Research map / Continual learning / Online and replay-based
Hyper-LifelongGAN: Scalable Lifelong Learning for Image Conditioned Generation
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 |
|---|---|---|---|
| swap conditional and real images between distillation and task losses (ablation) | Hyper-LifelongGAN · Paired image-conditioned generation, 4 tasks | Acc: 91.27 -> 92.55 | Helped |
| swap conditional and real images between distillation and task losses (ablation) | Hyper-LifelongGAN · Paired image-conditioned generation, 4 tasks | FID: 118.98 -> 101.43 | Helped |
| use random noise as the distillation input instead of real images (ablation) | Hyper-LifelongGAN · Paired image-conditioned generation, 4 tasks | Acc: 92.55 -> 91.92 | Hurt |
| use random noise as the distillation input instead of real images (ablation) | Hyper-LifelongGAN · Paired image-conditioned generation, 4 tasks | FID: 101.43 -> 110.84 | Hurt |
| increase base filter size from K=3 to K=5 (ablation) | Hyper-LifelongGAN · Cityscapes initial task | Acc: 66.00 -> 76.60 | Helped |
| increase base filter size from K=3 to K=5 (ablation) | Hyper-LifelongGAN · Cityscapes initial task | FID: 72.96 -> 72.25 | Helped |
| increase base filter size from K=5 to K=7 (ablation) | Hyper-LifelongGAN · Cityscapes initial task | Acc: 76.60 -> 75.40 | Hurt |
| increase base filter size from K=5 to K=7 (ablation) | Hyper-LifelongGAN · Cityscapes initial task | FID: 72.25 -> 57.58 | Helped |
| add dynamic base filters generated by task-specific filter generators (ablation) | Hyper-LifelongGAN · Paired image-conditioned generation, 4 tasks | FID: 137.99 -> 114.73 | Helped |
| add separate task-specific base filters (ablation) | Hyper-LifelongGAN · Paired image-conditioned generation, 4 tasks | Acc: 89.60 -> 91.77 | Helped |
| add task-specific coefficients for combining base filters (ablation) | Hyper-LifelongGAN · Paired image-conditioned generation, 4 tasks | Acc: 91.77 -> 92.55 | Helped |
| add task-specific coefficients for combining base filters (ablation) | Hyper-LifelongGAN · Paired image-conditioned generation, 4 tasks | FID: 105.42 -> 101.43 | Helped |
Closest papers by meaning
- GAN Memory with No Forgetting (NeurIPS 2020 · similarity 0.89)
- Continual learning with hypernetworks (ICLR 2020 · similarity 0.89)
- Improved Schemes for Episodic Memory-based Lifelong Learning (NeurIPS 2020 · similarity 0.89)
- Recall-Oriented Continual Learning with Generative Adversarial Meta-Model (AAAI 2024 · similarity 0.87)
- Task-Free Continual Generation and Representation Learning via Dynamic Expansionable Memory Cluster (AAAI 2024 · similarity 0.87)
- ZeroFlow: Overcoming Catastrophic Forgetting is Easier than You Think (ICML 2025 · 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.