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 this paper, we propose a novel data-pruning approach calledmoving-one-sample-out (MoSo), which aims to identify and remove the leastinformative samples from the training set. The core insight behind MoSo is todetermine the importance of each sample by assessing its impact on the optimalempirical risk. This is achieved by measuring the extent to which the empiricalrisk changes when a particular sample is excluded from the training set.Instead of using the computationally expensive leaving-one-out-retrainingprocedure, we propose an efficient first-order approximator that only requiresgradient information from different training stages. The key idea behind ourapproximation is that samples with gradients that are consistently aligned withthe average gradient of the training set are more informative and shouldreceive higher scores, which could be intuitively understood as follows: if thegradient from a specific sample is consistent with the average gradient vector,it implies that optimizing the network using the sample will yield a similareffect on all remaining samples. Experimental results demonstrate that MoSoeffectively mitigates severe performance degradation at high pruning ratios andachieves satisfactory performance across various settings.

Data Pruning via Moving-one-Sample-out