Exploration-Guided Reward Shaping for Reinforcement Learning under Sparse Rewards

Authors: Rati Devidze, Parameswaran Kamalaruban, Adish Singla

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on several environments with sparse/noisy reward signals demonstrate the effectiveness of EXPLORS.
Researcher Affiliation Academia 1Max Planck Institute for Software Systems (MPI-SWS), Saarbrucken, Germany 2The Alan Turing Institute, London, UK
Pseudocode Yes Algorithm 1 Online Reward Shaping
Open Source Code Yes 1Github repo: https://github.com/machine-teaching-group/neurips2022_exploration-guided-reward-shaping.
Open Datasets No The paper describes custom environments (CHAIN, ROOM, LINEK) but does not provide concrete access information (link, DOI, specific citation) for these datasets to be publicly available.
Dataset Splits No The paper does not explicitly provide training/validation/test dataset splits. It describes evaluation during training based on extrinsic rewards.
Hardware Specification No The paper states "Details are provided in appendices" regarding compute resources, but no specific hardware details are given in the main text.
Software Dependencies No The paper mentions using "tabular REINFORCE agent [7]" and "tabular Q-learning agent [7]", and a "neural REINFORCE agent", but does not provide specific version numbers for any software dependencies (e.g., Python, PyTorch, TensorFlow).
Experiment Setup Yes We give an overview of main results here, and provide a more detailed description of the setup and additional implementation details in appendices.