High-Likelihood Area Matters --- Rewarding Correct,Rare Predictions Under Imbalanced Distributions

guangxiang zhao, Lei Li, Xuancheng Ren, Xu Sun, Bin He

Keywords: classification, imbalance, long-tailed, likelihood, focal loss

Abstract: Learning from natural datasets poses significant challenges for traditional classification methods based on the cross-entropy objective due to imbalanced class distributions. It is intuitive to assume that the examples from rare classes are harder to learn so that the classifier is uncertain of the prediction, which establishes the low-likelihood area. Based on this, existing approaches drive the classifier actively to correctly predict those incorrect, rare examples. However, this assumption is one-sided and could be misleading. We find in practice that the high-likelihood area contains correct predictions for rare class examples and it plays a vital role in learning imbalanced class distributions. In light of this finding, we propose the Eureka Loss, which rewards the classifier when examples belong to rare classes in the high-likelihood area are correctly predicted. Experiments on the large-scale long-tailed iNaturalist 2018 classification dataset and the ImageNet-LT benchmark both validate the proposed approach. We further analyze the influence of the Eureka Loss in detail on diverse data distributions.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

One-sentence Summary: We chanllenge the intuition that high-likelihood area should be weaken for learning under imbalanced distributions and find that the correct predictions of rare classes paly an important role.

Reviewed Version (pdf): https://openreview.net/references/pdf?id=amM3npxm-I