📡 Embodied AI Radar具身智能雷达

Topics主题 › World Models世界模型

Paper论文 2019 · University of Toronto, Google

Dream to Control: Learning Behaviors by Latent Imagination

The first Dreamer — learns a latent world model from pixels and trains the policy entirely on imagined rollouts.第一代 Dreamer:从像素学习潜空间世界模型,并完全在想象的轨迹上训练策略。

BibTeX
@misc{hafner2019dream,
  title = {{Dream to Control: Learning Behaviors by Latent Imagination}},
  author = {Danijar Hafner and Timothy Lillicrap and Jimmy Ba and Mohammad Norouzi},
  year = {2019},
  eprint = {1912.01603},
  archivePrefix = {arXiv},
  primaryClass = {cs.LG},
  url = {https://arxiv.org/abs/1912.01603},
}
World Models世界模型model-based-rllatent-dynamics

Authors (from arXiv): 作者(来自 arXiv):Danijar Hafner, Timothy Lillicrap, Jimmy Ba, Mohammad Norouzi

More from: 同机构更多条目:Google DeepMind / Google谷歌 DeepMind / 谷歌 · University of Toronto多伦多大学

More in World Models更多世界模型条目

All World Models entries →查看全部世界模型条目 →