Softmax Deep Double Deterministic Policy Gradients

Authors: Ling Pan, Qingpeng Cai, Longbo Huang

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive experiments on challenging continuous control tasks, and results show that SD3 outperforms state-of-the-art methods.
Researcher Affiliation Collaboration Ling Pan1, Qingpeng Cai2, Longbo Huang1 1Institute for Interdisciplinary Information Sciences, Tsinghua University pl17@mails.tsinghua.edu.cn, longbohuang@tsinghua.edu.cn 2Alibaba Group qingpeng.cqp@alibaba-inc.com
Pseudocode Yes Algorithm 1 SD3
Open Source Code Yes Details for hyperparameters are in Appendix E.1, and the implementation details are publicly available at https://github.com/ling-pan/SD3.
Open Datasets Yes We conduct extensive experiments in standard continuous control tasks from Open AI Gym [6] to evaluate the SD3 algorithm.
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning. It uses standard environments but does not detail their splits.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper mentions software like Open AI Gym, Mu Jo Co, Box2d, and refers to implementations of DDPG and TD3, but it does not provide specific version numbers for these or any other ancillary software components needed to replicate the experiment.
Experiment Setup Yes For the softmax operator in SD3, the number of noises to sample K is 50, and the parameter β is mainly chosen from {10 3, 5 10 3, 10 2, 5 10 2, 10 1, 5 10 1} using grid search. All other hyperparameters of SD3 are set to be the same as the default setting for TD3 on all tasks except for Humanoid-v2... Details for hyperparameters are in Appendix E.1