Abstract: Deep learning has rapidly emerged as a transformative technology that permeates all modern software, from autonomous driving systems to malware-detection tools. Considering the critical role of this software in our technologies, it must behave as intended. However, the complexity introduced by deep-learning components complicates formal reasoning about the behavior of such software, frequently resulting in solutions that offer only empirical or no guarantees. My research contributes techniques and algorithms that increase the trustworthiness of deep-learning-powered software by providing strong provable guarantees across the components existing in the entire deep learning pipeline.
Bio: Yuhao Zhang is a Ph.D. candidate at the University of Wisconsin–Madison’s Department of Computer Science, advised by Prof. Loris D’Antoni and Prof. Aws Albarghouthi. He earned his B.S. in Computer Science at Peking University in 2019 with the distinguished Summa Cum Laude honor. Yuhao’s research spans the fields of software engineering, programming languages, and deep learning. His research aims to ensure the safety, security, and reliability of all components of deep-learning-powered software by providing provable guarantees. Yuhao’s work has appeared in prestigious conferences, including ICSE, FSE, ISSTA, OOPSLA, NeurIPS, ICML, and EMNLP.