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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Class imbalance is a ubiquitous phenomenon occurring in real world datadistributions. To overcome its detrimental effect on training accurateclassifiers, existing work follows three major directions: class re-balancing,information transfer, and representation learning. In this paper, we propose anew and complementary direction for improving performance on long taileddatasets - optimizing the backbone architecture through neural architecturesearch (NAS). We find that an architecture's accuracy obtained on a balanceddataset is not indicative of good performance on imbalanced ones. This posesthe need for a full NAS run on long tailed datasets which can quickly becomeprohibitively compute intensive. To alleviate this compute burden, we aim toefficiently adapt a NAS super-network from a balanced source dataset to animbalanced target one. Among several adaptation strategies, we find that themost effective one is to retrain the linear classification head with reweightedloss, while freezing the backbone NAS super-network trained on a balancedsource dataset. We perform extensive experiments on multiple datasets andprovide concrete insights to optimize architectures for long tailed datasets.

IMB-NAS: Neural Architecture Search for Imbalanced Datasets