AI Paper Atlas

Research map / Continual learning / Online and replay-based

Memory Aware Synapses: Learning what (not) to forget

Rahaf Aljundi, Francesca Babiloni, Mohamed Elhoseiny, Marcus Rohrbach, Tinne Tuytelaars

ECCV 2018 · Continual learning › Online and replay-based · approach: Gradient projection and plasticity preservation · 1,702 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
use local Hebbian layerwise importance (l-MAS) instead of global gradients (ablation)MAS · 2-task object recognitionaverage forgetting (%): 1% -> 3%Hurt
compute gradients of squared l2 norm instead of vector output (ablation)MAS · Flower -> Scenes and Birdsdrop on first task (%): 0.51% -> 0.50%No clear effect
compute importance on unlabeled test data instead of training data (ablation)MAS · Flower -> Scenesfirst-task accuracy (%): 77.24 -> 77.45No clear effect
compute importance on unlabeled test data instead of training data (ablation)l-MAS · Flower -> Birdsfirst-task accuracy (%): 73.96 -> 74.48No clear effect
compute importance on both training and test dataMAS · 6DS fact learningMAP on all facts: 0.29 -> 0.30Improved
compute importance only on the class of interest (adaptation)MAS · Krishna Kamal flower, 8 tasksforgetting on subset (%): 8% -> 2%Improved
compute importance only on the class of interest (adaptation)MAS · Krishna Kamal flower, 8 taskslast task accuracy (%): 69% -> 84%Improved
compute importance only on sport subsetMAS · 6DS sport subsetMAP at end of sequence: 0.20 -> 0.50Improved
compute importance on same subset used for evaluationMAS · 6DS two-task adaptationMAP on T11: 0.451 -> 0.472Improved
use MAS instead of SIMAS · 8-task object recognitionaverage accuracy (%): 50.49 -> 52.695Improved
use MAS instead of SIMAS · 6DS fact learning, 4 tasksMAP on all facts: 0.25 -> 0.29Improved
use MAS instead of SIMAS · Scenes -> Birds, first taskaccuracy (%): 57.46 -> 57.61No clear effect

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