Neural Collapse in Multi-label Learning with Pick-all-label Loss
This work addresses the challenge of improving multi-label learning for applications like image tagging or text categorization, though it is incremental as it builds on prior multi-class neural collapse studies.
The paper tackles the problem of understanding deep neural networks in multi-label classification by extending the neural collapse phenomenon from multi-class to multi-label settings, proving that a generalized neural collapse holds with the pick-all-label loss and showing it leads to better test performance and more efficient training.
We study deep neural networks for the multi-label classification (MLab) task through the lens of neural collapse (NC). Previous works have been restricted to the multi-class classification setting and discovered a prevalent NC phenomenon comprising of the following properties for the last-layer features: (i) the variability of features within every class collapses to zero, (ii) the set of feature means form an equi-angular tight frame (ETF), and (iii) the last layer classifiers collapse to the feature mean upon some scaling. We generalize the study to multi-label learning, and prove for the first time that a generalized NC phenomenon holds with the "pick-all-label" formulation, which we term as MLab NC. While the ETF geometry remains consistent for features with a single label, multi-label scenarios introduce a unique combinatorial aspect we term the "tag-wise average" property, where the means of features with multiple labels are the scaled averages of means for single-label instances. Theoretically, under proper assumptions on the features, we establish that the only global optimizer of the pick-all-label cross-entropy loss satisfy the multi-label NC. In practice, we demonstrate that our findings can lead to better test performance with more efficient training techniques for MLab learning.