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

Supervised learning datasets may contain multiple cues that explain thetraining set equally well, i.e., learning any of them would lead to the correctpredictions on the training data. However, many of them can be spurious, i.e.,lose their predictive power under a distribution shift and consequently fail togeneralize to out-of-distribution (OOD) data. Recently developed"diversification" methods (Lee et al., 2023; Pagliardini et al., 2023) approachthis problem by finding multiple diverse hypotheses that rely on differentfeatures. This paper aims to study this class of methods and identify the keycomponents contributing to their OOD generalization abilities.  We show that (1) diversification methods are highly sensitive to thedistribution of the unlabeled data used for diversification and canunderperform significantly when away from a method-specific sweet spot. (2)Diversification alone is insufficient for OOD generalization. The choice of theused learning algorithm, e.g., the model's architecture and pretraining, iscrucial. In standard experiments (classification on Waterbirds and Office-Homedatasets), using the second-best choice leads to an up to 20\% absolute drop inaccuracy. (3) The optimal choice of learning algorithm depends on the unlabeleddata and vice versa i.e. they are co-dependent. (4) Finally, we show that, inpractice, the above pitfalls cannot be alleviated by increasing the number ofdiverse hypotheses, the major feature of diversification methods.  These findings provide a clearer understanding of the critical design factorsinfluencing the OOD generalization abilities of diversification methods. Theycan guide practitioners in how to use the existing methods best and guideresearchers in developing new, better ones.

Unraveling the Key Components of OOD Generalization via Diversification