黑盒系统的贝叶斯安全验证（英文版）

英文标题：Bayesian Safety Validation for Black-Box Systems中文摘要：本文提出了一种基于贝叶斯优化的安全验证算法，它通过迭代地适配概率代理模型来高效预测故障，并通过重要性采样来估计操作域内的故障概率。经实验表明，该算法在减少样本数和各种安全验证指标方面表现良好，并可用于补充机器学习组件的 FAA 认证过程。英文摘要：Accurately estimating the probability of failure for safety-critical systemsis important for certification. Estimation is often challenging due tohigh-dimensional input spaces, dangerous test scenarios, and computationallyexpensive simulators; thus, efficient estimation techniques are important tostudy. This work ref