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

Active learning promises to alleviate the massive data needs of supervisedmachine learning: it has successfully improved sample efficiency by an order ofmagnitude on traditional tasks like topic classification and objectrecognition. However, we uncover a striking contrast to this promise: across 5models and 4 datasets on the task of visual question answering, a wide varietyof active learning approaches fail to outperform random selection. Tounderstand this discrepancy, we profile 8 active learning methods on aper-example basis, and identify the problem as collective outliers -- groups ofexamples that active learning methods prefer to acquire but models fail tolearn (e.g., questions that ask about text in images or require externalknowledge). Through systematic ablation experiments and qualitativevisualizations, we verify that collective outliers are a general phenomenonresponsible for degrading pool-based active learning. Notably, we show thatactive learning sample efficiency increases significantly as the number ofcollective outliers in the active learning pool decreases. We conclude with adiscussion and prescriptive recommendations for mitigating the effects of theseoutliers in future work.

Mind Your Outliers! Investigating the Negative Impact of Outliers on Active Learning for Visual Question Answering