Uploaded July 2024 | Updated November 2026, 3 days ago
Institute for Assured Autonomy & Computer Science Seminar Series
July 19, 2022
“How We Trust a Black Box: Formal Verification of Deep Neural Networks”
Huan Zhang, Carnegie Mellon University
Neural networks have become a crucial element in modern artificial intelligence. When applying neural networks to mission-critical systems such as autonomous driving and aircraft control, it is often desirable to formally verify their trustworthiness in regards to safety and robustness. In this talk, Huan Zhang will first introduce the problem of neural network verification and the challenges of guaranteeing the behavior of a neural network given input specifications. Then he will discuss bound-propagation-based algorithms (e.g., CROWN and beta-CROWN), which are efficient, scalable, and powerful techniques for the formal verification of neural networks and which can also be generalizable to computational graphs beyond neural networks. His talk will highlight state-of-the-art verification techniques used in his α,β-CROWN (alpha-beta-CROWN) verifier that won the 2nd International Verification of Neural Networks Competition as well as novel applications of neural network verification.
Huan Zhang is a postdoctoral researcher at Carnegie Mellon University supervised by Professor Zico Kolter. Zhang received his PhD at the University of California, Los Angeles in 2020. His research focuses on the trustworthiness of artificial intelligence, especially on developing formal verification methods to guarantee the robustness and safety of machine learning. Zhang was awarded an IBM PhD Fellowship and led the winning team in the 2021 International Verification of Neural Networks Competition. He also received the 2021 New Frontiers in Adversarial Machine Learning Rising Star Award, sponsored by the MIT-IBM Watson AI Lab.
Institute for Assured Autonomy & Computer Science Seminar Series
July 19, 2022
“How We Trust a Black Box: Formal Verification of Deep Neural Networks”
Huan Zhang, Carnegie Mellon University
Neural networks have become a crucial element in modern artificial intelligence. When applying neural networks to mission-critical systems such as autonomous driving and aircraft control, it is often desirable to formally verify their trustworthiness in regards to safety and robustness. In this talk, Huan Zhang will first introduce the problem of neural network verification and the challenges of guaranteeing the behavior of a neural network given input specifications. Then he will discuss bound-propagation-based algorithms (e.g., CROWN and beta-CROWN), which are efficient, scalable, and powerful techniques for the formal verification of neural networks and which can also be generalizable to computational graphs beyond neural networks. His talk will highlight state-of-the-art verification techniques used in his α,β-CROWN (alpha-beta-CROWN) verifier that won the 2nd International Verification of Neural Networks Competition as well as novel applications of neural network verification.
Huan Zhang is a postdoctoral researcher at Carnegie Mellon University supervised by Professor Zico Kolter. Zhang received his PhD at the University of California, Los Angeles in 2020. His research focuses on the trustworthiness of artificial intelligence, especially on developing formal verification methods to guarantee the robustness and safety of machine learning. Zhang was awarded an IBM PhD Fellowship and led the winning team in the 2021 International Verification of Neural Networks Competition. He also received the 2021 New Frontiers in Adversarial Machine Learning Rising Star Award, sponsored by the MIT-IBM Watson AI Lab.










