AI Paper Atlas

Research map / Continual learning / Online and replay-based

Growing a Brain with Sparsity-Inducing Generation for Continual Learning

Hyundong Jin, Gyeong-hyeon Kim, Chanho Ahn, Eunwoo Kim

ICCV 2023 · Continual learning › Online and replay-based · approach: Gradient projection and plasticity preservation · 8 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
transform old parameters with hypernetwork-generated sparse parameters instead of keeping them fixed (ablation)GrowBrain · ImageNet+FG class-incrementalaccuracy: 45.3 -> 60.1Improved
transform old parameters with hypernetwork-generated sparse parameters instead of keeping them fixed (ablation)GrowBrain · UCF-101 class-incrementalaccuracy: 40.4 -> 64.7Improved
add sparsity-inducing generation (lambda_sp > 0) (ablation)GrowBrain · ImageNet+FG class-incrementalaccuracy: 58.1 -> 60.1Improved
add sparsity-inducing generation (lambda_sp > 0) (ablation)GrowBrain · UCF-101 class-incrementalaccuracy: 64.0 -> 64.7Improved
add channel-adaptive segments and matrix decomposition (scalability) (ablation)GrowBrain · ImageNet+FG class-incrementalnumber of layer embeddings: 3262 -> 133Improved
add channel-adaptive segments and matrix decomposition (scalability) (ablation)GrowBrain · ImageNet+FG class-incrementalaccuracy: 59.8 -> 60.1No clear effect
raise lambda_sp from 0.1 to 1.0 (ablation)GrowBrain · ImageNet+FG class-incrementalaccuracy and parameters: 15.7M -> 9.4M params, +0.7% accuracyImproved
use parameter isolation with transformation instead of regularization methods (LwF, EWC-On)GrowBrain · ImageNet+FG class-incrementalaccuracy: gaps of 48.8% and 31.7% vs EWC-On and LwFImproved
use GrowBrain instead of LwF on ActivityNet class-incremental T=5GrowBrain · ActivityNet class-incremental (T=5, Avg)accuracy: 42.8 vs 44.1 (LwF)Hurt
use all previous parameter sets (PackNet) on dissimilar domainsPackNet · ImageNet+FG class-incrementalaccuracy: drops of 12.2% and 9.0% after WikiArt and SketchHurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

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