AI Paper Atlas

Research map / Continual learning / Online and replay-based

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

Hasan Abed Al Kader Hammoud, Ameya Prabhu, Ser-Nam Lim, Philip H.S. Torr, Adel Bibi, Bernard Ghanem

ICCV 2023 · Continual learning › Online and replay-based · approach: Plug-and-play online training enhancements · 15 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add FIFO sampling to the full-model/FC-Only learner (vs uniform) (ablation)OverAdapt · CGLM online accuracyonline accuracy: 34 -> 78Improved
add FIFO sampling (vs uniform) to full-model training (ablation)ER full-model · CGLM information retention (BWT)accuracy: 42 -> 7Hurt
add FIFO sampling (vs uniform) to full-model training (ablation)ER full-model · CLOC information retention (BWT)accuracy: 18 -> 1Hurt
add FIFO sampling (vs uniform) to full-model training (ablation)ER full-model · CLOC online accuracyonline accuracy: 14 -> 24Improved
use FC-Only training instead of full-model training under FIFO (ablation)OverAdapt · CGLM online accuracyonline accuracy: 78 -> 82Improved
use FC-Only training instead of full-model training (ablation)OverAdapt · CGLM/CLOC online and near-future accuracyaccuracyMixed
replace online accuracy with near-future accuracy (shift S=256 CGLM, 16384 CLOC)FIFO-sampled OverAdapt variants · CGLM and CLOCadaptation accuracy: about 70% drop in CGLM and 20-40% on CLOCHurt
replace online accuracy with near-future accuracyuniform-sampling variants · CGLM and CLOCadaptation accuracyNo clear effect
add mixed sampling (half FIFO, half uniform) instead of uniform (ablation)ER (Replay Only) · CGLM near-future accuracynear-future accuracyNo clear effect
add mixed sampling instead of uniform (ablation)ER (Replay Only) · CLOC near-future accuracynear-future accuracyMixed
change weight decay (ablation)ER (Replay Only) · CGLMnear-future accuracy and BWTNo clear effect
change learning rate by an order of magnitude from 0.005 (ablation)ER (Replay Only) · CGLMnear-future accuracy and BWTHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy