Hire an Illini
Linyi Li
- Advisor:
- Bo Li and Tao Xie
- Departments:
- Areas of Expertise:
- Security
- Machine Learning
- Thesis Title:
- Enabling large-scale certifiably trustworthy deep learning systems
- Thesis abstract:
- Given the rising societal safe and ethical concerns for modern deep learning systems in deployment, designing certifiable large-scale deep learning systems for real-world requirements is in urgent demand. This thesis proposes a framework for building certifiable large-scale deep learning systems towards trustworthy machine learning, achieving robustness against noise perturbations, semantic transformations, poisoning attacks, distributional shifts; fairness; and reliability against numerical defects. The framework is applied to modern deep reinforcement learning and computer vision systems, demonstrating its effectiveness. The shared core backbone for designing certifiable deep learning systems includes threat-model-dependent smoothing, efficient and exact model state abstraction, statistical worst-case characterization, and diversity-enhanced model training. The end of the thesis summarizes several challenges in certifiable ML, such as scalability challenges, tightness challenges, deployment challenges, the gap between theory and practice, and the societal implications/impacts of certifiably trustworthy ML.
- Downloads:
Contact information:
linyi2@illinois.edu