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

It is well-known that training neural networks for image classification withempirical risk minimization (ERM) makes them vulnerable to relying on spuriousattributes instead of causal ones for prediction. Previously, deep featurere-weighting (DFR) has proposed retraining the last layer of a pre-trainednetwork on balanced data concerning spurious attributes, making it robust tospurious correlation. However, spurious attribute annotations are not alwaysavailable. In order to provide group robustness without such annotations, wepropose a new method, called loss-based feature re-weighting (LFR), in which weinfer a grouping of the data by evaluating an ERM-pre-trained model on a smallleft-out split of the training data. Then, a balanced number of samples ischosen by selecting high-loss samples from misclassified data points andlow-loss samples from correctly-classified ones. Finally, we retrain the lastlayer on the selected balanced groups to make the model robust to spuriouscorrelation. For a complete assessment, we evaluate LFR on various versions ofWaterbirds and CelebA datasets with different spurious correlations, which is anovel technique for observing the model's performance in a wide range ofspuriosity rates. While LFR is extremely fast and straightforward, itoutperforms the previous methods that do not assume group label availability,as well as the DFR with group annotations provided, in cases of high spuriouscorrelation in the training data.

Annotation-Free Group Robustness via Loss-Based Resampling