A Survey of Deep Long-Tail Classification Advancements

TL;DR

This survey reviews recent advancements in deep long-tail classification, highlighting methods addressing class imbalance, analyzing their performance, and discussing future challenges in real-world applications.

Contribution

It provides a comprehensive taxonomy of recent methods within a unified framework, along with performance comparisons and analysis of remaining challenges.

Findings

Recent methods improve accuracy on imbalanced datasets

Standard metrics and convergence studies are discussed

Future research directions are identified

Abstract

Many data distributions in the real world are hardly uniform. Instead, skewed and long-tailed distributions of various kinds are commonly observed. This poses an interesting problem for machine learning, where most algorithms assume or work well with uniformly distributed data. The problem is further exacerbated by current state-of-the-art deep learning models requiring large volumes of training data. As such, learning from imbalanced data remains a challenging research problem and a problem that must be solved as we move towards more real-world applications of deep learning. In the context of class imbalance, state-of-the-art (SOTA) accuracies on standard benchmark datasets for classification typically fall less than 75%, even for less challenging datasets such as CIFAR100. Nonetheless, there has been progress in this niche area of deep learning. To this end, in this survey, we provide a taxonomy of various methods proposed for addressing the problem of long-tail classification, focusing on works that happened in the last few years under a single mathematical framework. We also discuss standard performance metrics, convergence studies, feature distribution and classifier analysis. We also provide a quantitative comparison of the performance of different SOTA methods and conclude the survey by discussing the remaining challenges and future research direction.