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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

We revisit the common practice of evaluating adaptation of Online ContinualLearning (OCL) algorithms through the metric of online accuracy, which measuresthe accuracy of the model on the immediate next few samples. However, we showthat this metric is unreliable, as even vacuous blind classifiers, which do notuse input images for prediction, can achieve unrealistically high onlineaccuracy by exploiting spurious label correlations in the data stream. Ourstudy reveals that existing OCL algorithms can also achieve high onlineaccuracy, but perform poorly in retaining useful information, suggesting thatthey unintentionally learn spurious label correlations. To address this issue,we propose a novel metric for measuring adaptation based on the accuracy on thenear-future samples, where spurious correlations are removed. We benchmarkexisting OCL approaches using our proposed metric on large-scale datasets undervarious computational budgets and find that better generalization can beachieved by retaining and reusing past seen information. We believe that ourproposed metric can aid in the development of truly adaptive OCL methods. Weprovide code to reproduce our results athttps://github.com/drimpossible/EvalOCL.

Rapid Adaptation in Online Continual Learning: Are We Evaluating It Right?