Research map / Continual learning / Online and replay-based
Overcoming Forgetting Catastrophe in Quantization-Aware Training
Ting-An Chen, De-Nian Yang, Ming-Syan Chen
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 BaLL loss to LifeQuant (ablation) | LifeQuant 4-bit ResNet-20 · CIFAR-100 (gamma=25) | accuracy (%): 47.73 -> 50.15 | Improved |
| add ProxQ to LifeQuant (ablation) | LifeQuant 4-bit ResNet-20 · CIFAR-100 (gamma=25) | accuracy (%): 46.17 -> 50.15 | Improved |
| add ProxQ to LifeQuant (ablation) | LifeQuant 3-bit MobileNet-V2 · ImageCLEF I->P->C | accuracy (%): 71.94 -> 76.38 | Improved |
| add BaLL loss to LifeQuant (ablation) | LifeQuant 3-bit MobileNet-V2 · ImageCLEF I->P->C | accuracy (%): 43.61 -> 76.38 | Improved |
| add proximal regularization (ProxQ) to reduce space shift (ablation) | LifeQuant 2-bit ResNet-20 · CIFAR-100 (gamma=25) | space shift RMSE: 0.0034 -> 0.0032 | Improved |
| add class-based rebalancing factor to BaLL loss (ablation) | BaLL 3-bit MobileNet-V2, delta=20 · ImageCLEF B->C->I | accuracy (%): 47.22 -> 72.78 | Improved |
| increase replay ratio from 10% to 20% (ablation) | BaLL 3-bit MobileNet-V2 · ImageCLEF B->C->I | accuracy (%): 50.55 -> 72.78 | Improved |
| increase replay ratio from 20% to 35% (ablation) | BaLL 3-bit MobileNet-V2 · ImageCLEF B->C->I | accuracy (%): 72.78 -> 70.83 | Mixed |
| replace prior lifelong learning losses (EWC etc.) with BaLL | quantized MobileNet-V2 · ImageCLEF B->C->I (3-bit, delta=20) | forgetting (%): 38.71 -> 2.64 | Improved |
| replace AlignQ with LifeQuant | 3-bit MobileNet-V2 · ImageCLEF B->C->I | accuracy (%): 57.50 -> 76.11 | Improved |
Closest papers by meaning
- Continual Learning via Bit-Level Information Preserving (CVPR 2021 · similarity 0.87)
- Optimal and Approximate Adaptive Stochastic Quantization (NeurIPS 2024 · similarity 0.85)
- Layerwise Proximal Replay: A Proximal Point Method for Online Continual Learning (ICML 2024 · similarity 0.85)
- Measuring Representational Shifts in Continual Learning: A Linear Transformation Perspective (ICML 2025 · similarity 0.84)
- Mitigating Catastrophic Forgetting in Online Continual Learning by Modeling Previous Task Interrelations via Pareto Optimization (ICML 2024 · similarity 0.84)
- Online Task-Free Continual Learning via Dynamic Expansionable Memory Distribution (CVPR 2025 · similarity 0.84)
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.
