MAME : Model-Agnostic Meta-Exploration

被引:0
作者
Gurumurthy, Swaminathan [1 ]
Kumar, Sumit [1 ]
Sycara, Katia [1 ]
机构
[1] Carnegie Mellon Univ, Inst Robot, Pittsburgh, PA 15213 USA
来源
CONFERENCE ON ROBOT LEARNING, VOL 100 | 2019年 / 100卷
关键词
Meta-Learning; Exploration; Self-Supervised Learning;
D O I
暂无
中图分类号
TP39 [计算机的应用];
学科分类号
081203 ; 0835 ;
摘要
Meta-Reinforcement learning approaches aim to develop learning procedures that can adapt quickly to a distribution of tasks with the help of a few examples. Developing efficient exploration strategies capable of finding the most useful samples becomes critical in such settings. Existing approaches towards finding efficient exploration strategies add auxiliary objectives to promote exploration by the pre-update policy, however, this makes the adaptation using a few gradient steps difficult as the pre-update (exploration) and post-update (exploitation) policies are often quite different. Instead, we propose to explicitly model a separate exploration policy for the task distribution. Having two different policies gives more flexibility in training the exploration policy and also makes adaptation to any specific task easier. We show that using self-supervised or supervised learning objectives for adaptation allows for more efficient inner-loop updates and also demonstrate the superior performance of our model compared to prior works in this domain.
引用
收藏
页数:13
相关论文
共 27 条
  • [1] Stadie BC, 2019, Arxiv, DOI arXiv:1803.01118
  • [2] Kulkarni TD, 2016, Arxiv, DOI arXiv:1606.02396
  • [3] Duan Y, 2016, PR MACH LEARN RES, V48
  • [4] Duan Y, 2016, Arxiv, DOI [arXiv:1611.02779, 10.48550/arXiv.1611.02779.82]
  • [5] Finn C, 2018, Arxiv, DOI arXiv:1710.11622
  • [6] Finn C, 2017, PR MACH LEARN RES, V70
  • [7] Florensa C, 2019, Arxiv, DOI arXiv:1901.00943
  • [8] Foerster J, 2018, Arxiv, DOI [arXiv:1802.05098, DOI 10.48550/ARXIV.1802.05098]
  • [9] Gupta Abhishek, 2018, Advances in Neural Information Processing Systems, V31
  • [10] Kahn G, 2018, IEEE INT CONF ROBOT, P5129, DOI 10.1109/ICRA.2018.8460655