Although deep neural networks achieve tremendous success on various classification tasks, the generalization ability drops sheer when training datasets exhibit long-tailed distributions. One of the reasons is that the learned representations (i.e. features) from the imbalanced datasets are less effective than those from balanced datasets. Specifically, the learned representation under class-balanced distribution will present the Neural Collapse (NC) phenomena. NC indicates the features from the same category are close to each other and from different categories are maximally distant, showing an optimal linear separable state of classification. However, the pattern differs on imbalanced datasets and is partially responsible for the reduced performance of the model. In this work, we propose two explicit feature regularization terms to learn high-quality representation for class-imbalanced data. With the proposed regularization, NC phenomena will appear under the class-imbalanced distribution, and the generalization ability can be significantly improved. Our method is easily implemented, highly effective, and can be plugged into most existing methods. The extensive experimental results on widely-used benchmarks show the effectiveness of our method

This survey by Yifan Zhang, Bingyi Kang, Bryan Hooi, Shuicheng Yan and Jiashi Feng covers the following topic in far grater detail and I highly recommend checking it out for a more thorough discussion the ideas discussed in this article. With the massive success of Deep Learning in the field of image recognition comes the need to apply these techniques to solve real-world problems. An issue that arises here, however, is that in real world applications, training samples typically a long-tailed class distribution, where a small portion of classes have massive sample points but the others are associated with only a few samples. Thus, a model can be easily biased towards the head classes, resulting in a poor performance on the tail classes [1]. Many methods to counter such class imbalances in the data have been studied, mostly grouped within 3 categories.