Weighted model estimation for offline model-based reinforcement learning

Authors: Toru Hishinuma, Kei Senda

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

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical experiments demonstrate the effectiveness of weighting with the artificial weight. 6 Numerical Experiment
Researcher Affiliation Academia Toru Hishinuma Kyoto University hishinuma.toru.43n@kyoto-u.jp Kei Senda Kyoto University senda@kuaero.kyoto-u.ac.jp
Pseudocode Yes Algorithm 1 Weighted model estimation for policy evaluation (full version).
Open Source Code No The paper mentions modifying existing code ('This paper implements SAC by modifying the implementation code by [36]') but does not explicitly state that the source code for their own methodology is made publicly available or provide a link to it.
Open Datasets Yes This paper studies policy optimization on the D4RL Benchmark [33] based on the Mu Jo Co simulator [34].
Dataset Splits No The paper mentions using the D4RL Benchmark datasets but does not explicitly provide specific training/validation/test dataset splits, such as percentages or sample counts, within its text.
Hardware Specification No The paper does not provide specific hardware details such as GPU or CPU models, or any other computer specifications used for running its experiments.
Software Dependencies No The paper mentions using PyTorch (implicitly via a reference to a PyTorch SAC implementation) and the MuJoCo simulator, but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes The agent uses Pθ represented by two-layer neural networks with 8 units with tanh activation.