Multi-objective Optimization-based Selection for Quality-Diversity by Non-surrounded-dominated Sorting

Authors: Ren-Jian Wang, Ke Xue, Haopu Shang, Chao Qian, Haobo Fu, Qiang Fu

IJCAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We perform experiments on synthetic functions and several complex tasks (i.e., QDGym, robotic arm, and Mario environment generation), showing that NSS achieves better performance than not only other MO-based selection methods but also state-of-the-art selection methods in QD.
Researcher Affiliation Collaboration 1State Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, China 2Tencent AI Lab, Shenzhen, China {wangrj, xuek, shanghp, qianc}@lamda.nju.edu.cn, {haobofu, leonfu}@tencent.com
Pseudocode Yes Algorithm 1 Non-surrounded-dominated Sorting
Open Source Code Yes Our code is available at https://github.com/lamda-bbo/NSS.
Open Datasets Yes We conduct experiments on four different environments, i.e., QD Hopper, Walker, Half Cheetah, and Ant. These tasks aim to generate a set of policies that move forward as fast as possible and are diverse in the frequency of feet use. Thus, the objective function is determined by the agent s forward speed, and the behavior descriptor functions are defined as the fraction of time each foot was touching the ground during an episode.
Dataset Splits No The paper discusses the iterative process of QD algorithms and performance metrics like QD-Score, but it does not specify any training, validation, or test dataset splits in the context of standard supervised learning benchmarks.
Hardware Specification No The paper does not provide specific details about the hardware used for the experiments, such as GPU models, CPU types, or memory specifications. It only discusses running time.
Software Dependencies No The paper mentions various algorithms and frameworks (e.g., ME, PGA-ME, OG-ME) but does not list any specific software dependencies or their version numbers, such as programming languages, libraries, or solvers.
Experiment Setup Yes The detailed settings of experiments are provided in Appendix B.