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

Concept shift is a prevailing problem in natural tasks like medical imagesegmentation where samples usually come from different subpopulations withvariant correlations between features and labels. One common type of conceptshift in medical image segmentation is the "information imbalance" betweenlabel-sparse samples with few (if any) segmentation labels and label-densesamples with plentiful labeled pixels. Existing distributionally robustalgorithms have focused on adaptively truncating/down-weighting the "lessinformative" (i.e., label-sparse in our context) samples. To exploit datafeatures of label-sparse samples more efficiently, we propose an adaptivelyweighted online optimization algorithm -- AdaWAC -- to incorporate dataaugmentation consistency regularization in sample reweighting. Our methodintroduces a set of trainable weights to balance the supervised loss andunsupervised consistency regularization of each sample separately. At thesaddle point of the underlying objective, the weights assign label-densesamples to the supervised loss and label-sparse samples to the unsupervisedconsistency regularization. We provide a convergence guarantee by recasting theoptimization as online mirror descent on a saddle point problem. Our empiricalresults demonstrate that AdaWAC not only enhances the segmentation performanceand sample efficiency but also improves the robustness to concept shift onvarious medical image segmentation tasks with different UNet-style backbones.

Adaptively Weighted Data Augmentation Consistency Regularization for Robust Optimization under Concept Shift