Breadcrumbs: Adversarial Class-Balanced Sampling for Long-tailed Recognition
This work addresses the challenge of long-tailed recognition in machine learning, where class imbalance leads to poor performance on minority classes, offering an incremental improvement over prior methods.
The paper tackled the problem of overfitting in class-balanced sampling for long-tailed recognition by proposing Breadcrumb, an adversarial class-balanced sampling method that uses feature augmentation without extra computation, resulting in classifiers that outperform existing solutions on three datasets.
The problem of long-tailed recognition, where the number of examples per class is highly unbalanced, is considered. While training with class-balanced sampling has been shown effective for this problem, it is known to over-fit to few-shot classes. It is hypothesized that this is due to the repeated sampling of examples and can be addressed by feature space augmentation. A new feature augmentation strategy, EMANATE, based on back-tracking of features across epochs during training, is proposed. It is shown that, unlike class-balanced sampling, this is an adversarial augmentation strategy. A new sampling procedure, Breadcrumb, is then introduced to implement adversarial class-balanced sampling without extra computation. Experiments on three popular long-tailed recognition datasets show that Breadcrumb training produces classifiers that outperform existing solutions to the problem.