AI Paper Atlas

Research map / Continual learning / Class-incremental

Split-and-Bridge: Adaptable Class Incremental Learning within a Single Neural Network

Jong-Yeong Kim, Dong-Wan Choi

AAAI 2021 · Continual learning › Class-incremental · approach: Class-balanced knowledge retention · 24 citations (OpenAlex)

Read the paper

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 Split-and-Bridge instead of the strongest listed baselineResNet-18 · CIFAR-100, 5 tasksaverage accuracy: 67.68 -> 68.62Improved
Use Split-and-Bridge instead of the strongest listed baselineResNet-18 · CIFAR-100, 10 tasksaverage accuracy: 63.72 -> 66.97Improved
Use Split-and-Bridge instead of the strongest listed baselineResNet-18 · CIFAR-100, 20 tasksaverage accuracy: 58.08 -> 61.12Improved
Use Split-and-Bridge instead of the strongest listed baselineResNet-18 · Tiny-ImageNet, 2 tasksaverage accuracy: 58.33 -> 60.52Improved
Use Split-and-Bridge instead of the strongest listed baselineResNet-18 · Tiny-ImageNet, 5 tasksaverage accuracy: 56.80 -> 57.16Improved
Use Split-and-Bridge instead of the strongest listed baselineResNet-18 · Tiny-ImageNet, 10 tasksaverage accuracy: 53.12 -> 54.81Improved
Use Split-and-Bridge instead of the strongest listed baselineResNet-18 · Tiny-ImageNet, 20 tasksaverage accuracy: 48.14 -> 51.63Improved
Add KD loss to CE loss (ablation)ResNet-18 · CIFAR-100, 2 tasksoverall accuracy: 74.25 -> 75.67Improved
Add KD loss to CE loss (ablation)ResNet-18 · CIFAR-100, 2 tasksnew accuracy: 89.6 -> 85.75Hurt
Add KD loss to CE loss (ablation)ResNet-18 · CIFAR-100, 2 tasksintra-new accuracy: 90.85 -> 88.30Hurt
Separate old-task and new-task upper-layer branches during split-phase trainingSplit-and-Bridge · CIFAR-100 and Tiny-ImageNetintra-new accuracyImproved
Use two separately trained networks and combine them with double distillationDD with WA · CIFAR-100 and Tiny-ImageNetaverage accuracyMixed

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