On the Robustness of Average Losses for Partial-Label Learning
This work addresses partial-label learning, a domain-specific problem in machine learning, by providing theoretical robustness analysis and experimental gains, though it is incremental as it builds on existing strategies.
The paper tackles the problem of partial-label learning (PLL) by analyzing why average-based strategies (ABS) underperform and proposing improvements, showing that ABS with bounded losses can match or exceed state-of-the-art performance of identification-based strategies (IBS) and that using robust average PL losses to warm start IBS further boosts performance.
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 candidate labels equally for training and let trained models be able to predict TL. Although PLL research has focused on IBS for better performance, ABS is also worthy of study since modern IBS behaves like ABS in the beginning of training to prepare for PL purification and TL selection. In this paper, we analyze why ABS was unsatisfactory and propose how to improve it. Theoretically, we formalize five problem settings of PLL and prove that average PL losses (APLLs) with bounded multi-class losses are always robust, while APLLs with unbounded losses may be non-robust, which is the first robustness analysis for PLL. Experimentally, we have two promising findings: ABS using bounded losses can match/exceed state-of-the-art performance of IBS using unbounded losses; after using robust APLLs to warm start, IBS can further improve upon itself. Our work draws attention to ABS research, which can in turn boost IBS and push forward the whole PLL.