Global Guarantee

Global guarantee research focuses on establishing rigorous mathematical assurances for the performance and stability of complex systems, particularly those involving machine learning models. Current efforts concentrate on developing algorithms and model architectures (like Extended Linearized Contracting Dynamics and smoothed online quadratic optimization methods) that provide provable global guarantees, such as global stability and convergence to optimal solutions, even in the presence of uncertainty or non-convexity. This work is crucial for deploying machine learning in safety-critical applications (e.g., power grid management, control systems) where reliable performance is paramount, and for advancing the theoretical understanding of non-convex optimization.