Ranked MNISTs
收藏资源简介:
Ranked MNISTs数据集是由Technical University of Munich等研究机构创建的,包含8个合成数据集,用于多标签排序(MLR)研究。这些数据集在生成时考虑了不同的显著性决定因素,提供了一个丰富且可控的实验环境。Ranked MNISTs数据集旨在解决多标签排序中数据稀缺和标注偏差的问题,通过引入正标签之间的排序,揭示了标签的隐含重要性和相关度。这些数据集对于研究MLR领域的新方法提供了宝贵的资源,并有助于理解和解决MLR中的挑战。
The Ranked MNISTs dataset was developed by research institutions including the Technical University of Munich, and comprises 8 synthetic datasets tailored for multi-label ranking (MLR) research. These datasets are constructed with diverse salience determinants taken into account, creating a rich and controllable experimental environment. The Ranked MNISTs dataset is designed to tackle the problems of data scarcity and annotation bias in multi-label ranking, and unveils the implicit importance and relevance of labels by introducing ranking among positive labels. These datasets serve as a valuable resource for researching novel methods in the MLR field, and facilitate the understanding and resolution of challenges in MLR.




