Activations Dataset

该数据集包含了机器人在与环境互动过程中收集的隐藏激活数据，以及相应的行为记录。该数据集被用于训练变分自编码器（VAE），旨在重建神经元激活状态，并以此提高演员-评论家模型的性能。数据在2000个剧集的基础上进行收集，任务是通过神经元激活所得到的内部状态来训练演员-评论家模型。

This dataset comprises hidden activation data collected during a robot's interaction with its environment, as well as corresponding behavioral records. It is employed for training a Variational Autoencoder (VAE), with the aim of reconstructing neuronal activation states and consequently improving the performance of the Actor-Critic model. The data was collected across 2000 episodes, and its target task is to train the Actor-Critic model using the internal states derived from neuronal activations.