Seminars and Events

Artificial Intelligence Seminar

How Can We Trust a Black-box? A Quest for Scalable and Powerful Neural Network Verifiers

Event Details

Neural networks have become a crucial element in modern artificial intelligence. However, they are often black-boxes and can behave unexpectedly and produce surprisingly wrong results under malicious inputs. When applying neural networks to mission-critical systems such as autonomous driving and aircraft control, it is often desirable to formally verify their trustworthiness such as safety and robustness. Unfortunately, the complexity of neural networks has made the task of formally verifying their properties very challenging. To tackle this challenge, I first propose an efficient verification algorithm based on linear relaxations of neural networks, which produces guaranteed output bounds given bounded input perturbations. The algorithm propagates linear inequalities through the network efficiently in a backward manner and can be applied to arbitrary network architectures. To reduce relaxation errors, I develop an efficient optimization procedure that can tighten verification bounds rapidly on machine learning accelerators such as GPUs. Lastly, I discuss how to further empower the verifier with branch and bound by incorporating the additional branching constraints into the bound propagation procedure. The combination of these advanced neural network verification techniques leads to α,β-CROWN, a scalable, powerful and GPU-based neural network verifier that is up to 2 to 3 orders of magnitudes faster than traditional CPU based verifier and won the 2nd International Verification of Neural Networks Competition (VNN-COMP’21) with the highest total score over a set of 8 benchmarks.

Speaker Bio

Huan Zhang is a postdoctoral researcher at Carnegie Mellon University, supervised by Prof. Zico Kolter. He received his Ph.D. degree in Computer Science at UCLA in 2020, advised by Prof. Cho-Jui Hsieh. Huan received his Master of Science degree at UC Davis in 2014. Huan's research focuses on the robustness and trustworthiness of artificial intelligence (AI), especially on using formal verification and provable methods to evaluate and enhance the robustness of machine learning models, such as deep neural networks and tree ensembles. Huan systematically studied the robustness of many machine learning scenarios including reinforcement learning, natural language processing, and image generation, and he has co-authored over 40 papers in top-tier machine learning and AI conferences. Huan led the development of α,β-CROWN, a toolbox for neural network robustness verification, which won the first prize in Verification of Neural Networks Competition (VNN-COMP'21), outperforming 11 competing verification tools. Huan Zhang was awarded an IBM Ph.D. fellowship during 2018 - 2020.

Host: Muhao Chen, POC: Amy Feng

YOU ONLY NEED TO REGISTER ONCE TO ATTEND THE ENTIRE SERIES - We will send you email announcements with details of the upcoming speakers.

Register in advance for this webinar:

After registering, you will receive an email confirmation containing information about joining the Zoom webinar. Please save the Zoom webinar link to your calendar.

The recording for this AI Seminar talk will be posted on our USC/ISI YouTube page within 1-2 business days: