Research map / Continual learning / Online and replay-based
GAN Memory with No Forgetting
Read the paperCheck 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 |
|---|---|---|---|
| add parameter compression and sharing (ablation) | GAN memory · butterfly task D1 | FID: 28.89 -> 27.67 | Helped |
| add parameter compression and sharing (ablation) | GAN memory · butterfly task D2 | FID: 22.80 -> 23.49 | Hurt |
| add parameter compression and sharing (ablation) | GAN memory · butterfly task D3 | FID: 34.36 -> 28.90 | Helped |
| add parameter compression and sharing (ablation) | GAN memory · butterfly task D4 | FID: 35.72 -> 31.07 | Helped |
| add parameter compression and sharing (ablation) | GAN memory · butterfly task D5 | FID: 29.50 -> 32.19 | Hurt |
| add parameter compression and sharing (ablation) | GAN memory · butterfly task D6 | FID: 40.03 -> 49.28 | Hurt |
| add compression and sharing (ablation) | GAN memory · butterfly task D6 | new parameters: 10.6M -> 0.3M | Helped |
| modulate a different source GAN (LSUN Bedrooms instead of CelebA) (ablation) | GAN memory · Flowers | FID: 14.8 -> 15.0 | No clear effect |
| use style modulation of frozen source instead of fine-tuning the whole model | GAN memory · Flowers, Cathedrals, Cats | FID | Helped |
| use GAN memory instead of MeRGAN replay | GAN memory · lifelong generation, 6 tasks | FID | Helped |
| use conditional GAN memory pseudo rehearsal instead of EWC or MeRGAN | conditional GAN memory · class-incremental ImageNet 36 classes | accuracy | Helped |
Closest papers by meaning
- Memory Replay GANs: Learning to Generate New Categories without Forgetting (NeurIPS 2018 · similarity 0.90)
- Hyper-LifelongGAN: Scalable Lifelong Learning for Image Conditioned Generation (CVPR 2021 · similarity 0.89)
- Memorization Precedes Generation: Learning Unsupervised GANs with Memory Networks (ICLR 2018 · similarity 0.88)
- Recall-Oriented Continual Learning with Generative Adversarial Meta-Model (AAAI 2024 · similarity 0.87)
- Do Your Best and Get Enough Rest for Continual Learning (CVPR 2025 · similarity 0.86)
- Resurrecting Old Classes with New Data for Exemplar-Free Continual Learning (CVPR 2024 · similarity 0.85)
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.