Research map / Continual learning / Online and replay-based
Learning Dynamic Latent Spaces for Lifelong Generative Modelling
Fei Ye, Adrian G. Bors
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 |
|---|---|---|---|
| Use recursive latent-space expansion with EGAM attention | ORVAE · Split CIFAR10 | FID: 146.30 -> 122.73 | Improved |
| Use recursive latent-space expansion with EGAM attention | ORVAE · Split TinyImageNet | FID: 170.62 -> 131.46 | Improved |
| Use ORVAE for classification | ORVAE · Split CIFAR10 | classification accuracy: 48.92 ± 1.32 -> 50.43 ± 0.15 | Improved |
| Use ORVAE for classification | ORVAE · Split CIFAR100 | classification accuracy: 21.62 ± 0.69 -> 22.83 ± 0.25 | Improved |
| Use ORVAE for density estimation | ORVAE · Split MNIST-Fashion | sample log-likelihood: -242.54 -> -241.40 | Improved |
| Use importance-weighted objectives rather than ELBO (ablation) | ORVAE · Split MNIST | sample log-likelihood: -108.21 -> -120.12; -108.21 -> -133.45 | Hurt |
| Use importance-weighted objectives rather than ELBO (ablation) | ORVAE · Split MNIST-Fashion | sample log-likelihood: -241.40 -> -241.70; -241.40 -> -246.48 | Hurt |
| Use a larger memory buffer (ablation) | ORVAE · Split MNIST | sample log-likelihood | Mixed |
| Increase the expansion threshold λ (ablation) | ORVAE · Split MNIST | sample log-likelihood | Mixed |
Closest papers by meaning
- Continual Variational Autoencoder via Continual Generative Knowledge Distillation (AAAI 2023 · similarity 0.91)
- Wasserstein Expansible Variational Autoencoder for Discriminative and Generative Continual Learning (ICCV 2023 · similarity 0.90)
- Dynamic Expansion Diffusion Learning for Lifelong Generative Modelling (AAAI 2025 · similarity 0.87)
- Learn more, but bother less: parameter efficient continual learning (NeurIPS 2024 · similarity 0.86)
- Task-Free Continual Generation and Representation Learning via Dynamic Expansionable Memory Cluster (AAAI 2024 · similarity 0.86)
- Knowledge Graph Enhanced Generative Multi-modal Models for Class-Incremental Learning (NeurIPS 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.
