AI Paper Atlas

Research map / Continual learning / Class-incremental

Rectification-Based Knowledge Retention for Continual Learning

Pravendra Singh, Pratik Mazumder, Piyush Rai, Vinay P. Namboodiri

CVPR 2021 · Continual learning › Class-incremental · approach: Class-balanced knowledge retention · 45 citations (OpenAlex)

Check 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 RKR instead of CCLLCIFAR-100 task-incremental learning with LeNet · CIFAR-100accuracy: 5.87%Improved
Use RKR instead of CCLLImageNet-1K task-incremental learning · ImageNet-1Ktop-5 accuracy: 3.1%Improved
Use RKR instead of LZSLincremental generalized zero-shot learning · aPY, AWA1, CUB, and SUNH: 5.65%, 6.91%, 6.33%, and 2.53%Improved
Add weight rectifications (ablation)RKR · CIFAR-100, LeNetaccuracy: 62.3% -> 69.6%Improved
Add weight rectifications (ablation)RKR · CIFAR-100, ResNet-18accuracy: 79.2% -> 87.6%Improved
Add output scaling (ablation)RKR · CIFAR-100, LeNetaccuracy: 69.1% -> 69.6%Improved
Add output scaling (ablation)RKR · CIFAR-100, ResNet-18accuracy: 87.5% -> 87.6%Improved
Add weight rectifications (ablation)RKR · AWA1, CUB, and SUNH: 6.8%, 9.28%, and 6.24%Improved
Add output scaling (ablation)RKR · AWA1, CUB, and SUNH: 3.71%, 4.57%, and 3.59%Improved
Initialize rectification and scaling parameters from the previous task instead of from scratch (ablation)RKR · CIFAR-100, ResNet-18accuracy: 1.45%Improved
Initialize rectification and scaling parameters from the previous task instead of randomly (ablation)RKR · AWA1, CUB, and SUNH: 3.72%, 7.57%, and 7.14%Improved
Increase K from 1 to 2 (ablation)RKR · CIFAR-100, ResNet-18accuracy: 85.55% -> 87.60%Improved

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