On the robustness of average losses for partial-label learning

• Main Authors: Lv, Jiaqi, Liu, Biao, Feng, Lei, Xu, Ning, Xu, Miao, An, Bo, Niu, Gang, Geng, Xin, Sugiyama, Masashi
• Other Authors: School of Computer Science and Engineering
• Format: Article
• Language: English
• Published: 2023
• Subjects: Engineering::Computer science and engineering; Partial-Label Learning; Weakly Supervised Learning
• Online Access: https://hdl.handle.net/10356/172190

Partial-label learning (PLL) utilizes instances with PLs, where a PL includes several candidate labels but only one is the true label (TL). In PLL, identification-based strategy (IBS) purifies each PL on the fly to select the (most likely) TL for training; average-based strategy (ABS) treats all can...