Meta Reinforcement Learning with Autonomous Inference of Subtask Dependencies

This repository is an implementation of our ICLR 2020 Meta Reinforcement Learning with Autonomous Inference of Subtask Dependencies in PyTorch. If you use this work, please cite:

@inproceedings{
        Sohn2020Meta,
        title={Meta Reinforcement Learning with Autonomous Inference of Subtask Dependencies},
        author={Sungryull Sohn and Hyunjae Woo and Jongwook Choi and Honglak Lee},
        booktitle={International Conference on Learning Representations},
        year={2020},
        url={https://openreview.net/forum?id=HkgsWxrtPB}
}

Our codebase is built on top of subtask-graph-execution and pytorch-a2c-ppo-acktr-gail for Mining & Playground and the older (deprecated) version of the current Reaver for StarCraft II domain.

We have provided two separate codebases for each of our grid-world and StarCraft II domains inside the grid-world and sc2 directories. Please refer to README in each directory for instructions on installation and running the code.

srsohn/msgi

Meta Reinforcement Learning with Autonomous Inference of Subtask Dependencies