Feature space and label space selection based on Error-correcting output codes for partial label learning

Guangyi Lin, Zi-Yang Xiao, Jia-Tong Liu, Beizhan Wang, KunHong Liu, Qingqiang Wu

Abstract: Partial label learning (PLL) is a type of weakly supervised learning. This paper proposes a new heuristic algorithm for Partial Label learning through the combination of Feature space And Label space using Error-correcting output codes (PL-FALE for short). In the proposed framework, the feature space of different training sample subsets is exploited to generate diverse positive group and negative group pairs by the divide-and-conquer strategy. Then, the training sample label space is exploited to deal with the partial label overlap in every group pair. A set of experiments are conducted on nine controlled UCI datasets and five real-world datasets, and the experimental results show that PL-FALE can improve the classification performance by utilizing the information embedded in the feature space and label space in most cases. Also, the proposed algorithm achieves more stable performance than other ECOC-based PLL algorithms. The source code of PL-FALE is publicly available for non-commercial and research use at: https://github.com/xiaoziyang1/plfale1.