EAT-C: Environment-Adversarial sub-Task Curriculum for Efficient Reinforcement Learning
Authors: Shuang Ao, Tianyi Zhou, Jing Jiang, Guodong Long, Xuan Song, Chengqi Zhang
ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on diverse tasks demonstrate the advantages of our method on improving RL s efficiency and generalization. |
| Researcher Affiliation | Academia | 1University of Technology Sydney 2University of Washington, Seattle 3University of Maryland, College Park 4Southern University of Science and Technology. |
| Pseudocode | Yes | Algorithm 1 Top-Down Planning of Sub-task Curriculum |
| Open Source Code | No | The paper provides a link to a third-party baseline's code (POET) but does not state that the code for EAT-C is open-source or provide a link. |
| Open Datasets | Yes | 2D pusher (Yamada et al., 2020). Discrete space tasks (Maxime Chevalier-Boisvert & Pal, 2018). (Jurgenson et al., 2020) for the robotic arm. |
| Dataset Splits | No | The paper states 'We train and test RL policies on three types of compositional tasks' and 'We randomly sample diverse environments and tasks for training and test', but does not specify a validation dataset split or explicit percentages for train/test/validation. |
| Hardware Specification | No | The paper mentions using 'MuJoCo as the simulator' but does not provide specific details about the hardware used to run experiments, such as GPU/CPU models or memory. |
| Software Dependencies | No | The paper mentions 'Soft Actor Critic (SAC)' and 'MuJoCo' but does not provide specific version numbers for these or other software dependencies. |
| Experiment Setup | Yes | A complete list of hyperparameters for SAC in 2D-pusher tasks is given in Table 5. A complete list of hyperparameters of SAC in the discrete space tasks is given in Table. 6. |