural Networks (NNs) have attracted tremendous attention
both in academia and industry. This is due to the incredible success of NN
based techniques in a wide spectrum of domains including computer vision,
speech recognition, and natural language processing. However, due to the
increasing adoption of NNs in safety-critical and socially sensitive domains
such as self-driving cars, robotics, computer security, criminal justice,
and medical diagnosis, there is a pressing need for developing verification
techniques that can provide guarantees about dependability and safety of
NN applications. Formal verification techniques can provide guarantees of
correctness, however, their scalability can be limited. In this seminar
we will discuss recent research on verification of neural networks.
Instructors:
Tevfik Bultan
Yufei Ding
Weekly meeting time:
TBD
Units: This will be a 2 unit seminar
Course Work
- Each student has to read all the papers that are presented every week and participate in the discussion.
- Each student will be asked to present a paper:
- Please use/prepare slides for presenting the paper.
- After the presentation and discussion, the presenter should prepare a summary of the discussion and future research directions based on the presented work.
Schedule
- Week 1: Thursday, March 31st, 1:00pm. Organizational meeting (via zoom).
- Week 2: Friday, April 8th, 2:00pm, HFH 1132.
- Zheng Wang will present "Faith: An Efficient Framework for Transformer Verification on GPUs"
- Mara Downing will present "Quantitative Verification of Quantized Neural Networks"
Paper List
Paper list.
If there are other papers thati you would like to discuss on neural network verification, please email the instructors with the paper information.