Machine Learning and Friends Lunch: Andrew Wu, Marabou 2.0: A Versatile Formal Analyzer of Neural Networks
Content
Speaker
Abstract
Deep neural networks are revolutionizing the way complex systems are designed. Consequently, there is a pressing need for tools and techniques to obtain formal guarantees on the behaviors of neural networks. To address that need, we present version 2.0 of Marabou, a toolkit for formally verifying user-defined properties on deep neural networks. We discuss the tool’s architectural design and highlight a few of its many recent applications, in robotics, vision, and systems domains.
Bio
Andrew (Haoze) Wu is an Assistant Professor in Computer Science at Amherst College. He recently obtained his PhD in Computer Science from Stanford University. Andrew's work aims to bring together automated reasoning and machine learning to create safer and smarter computer systems. He is the lead developer of the Marabou framework, a state-of-the-art neural network verification tool widely used in academia and industry. His work has been published at top venues in formal methods and artificial intelligence.