DQN Replay Dataset
收藏arXiv2025-09-30 收录
下载链接：
https://offline-rl.github.io/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了50,000个在在线DQN智能体训练过程中收集的转换，每个转换的形式为（s_t^i, a_t^i, r_t^i, s_{t+1}^i）。此外，每个剧集结束后，该数据集会被划分为80%用于训练，20%用于评估。这一数据集的规模为50,000个转换，其任务是针对恶化的水管制定最优的修复策略。
This dataset contains 50,000 transitions collected during the training process of an online DQN agent, where each transition takes the form of (s_t^i, a_t^i, r_t^i, s_{t+1}^i). After each episode concludes, the dataset is split into 80% for training and 20% for evaluation. With a total of 50,000 transitions, this dataset is designed to develop optimal repair strategies for deteriorated water pipelines.
提供机构：
Generated from the DQN training process
