arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.01629cs.LGcs.ROcs.SYeess.SY
收藏

AdaptNC: 适应性非符合分数用于分布偏移下的符合预测

AdaptNC: Adaptive Nonconformity Scores for Conformal Prediction under Distribution Shift

Department of Electrical and Systems Engineering, University of Pennsylvania(宾夕法尼亚大学电气与系统工程系)

Renukanandan Tumu, Aditya Singh, Rahul Mangharam

AI总结:

本文提出AdaptNC框架,通过联合在线调整非符合分数参数和符合阈值,解决分布偏移下的预测区域保守问题,实验表明其在保持覆盖率的同时显著减少预测区域体积。

摘要:

严格不确定性量化对于在无约束环境中安全部署自主系统至关重要。符合预测(CP)提供了一个无分布的框架来完成这项任务,但其标准形式依赖于交换性假设,这些假设在现实世界机器人中会被分布偏移所违反。现有的在线CP方法通过自适应地调整符合阈值来保持目标覆盖率,但通常使用固定的非符合分数函数。我们显示,这种固定的几何结构在环境经历结构性偏移时会导致高度保守、体积低效的预测区域。为此,我们提出AdaptNC,一个用于联合在线调整非符合分数参数和符合阈值的框架。AdaptNC利用自适应重加权方案来优化分数函数,并引入一个回放缓冲机制以减轻分数转换期间出现的覆盖率不稳定性。我们在涉及多智能体策略变化、环境变化和传感器退化等多种机器人基准上评估了AdaptNC。我们的结果表明,与现有最先进的仅阈值基线相比,AdaptNC在保持目标覆盖率水平的同时显著减少了预测区域体积。

英文摘要:

Rigorous uncertainty quantification is essential for the safe deployment of autonomous systems in unconstrained environments. Conformal Prediction (CP) provides a distribution-free framework for this task, yet its standard formulations rely on exchangeability assumptions that are violated by the distribution shifts inherent in real-world robotics. Existing online CP methods maintain target coverage by adaptively scaling the conformal threshold, but typically employ a static nonconformity score function. We show that this fixed geometry leads to highly conservative, volume-inefficient prediction regions when environments undergo structural shifts. To address this, we propose $\textbf{AdaptNC}$, a framework for the joint online adaptation of both the nonconformity score parameters and the conformal threshold. AdaptNC leverages an adaptive reweighting scheme to optimize score functions, and introduces a replay buffer mechanism to mitigate the coverage instability that occurs during score transitions. We evaluate AdaptNC on diverse robotic benchmarks involving multi-agent policy changes, environmental changes and sensor degradation. Our results demonstrate that AdaptNC significantly reduces prediction region volume compared to state-of-the-art threshold-only baselines while maintaining target coverage levels.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑