AI Paper Atlas

Research map / Continual learning / Online and replay-based

Does Continual Learning Meet Compositionality? New Benchmarks and An Evaluation Framework

Weiduo Liao, Ying Wei, Mingchen Jiang, Qingfu Zhang, Hisao Ishibuchi

NeurIPS 2023 · Continual learning › Online and replay-based · approach: Unified and unsupervised continual frameworks · 0 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 a modularity-based method (RPSnet) instead of Finetune in class-ILcontinual learner, ResNet-18 · COBJAcon: 6.37 -> 21.93Improved
use RPSnet (modularity-based) instead of ER for compositional generalizationcontinual learner, ResNet-18 · CGQAHn: 68.46 -> 59.94Hurt
use RPSnet instead of ER for compositional generalizationcontinual learner, ResNet-18 · COBJHn: 45.86 -> 70.47Improved
use MNTDP* (frozen modules) instead of ER* in task-ILcontinual learner, ResNet-18 · CGQAHn: 70.19 -> 47.39Hurt
use LwF instead of Finetune in class-ILcontinual learner, ResNet-18 · CGQAHn: 63.75 -> 71.03Improved
use joint multi-task training instead of sequential continual trainingResNet-18 · CGQAHn: 71.03 -> 80.35Improved
use single-head (class-IL) instead of multi-head (task-IL) classifierFinetune · CGQAHn: 69.66 -> 63.75Hurt
use single-head (class-IL) instead of multi-head (task-IL) classifierLwF · CGQAHn: 69.40 -> 71.03No clear effect
add experience replay (ER) to Finetunecontinual learner, ResNet-18 · CGQAAcon: 8.38 -> 19.78Improved
use regularization (GEM, EWC) instead of replaycontinual learner, ResNet-18 · CGQAHn: GEM 67.33, EWC 66.95 vs Finetune 63.75No clear effect
test on novel combinations of old concepts vs fresh conceptscontinual learners · CGQAS(sys) relative dropHurt
increase training samples per class in continual taskscontinual learners · CGQAAsys and AnonImproved

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