Research map / Continual learning / Online and replay-based
Residual Continual Learning
Janghyeon Lee, Donggyu Joo, Hyeong Gwon Hong, Junmo Kim
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use ResCL's layer-wise residual reparameterization and decay loss instead of LwF | pre-activation ResNet-32 · CIFAR-10 → CIFAR-100 | maximum achievable average accuracy (%): 77.67±0.10 -> 78.80±0.17 | Improved |
| Use ResCL's layer-wise residual reparameterization and decay loss instead of Mean-IMM | pre-activation ResNet-32 · CIFAR-100 → CIFAR-10 | maximum achievable average accuracy (%): 76.41±0.19 -> 77.13±0.12 | Improved |
| Use ResCL's layer-wise residual reparameterization and decay loss instead of Mean-IMM | pre-activation ResNet-32 · CIFAR-10 → SVHN | maximum achievable average accuracy (%): 79.91±1.47 -> 89.49±0.32 | Improved |
| Use ResCL instead of LwF | pre-activation ResNet · CIFAR-10 → CIFAR-100 → SVHN | maximum achievable average accuracy (%): 58.44 -> 78.53 | Improved |
| Use ResCL instead of Mean-IMM | AlexNet · ImageNet → CUB | maximum achievable average accuracy (%): 52.12 -> 53.51 | Improved |
| Use ResCL instead of LwF | VGG · ImageNet → CUB | maximum achievable average accuracy (%): 68.82 -> 68.95 | Improved |
| Use ResCL instead of LwF at the required target accuracy | pre-activation ResNet-32 · CIFAR-10 → CIFAR-100 | source accuracy (%): 89.59 -> 90.65 | Improved |
| Use ResCL instead of LwF at the required target accuracy | pre-activation ResNet-32 · CIFAR-100 → CIFAR-10 | source accuracy (%): 63.91 -> 68.13 | Improved |
| Use ResCL instead of LwF at the required target accuracy | pre-activation ResNet-32 · CIFAR-10 → SVHN | source accuracy (%): 38.08 -> 76.83 | Improved |
| Increase the decay-loss multiplier λ | ResCL on pre-activation ResNet-32 · CIFAR-10 → SVHN | source accuracy | Improved |
| Increase the decay-loss multiplier λ | ResCL on pre-activation ResNet-32 · CIFAR-10 → SVHN | target accuracy | Hurt |
Closest papers by meaning
- Continual learning with hypernetworks (ICLR 2020 · similarity 0.89)
- Theory on Forgetting and Generalization of Continual Learning (ICML 2023 · similarity 0.89)
- A continual learning survey: Defying forgetting in classification tasks (TPAMI 2021 · similarity 0.89)
- Advancing Prompt-Based Methods for Replay-Independent General Continual Learning (ICLR 2025 · similarity 0.89)
- CRNet: A Fast Continual Learning Framework With Random Theory (TPAMI 2023 · similarity 0.88)
- Towards Understanding Catastrophic Forgetting in Two-layer Convolutional Neural Networks (ICML 2025 · similarity 0.88)
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.
