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

Training machine learning models robust to distribution shifts is criticalfor real-world applications. Some robust training algorithms (e.g., Group DRO)specialize to group shifts and require group information on all trainingpoints. Other methods (e.g., CVaR DRO) that do not need group annotations canbe overly conservative, since they naively upweight high loss points which mayform a contrived set that does not correspond to any meaningful group in thereal world (e.g., when the high loss points are randomly mislabeled trainingpoints). In this work, we address limitations in prior approaches by assuming amore nuanced form of group shift: conditioned on the label, we assume that thetrue group function (indicator over group) is simple. For example, we mayexpect that group shifts occur along low bitrate features (e.g., imagebackground, lighting). Thus, we aim to learn a model that maintains highaccuracy on simple group functions realized by these low bitrate features, thatneed not spend valuable model capacity achieving high accuracy on contrivedgroups of examples. Based on this, we consider the two-player game formulationof DRO where the adversary's capacity is bitrate-constrained. Our resultingpractical algorithm, Bitrate-Constrained DRO (BR-DRO), does not require groupinformation on training samples yet matches the performance of Group DRO ondatasets that have training group annotations and that of CVaR DRO onlong-tailed distributions. Our theoretical analysis reveals that in somesettings BR-DRO objective can provably yield statistically efficient and lessconservative solutions than unconstrained CVaR DRO.

Bitrate-Constrained DRO: Beyond Worst Case Robustness To Unknown Group Shifts