Variational Empowerment as Representation Learning for Goal-Conditioned Reinforcement Learning

Authors: Jongwook Choi, Archit Sharma, Honglak Lee, Sergey Levine, Shixiang Shane Gu

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Through principled mathematical derivations and careful experimental studies, our work lays a novel foundation from which to evaluate, analyze, and develop representation learning techniques in goal-based RL.
Researcher Affiliation Collaboration Jongwook Choi 1 Archit Sharma 2 Honglak Lee 1 3 Sergey Levine 4 5 Shixiang Shane Gu 4 Work done while an intern at Google. 1University of Michigan 2Stanford University 3LG AI Research 4Google Research 5University of California, Berkeley.
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any statements or links indicating that source code for the described methodology is publicly available.
Open Datasets Yes We evaluate the performance of several variants of VGCRL on standard locomotion tasks (Brockman et al., 2016).
Dataset Splits No The paper mentions evaluating performance on standard locomotion tasks but does not specify concrete training, validation, or test data splits.
Hardware Specification No The paper mentions using Mujoco for simulations but does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used for running the experiments or training.
Software Dependencies No The paper mentions
Experiment Setup Yes Evaluation of Latent Goal-Reaching Metric on Mu Jo Co control suites, after a total of 10M environment steps of training.