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

In leading collaborative filtering (CF) models, representations of users anditems are prone to learn popularity bias in the training data as shortcuts. Thepopularity shortcut tricks are good for in-distribution (ID) performance butpoorly generalized to out-of-distribution (OOD) data, i.e., when popularitydistribution of test data shifts w.r.t. the training one. To close the gap,debiasing strategies try to assess the shortcut degrees and mitigate them fromthe representations. However, there exist two deficiencies: (1) when measuringthe shortcut degrees, most strategies only use statistical metrics on a singleaspect (i.e., item frequency on item and user frequency on user aspect),failing to accommodate the compositional degree of a user-item pair; (2) whenmitigating shortcuts, many strategies assume that the test distribution isknown in advance. This results in low-quality debiased representations. Worsestill, these strategies achieve OOD generalizability with a sacrifice on IDperformance. In this work, we present a simple yet effective debiasingstrategy, PopGo, which quantifies and reduces the interaction-wise popularityshortcut without any assumptions on the test data. It first learns a shortcutmodel, which yields a shortcut degree of a user-item pair based on theirpopularity representations. Then, it trains the CF model by adjusting thepredictions with the interaction-wise shortcut degrees. By taking both causal-and information-theoretical looks at PopGo, we can justify why it encouragesthe CF model to capture the critical popularity-agnostic features while leavingthe spurious popularity-relevant patterns out. We use PopGo to debias twohigh-performing CF models (MF, LightGCN) on four benchmark datasets. On both IDand OOD test sets, PopGo achieves significant gains over the state-of-the-artdebiasing strategies (e.g., DICE, MACR).

Robust Collaborative Filtering to Popularity Distribution Shift