LHD

LHD是由中国科学院软件研究所的研究人员Jiuling Zhang和Zhiming Ding提出的一种新的数据集，旨在改进DARTS搜索空间，提供一个更大且更具挑战性的基准。该数据集通过结合未充分探索的影响因素，如传导鲁棒性和离散化策略，构建了一个基于多条件评估的基准。LHD数据集的应用领域主要集中在神经架构搜索（NAS）的评估和改进，旨在解决现有数据集在评估方法排名时存在的准确度范围狭窄的问题。

LHD is a novel dataset proposed by researchers Jiuling Zhang and Zhiming Ding from the Institute of Software, Chinese Academy of Sciences. It is designed to improve the DARTS search space and provide a larger, more challenging benchmark. This dataset constructs a multi-condition evaluation-based benchmark by integrating under-explored influencing factors such as conduction robustness and discretization strategies. The application domains of the LHD dataset primarily focus on the evaluation and improvement of Neural Architecture Search (NAS), aiming to address the problem of narrow accuracy range in existing datasets when ranking evaluation methods.