Automatic Successive Reinforcement Learning with Multiple Auxiliary Rewards

Authors: Zhao-Yang Fu, De-Chuan Zhan, Xin-Chun Li, Yi-Xing Lu

IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments and simulations have shown the superiority of our proposed ASR on a range of environments, including Open AI classical control domains and video games; Freeway and Catcher.
Researcher Affiliation Academia National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, China {fuzy,lixc}@lamda.nju.edu.cn, zhandc@nju.edu.cn, Yixing Lu97@gmail.com
Pseudocode Yes A summary of our ASR framework is shown in Algorithm 1.
Open Source Code No The paper references 'Open AI Baselines' with a GitHub link, but does not provide a specific link or statement for the source code of their proposed ASR framework.
Open Datasets Yes Open AI classical control domains Mountain Car, Cart Pole and Acrobot [Brockman et al., 2016], PLE game Catcher1 and Atari game Freeway.
Dataset Splits No The paper does not explicitly provide details on training/validation/test dataset splits, or how data is partitioned for a validation set.
Hardware Specification No No specific hardware (GPU/CPU models, memory, etc.) used for running experiments is explicitly mentioned in the paper.
Software Dependencies No The paper mentions 'Open AI Baselines' but does not provide specific version numbers for software dependencies or libraries used.
Experiment Setup Yes For classical control environments, we perform one million time steps of training for each method. For video games, we perform ten million time steps of training for each method. Each method is run with five random seeds.