Research map / Continual learning / Class-incremental
Masked Autoencoders are Efficient Class Incremental Learners
Jiang-Tian Zhai, Xialei Liu, Andrew D. Bagdanov, Ke Li, Ming-Ming Cheng
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 |
|---|---|---|---|
| add MAE-generated replay data (ablation) | Bilateral MAE baseline · CIFAR-100 10 tasks | Avg accuracy: 73.40 -> 75.88 | Improved |
| add self-supervised reconstruction loss (ablation) | MAE with replay · CIFAR-100 10 tasks | Avg accuracy: 75.88 -> 77.48 | Improved |
| add bilateral detailed branch (ablation) | MAE with replay and reconstruction · CIFAR-100 10 tasks | Avg accuracy: 77.48 -> 79.12 | Improved |
| masking ratio 0.75 versus 0.60 or 0.90 (ablation) | Bilateral MAE · CIFAR-100 10 tasks | Avg accuracy: 77.50 / 77.12 -> 79.12 | Improved |
| replace real-image replay with MAE-generated replay of equal storage (ablation) | Bilateral MAE · CIFAR-100 10 tasks | Avg accuracy: 79.57 -> 79.12 | No clear effect |
| apply detail loss in frequency domain instead of spatial domain (ablation) | Bilateral MAE · CIFAR-100 10 tasks | Last accuracy: 65.93 -> 68.40 | Improved |
| give DyTox the same 80 masked exemplars (ablation) | DyTox · CIFAR-100 10 tasks | Last accuracy: 62.10 -> 65.46 | Improved |
| reduce MLP size from 1536 to 768 (fewer parameters) (ablation) | Bilateral MAE · CIFAR-100 10 tasks | Last accuracy: 68.40 -> 67.52 | Hurt |
| use MAE patch replay instead of JPEG compression or down-scaling | memory-efficient replay · CIFAR-100 10 tasks | Last accuracy: 61.32 -> 68.40 | Improved |
Closest papers by meaning
- Contrastive Tuning: A Little Help to Make Masked Autoencoders Forget (AAAI 2024 · similarity 0.88)
- Self-Guided Masked Autoencoder (NeurIPS 2024 · similarity 0.88)
- Autoencoder-Based Hybrid Replay for Class-Incremental Learning (ICML 2025 · similarity 0.87)
- The Dynamic Duo of Collaborative Masking and Target for Advanced Masked Autoencoder Learning (AAAI 2025 · similarity 0.87)
- R-MAE: Regions Meet Masked Autoencoders (ICLR 2024 · similarity 0.87)
- Learning Mask Invariant Mutual Information for Masked Image Modeling (ICLR 2025 · similarity 0.87)
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.
