ASPiRe: Adaptive Skill Priors for Reinforcement Learning

Authors: Mengda Xu, Manuela Veloso, Shuran Song

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

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments demonstrate that ASPi Re can significantly accelerate the learning of new downstream tasks in the presence of multiple priors and show improvement on competitive baselines. We evaluate our method in three modified environments from D4RL [48] and one modified environment from robosuite environment [49].
Researcher Affiliation Collaboration Mengda Xu 1, 2, Manuela Veloso 2,3, Shuran Song 1 1 Department of Computer Science, Columbia University 2 J.P. Morgan AI Research 3 School of Computer Science, Carnegie Mellon University (emeritus)
Pseudocode Yes Algorithm 1 ASPi Re Algorithm
Open Source Code Yes The code is in the supplement material. We provide the codes to regenerate the datasets and downstream task learning and all instructions are inside readme.md
Open Datasets Yes We evaluate our method in three modified environments from D4RL [48] and one modified environment from robosuite environment [49]. We use the D4RL benchmark and cite in the section 4.
Dataset Splits Yes Did you specify all the training details (e.g., data splits, hyperparameters, how they were chosen)? [Yes] See appendix A.4
Hardware Specification Yes Did you include the total amount of compute and the type of resources used (e.g., type of GPUs, internal cluster, or cloud provider)? [Yes] See appendix A.4
Software Dependencies No No specific software version numbers (e.g., Python 3.x, PyTorch 1.x) are explicitly mentioned in the paper's main text or checklist for reproducibility.
Experiment Setup Yes See appendix A.4 for details on environment, data collection process and training. Did you specify all the training details (e.g., data splits, hyperparameters, how they were chosen)? [Yes] See appendix A.4