Research map / Continual learning / Online and replay-based
Continual Unsupervised Generative Modeling
Fei Ye, Adrian G. Borş
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 importance-weighted bound with 50 samples instead of ELBO in DEGM (ablation) | DEGM · Split MNIST | log-likelihood: -92.25 -> -88.04 | Improved |
| use 50 importance samples instead of 5 in single VAE with generative replay (ablation) | IWELBO-GR · Split MNIST | log-likelihood: -95.80 -> -93.57 | Improved |
| add a second stochastic layer to the replay VAE (ablation) | IWELBO-GR-50 · Split MNIST vs COFMI | log-likelihood: -93.57 -> -91.23 on Split MNIST; -172.10 -> -188.97 on COFMI | Mixed |
| replace single VAE with generative replay by DEGM dynamic expansion graph | IWELBO-GR-50 · COFMI | log-likelihood: -172.10 -> -167.14 | Improved |
| add adaptive weights (DEGAM) to DEGM (ablation) | DEGM · MSFIR | MSE: 26.77 -> 25.85 | Improved |
| add a new base node for every task (DEGM-2) instead of graph-connected specific nodes (ablation) | DEGM · MSFIR | MSE: 26.77 -> 23.84 | Improved |
| add a new base node for every task (DEGM-2) instead of graph-connected specific nodes (ablation) | DEGM · CCCOSCZC | MSE: 393.8 -> 381.3 | Improved |
| lower the expansion threshold tau (ablation) | DEGAM · MSFIR | MSE / number of base nodes | Improved |
| use DEGM instead of CN-DPM* expansion | CN-DPM* · CCCOSCZC | MSE: 477.2 -> 393.8 | Improved |
| use DEGAM instead of diffusion replay (DDGR) | DDGR · MSFIR | MSE: 49.68 -> 25.85 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DDGR — DDGR: Continual Learning with Deep Diffusion-based Generative Replay (ICML 2023) · better on 1 of 1 shared result
- CAM-GAN — CAM-GAN: Continual Adaptation Modules for Generative Adversarial Networks (NeurIPS 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Task-Free Continual Generation and Representation Learning via Dynamic Expansionable Memory Cluster (AAAI 2024 · similarity 0.89)
- Dynamic Expansion Diffusion Learning for Lifelong Generative Modelling (AAAI 2025 · similarity 0.88)
- Continual Unsupervised Generative Modelling via Online Optimal Transport (AAAI 2025 · similarity 0.88)
- Exploring Rationale Learning for Continual Graph Learning (AAAI 2025 · similarity 0.88)
- Knowledge Graph Enhanced Generative Multi-modal Models for Class-Incremental Learning (NeurIPS 2025 · similarity 0.88)
- Train with Perturbation, Infer after Merging: A Two-Stage Framework for Continual Learning (NeurIPS 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.
