ARNO

该数据集包含了在三种不同环境中实现的具有三个噪声级别的时序相关异常的ARNO场景，这些环境包括Cartpole、Reacher和Acrobot。具体来说，这些场景包含了轻度、中度和重度噪声，这些噪声会影响代理的奖励。数据集的规模覆盖了多种环境以及三个不同的噪声水平。其任务是强化学习中的分布外检测。

This dataset comprises ARNO scenarios with time-series dependent anomalies and three noise levels, implemented across three distinct environments: Cartpole, Reacher, and Acrobot. Specifically, these scenarios incorporate mild, moderate, and severe noise that impacts the agent's rewards. The dataset covers a range of environments alongside three distinct noise levels. The core task addressed by this dataset is out-of-distribution detection in reinforcement learning.