Reframing Long-Tailed Learning via Loss Landscape Geometry
About
Balancing performance trade-off on long-tail (LT) data distributions remains a long-standing challenge. In this paper, we posit that this dilemma stems from a phenomenon called "tail performance degradation" (the model tends to severely overfit on head classes while quickly forgetting tail classes) and pose a solution from a loss landscape perspective. We observe that different classes possess divergent convergence points in the loss landscape. Besides, this divergence is aggravated when the model settles into sharp and non-robust minima, rather than a shared and flat solution that is beneficial for all classes. In light of this, we propose a continual learning inspired framework to prevent "tail performance degradation". To avoid inefficient per-class parameter preservation, a Grouped Knowledge Preservation module is proposed to memorize group-specific convergence parameters, promoting convergence towards a shared solution. Concurrently, our framework integrates a Grouped Sharpness Aware module to seek flatter minima by explicitly addressing the geometry of the loss landscape. Notably, our framework requires neither external training samples nor pre-trained models, facilitating the broad applicability. Extensive experiments on four benchmarks demonstrate significant performance gains over state-of-the-art methods. The code is available at:https://gkp-gsa.github.io/.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Image Classification | ImageNet LT | Top-1 Accuracy58.9 | 264 | |
| Image Classification | iNaturalist 2018 (test) | Top-1 Accuracy74.4 | 207 | |
| Image Classification | ImageNet-LT (test) | -- | 159 | |
| Image Classification | CIFAR-10-LT (IF 50) | Top-1 Accuracy88.2 | 48 | |
| Image Classification | CIFAR-100 LT (IF=50) | Top-1 Acc57.6 | 42 | |
| Image Classification | CIFAR100 LT (r=100) | Top-1 Accuracy53.2 | 22 | |
| Image Classification | CIFAR-10-LT | Top-1 Accuracy86.3 | 17 | |
| Image Classification | CIFAR100 LT (r=10) | Top-1 Accuracy68.7 | 16 | |
| Image Classification | CIFAR10-LT r=10 | Top-1 Accuracy92.5 | 14 |