Ambiguity of distriubtions
Distributionally robust optimization addresses optimization problems under uncertain probability distributions by considering a set of potential distributions rather than a single known distribution. It aims to find solutions that perform well across all possible distributions within the set, providing robustness against model misspecification and ensuring performance under various scenarios.
Our research contributions include proposing various distributionally robust optimization methods, such as kernel-based and Sinkhorn-based approaches, and demonstrating their efficacy through numerical examples and optimization problems. These methods offer advantages over existing approaches and show superior performance in handling uncertainty.
S. Yang, Z. Li. Distributionally Robust Chance-Constrained Optimization with Deep Kernel Ambiguity Set. The 7th International Symposium on Advanced Control of Industrial Processes, Vancouver, Canada, 2022.
