AI Paper Atlas

Research map / Continual learning / Class-incremental

MAGR: Manifold-Aligned Graph Regularization for Continual Action Quality Assessment

Kanglei Zhou, Liyuan Wang, Xingxing Zhang, Hubert P. H. Shum, Frederick W. B. Li, Jianguo Li, Xiaohui Liang

ECCV 2024 · Continual learning › Class-incremental · approach: Class-balanced knowledge retention · 22 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 manifold projector to align old features to current manifold (ablation)MAGR · MTL-AQArho_avg: 0.6949 -> 0.8979Improved
add residual link in the manifold projector (ablation)MAGR · MTL-AQArho_avg: 0.8391 -> 0.8979Improved
add intra-inter-joint graph regularizer (ablation)MAGR · MTL-AQArho_avg: 0.7362 -> 0.8979Improved
add joint (global) graph regularizer on top of intra-inter (ablation)MAGR · MTL-AQArho_avg: 0.7839 -> 0.8979Improved
use KL divergence instead of MSE in graph regularization (ablation)MAGR · MTL-AQArho_avg: 0.8447 -> 0.8979Improved
use ordered uniform sampling instead of random sampling for the memory bank (ablation)MAGR · MTL-AQArho_aft: 0.0876 -> 0.0223Improved
use feature replay with updated backbone instead of raw data replay (Feature MER vs MER)MER · MTL-AQArho_avg: 0.8720 -> 0.7283Hurt
reduce memory to 3 features per sessionMAGR · MTL-AQArho_avgHurt
increase feature replay memory from 5 to 11 samples per sessionMAGR · MTL-AQArho_avgNo clear effect
use MAGR vs strongest baseline on larger feature deviation datasetsMAGR · FineDiving, MTL-AQA, UNLV-Diveoverall correlation gain (%): 5.66, 6.56, 15.64 at MSE deviations 26.85, 35.28, 51.75Improved
sequential fine-tuning without memorybaseline · MTL-AQArho_avg: 0.9360 -> 0.5458Hurt

Later papers that compared with this method

1 later accepted paper reports MAGR as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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