Research map / Continual learning / Online and replay-based
Continual Learning With Lifelong Vision Transformer
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 LVT instead of DER++ | continual learning on CIFAR100 · CIFAR100, 10 splits, memory 500, Class-IL | accuracy (%): 36.15±1.10 -> 43.51±1.06 | Helped |
| use LVT instead of ER | continual learning on CIFAR100 · CIFAR100, 10 splits, memory 500, Task-IL | accuracy (%): 74.97±0.41 -> 76.78±0.71 | Helped |
| use LVT instead of DER++ | continual learning on TinyImageNet · TinyImageNet, 10 splits, memory 500, Class-IL | accuracy (%): 19.33±1.41 -> 23.97±1.27 | Helped |
| use LVT instead of RM | continual learning on TinyImageNet · TinyImageNet, 10 splits, memory 500, Task-IL | accuracy (%): 52.08±0.84 -> 57.39±0.75 | Helped |
| use LVT instead of ERT | continual learning on ImageNet100 · ImageNet100, 10 splits, memory 500, Class-IL | accuracy (%): 20.42±1.13 -> 26.32±1.67 | Helped |
| use LVT instead of FDR | continual learning on ImageNet100 · ImageNet100, 10 splits, memory 500, Task-IL | accuracy (%): 42.60±0.64 -> 47.84±1.33 | Helped |
| add inter-task attention (ablation) | LVT · CIFAR100, 10 splits, memory 500, Class-IL | accuracy (%): 36.93 -> 43.51 | Helped |
| add inter-task attention (ablation) | LVT · TinyImageNet, 10 splits, memory 500, Class-IL | accuracy (%): 19.35 -> 23.97 | Helped |
| add the injection classifier (ablation) | LVT · CIFAR100, 10 splits, memory 500, Class-IL | accuracy (%): 39.76 -> 43.51 | Helped |
| add the accumulation classifier (ablation) | LVT · CIFAR100, 10 splits, memory 500, Class-IL | accuracy (%): 38.42 -> 43.51 | Helped |
| use confidence-aware memory update (ablation) | LVT · TinyImageNet, 10 splits, memory 500, Class-IL | accuracy (%): 21.16 -> 23.97 | Helped |
Closest papers by meaning
- Task-Free Dynamic Sparse Vision Transformer for Continual Learning (AAAI 2024 · similarity 0.90)
- Visual Prompt Tuning in Null Space for Continual Learning (NeurIPS 2024 · similarity 0.87)
- Convolutional Prompting meets Language Models for Continual Learning (CVPR 2024 · similarity 0.87)
- Evolving Parameterized Prompt Memory for Continual Learning (AAAI 2024 · similarity 0.87)
- Mixture of Experts Meets Prompt-Based Continual Learning (NeurIPS 2024 · similarity 0.87)
- Enhancing Visual Continual Learning with Language-Guided Supervision (CVPR 2024 · similarity 0.86)
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.