Balanced Product of Calibrated Experts for Long-Tailed Recognition

TL;DR

This paper introduces BalPoE, a balanced ensemble method for long-tailed recognition that combines calibrated experts to improve generalization across class distributions, achieving state-of-the-art results.

Contribution

The paper extends logit adjustment to ensembles, proposing BalPoE, which generalizes previous methods and provides a Fisher-consistent approach for unbiased long-tailed recognition.

Findings

Achieves state-of-the-art results on CIFAR-100-LT, ImageNet-LT, and iNaturalist-2018.

Demonstrates the importance of calibrated experts using mixup.

Provides theoretical proof of Fisher-consistency for the ensemble.

Abstract

Many real-world recognition problems are characterized by long-tailed label distributions. These distributions make representation learning highly challenging due to limited generalization over the tail classes. If the test distribution differs from the training distribution, e.g. uniform versus long-tailed, the problem of the distribution shift needs to be addressed. A recent line of work proposes learning multiple diverse experts to tackle this issue. Ensemble diversity is encouraged by various techniques, e.g. by specializing different experts in the head and the tail classes. In this work, we take an analytical approach and extend the notion of logit adjustment to ensembles to form a Balanced Product of Experts (BalPoE). BalPoE combines a family of experts with different test-time target distributions, generalizing several previous approaches. We show how to properly define these distributions and combine the experts in order to achieve unbiased predictions, by proving that the ensemble is Fisher-consistent for minimizing the balanced error. Our theoretical analysis shows that our balanced ensemble requires calibrated experts, which we achieve in practice using mixup. We conduct extensive experiments and our method obtains new state-of-the-art results on three long-tailed datasets: CIFAR-100-LT, ImageNet-LT, and iNaturalist-2018. Our code is available at https://github.com/emasa/BalPoE-CalibratedLT.