Rethinking Consistent Multi-Label Classification under Inexact Supervision
This work addresses the challenge of reducing annotation costs in multi-label classification for applications where precise labeling is impractical, though it appears incremental as it builds on existing weakly supervised paradigms.
The paper tackles the problem of multi-label classification under inexact supervision, such as partial or complementary labels, by proposing consistent approaches that avoid reliance on difficult-to-satisfy conditions like accurate label generation estimation or uniform distributions. The result includes theoretical consistency proofs and empirical validation showing effectiveness against state-of-the-art methods.
Partial multi-label learning and complementary multi-label learning are two popular weakly supervised multi-label classification paradigms that aim to alleviate the high annotation costs of collecting precisely annotated multi-label data. In partial multi-label learning, each instance is annotated with a candidate label set, among which only some labels are relevant; in complementary multi-label learning, each instance is annotated with complementary labels indicating the classes to which the instance does not belong. Existing consistent approaches for the two paradigms either require accurate estimation of the generation process of candidate or complementary labels or assume a uniform distribution to eliminate the estimation problem. However, both conditions are usually difficult to satisfy in real-world scenarios. In this paper, we propose consistent approaches that do not rely on the aforementioned conditions to handle both problems in a unified way. Specifically, we propose two unbiased risk estimators based on first- and second-order strategies. Theoretically, we prove consistency w.r.t. two widely used multi-label classification evaluation metrics and derive convergence rates for the estimation errors of the proposed risk estimators. Empirically, extensive experimental results validate the effectiveness of our proposed approaches against state-of-the-art methods.