Garnet MDPs
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/google-research/google-research/tree/master/rl_metrics_aaai2021
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一类随机生成的MDPs，其参数由状态和动作的数量决定，旨在评估不同指标对强化学习性能的影响。此外，Garnet MDPs相较于特定的MDPs，能够提供更少偏差的指标比较。任务是对不同指标对强化学习算法影响的实证评估。
This dataset consists of a class of randomly generated Markov Decision Processes (MDPs), whose parameters are determined by the number of states and actions. It is designed to evaluate the impact of different metrics on the performance of reinforcement learning algorithms. Moreover, compared with specific MDPs, Garnet MDPs can provide metric comparisons with lower bias. The associated task is to empirically evaluate the impact of different metrics on reinforcement learning algorithms.
