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

The generalization ability of machine learning models degrades significantlywhen the test distribution shifts away from the training distribution. Weinvestigate the problem of training models that are robust to shifts caused bychanges in the distribution of class-priors or group-priors. The presence ofskewed training priors can often lead to the models overfitting to spuriousfeatures. Unlike existing methods, which optimize for either the worst or theaverage performance over classes or groups, our work is motivated by the needfor finer control over the robustness properties of the model. We present anextremely lightweight post-hoc approach that performs scaling adjustments topredictions from a pre-trained model, with the goal of minimizing adistributionally robust loss around a chosen target distribution. Theseadjustments are computed by solving a constrained optimization problem on avalidation set and applied to the model during test time. Our constrainedoptimization objective is inspired by a natural notion of robustness tocontrolled distribution shifts. Our method comes with provable guarantees andempirically makes a strong case for distributional robust post-hoc classifiers.An empirical implementation is available athttps://github.com/weijiaheng/Drops.

Distributionally Robust Post-hoc Classifiers under Prior Shifts