Reproduce “Worst-Case Risk Quantification under Distributional Ambiguity using Kernel Mean Embedding in Moment Problem” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jia-Jie Zhu, Wittawat Jitkrittum, Moritz Diehl, Bernhard Schölkopf
The authors present a method for estimating worst-case constraint-violation probability despite zero observed violations. It represents nearby distributions using kernel mean embeddings and maximizes probability on violating states. In a Van der Pol control simulation, estimated risk rises as the state approaches the constraint. The finite estimate is a lower bound, not a safety guarantee; convergence is conditional.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.