Hierarchically Decoupled Imitation For Morphological Transfer
Authors: Donald Hejna, Lerrel Pinto, Pieter Abbeel
ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section we discuss empirical results on using hierarchically decoupled imitation for transferring policies across morphologies. |
| Researcher Affiliation | Academia | Donald J. Hejna III 1 Pieter Abbeel 1 Lerrel Pinto 1 2 1Department of EECS, University of California, Berkeley 2Computer Science, New York University. |
| Pseudocode | Yes | Algorithm 1 Low-level Alignment for high-level transfer |
| Open Source Code | Yes | Our code and videos can be found at https: //sites.google.com/berkeley.edu/ morphology-transfer. |
| Open Datasets | Yes | All of these environments are simulated in Mu Jo Co (Todorov et al., 2012) using the Open AI Gym interface. Images of the environments can be found in Figure 3. More details of the environments and the agents are provided in Appendix C. |
| Dataset Splits | No | The paper describes how policies are trained (e.g., “pre-train the low-level policy πlo B on uniformly sampled goals from G”) and evaluated (e.g., “results averaged across a hundred episodes per run” and “Maze E refers to the Maze End evaluation and Maze S refers to the Maze Sample evaluation”), but it does not specify explicit numerical percentages or counts for fixed train/validation/test dataset splits typical for supervised learning tasks. |
| Hardware Specification | No | The paper acknowledges “AWS for computing resources” but does not provide specific hardware details such as GPU models, CPU types, or memory specifications used for running experiments. |
| Software Dependencies | No | The paper states “We use SAC as our base RL optimizer, and re-purpose the open-source Stable Baselines (Hill et al., 2018) codebase for our methods.” and mentions “Mu Jo Co (Todorov et al., 2012) using the Open AI Gym interface”, but does not provide specific version numbers for these software dependencies. |
| Experiment Setup | No | The paper states “Additional training details including hyper-parameters set are included in Appendix D.” This indicates that detailed experimental setup information, including hyperparameters, is deferred to an appendix and not provided in the main text. |