English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

As the class size grows, maintaining a balanced dataset across many classesis challenging because the data are long-tailed in nature; it is evenimpossible when the sample-of-interest co-exists with each other in onecollectable unit, e.g., multiple visual instances in one image. Therefore,long-tailed classification is the key to deep learning at scale. However,existing methods are mainly based on re-weighting/re-sampling heuristics thatlack a fundamental theory. In this paper, we establish a causal inferenceframework, which not only unravels the whys of previous methods, but alsoderives a new principled solution. Specifically, our theory shows that the SGDmomentum is essentially a confounder in long-tailed classification. On onehand, it has a harmful causal effect that misleads the tail prediction biasedtowards the head. On the other hand, its induced mediation also benefits therepresentation learning and head prediction. Our framework elegantlydisentangles the paradoxical effects of the momentum, by pursuing the directcausal effect caused by an input sample. In particular, we use causalintervention in training, and counterfactual reasoning in inference, to removethe "bad" while keep the "good". We achieve new state-of-the-arts on threelong-tailed visual recognition benchmarks: Long-tailed CIFAR-10/-100,ImageNet-LT for image classification and LVIS for instance segmentation.

Long-Tailed Classification by Keeping the Good and Removing the Bad Momentum Causal Effect