micahr234/play_ns_frozenlake_random

该数据集是一个强化学习数据集，包含环境交互记录，涉及多个特征如环境名称（env_name）、环境索引（env_idx）、全局步数（global_step）、回合步数（episode_step）、动作（action）、奖励（reward）、转换后奖励（xformed_reward）、完成标志（done）、离散观察（observation_discrete）以及元数据（如P_flag、P、q_star）。数据划分为训练集（400,000个示例）和评估集（80,000个示例），适用于强化学习算法的训练和评估任务。This dataset is a reinforcement learning dataset containing records of environment interactions, with features such as environment name (env_name), environment index (env_idx), global step (global_step), episode step (episode_step), action (action), reward (reward), transformed reward (xformed_reward), done flag (done), discrete observation (observation_discrete), and metadata (e.g., P_flag, P, q_star). The data is split into training (400,000 examples) and evaluation (80,000 examples) sets, suitable for training and evaluating reinforcement learning algorithms.