Research map / Probabilistic and scientific modeling / Variational and generative sampling
Rethinking Controllable Variational Autoencoders
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 |
|---|---|---|---|
| Use DynamicVAE's incremental PI controller, moving average, and hybrid annealing | VAE · dSprites | RMIG: 0.4295 -> 0.4781 | Helped |
| Use DynamicVAE's controlled KL schedule | VAE · dSprites | MIG: 0.5233 -> 0.5578 | Helped |
| Use DynamicVAE's controlled KL schedule | VAE · dSprites | BetaVAE Score: 0.9608 -> 0.9981 | Helped |
| Replace positional PI control without large β(0) initialization with incremental PI control initialized to large β(0) (ablation) | DynamicVAE · dSprites | RMIG: 0.4641 ± 0.0240 -> 0.4781 ± 0.0172 | Helped |
| Add a ramp function to step-function annealing (ablation) | DynamicVAE · dSprites | RMIG: 0.4555 ± 0.0355 -> 0.4781 ± 0.0172 | Helped |
| Use moving-average feedback instead of instantaneous KL-divergence (ablation) | DynamicVAE · dSprites | RMIG: 0.4570 ± 0.0182 -> 0.4781 ± 0.0172 | Helped |
Closest papers by meaning
- ControlVAE: Controllable Variational Autoencoder (ICML 2020 · similarity 0.94)
- Guided Variational Autoencoder for Disentanglement Learning (CVPR 2020 · similarity 0.90)
- Property Controllable Variational Autoencoder via Invertible Mutual Dependence (ICLR 2021 · similarity 0.89)
- Why do Variational Autoencoders Really Promote Disentanglement? (ICML 2024 · similarity 0.89)
- $\alpha$TC-VAE: On the relationship between Disentanglement and Diversity (ICLR 2024 · similarity 0.88)
- Beyond Vanilla Variational Autoencoders: Detecting Posterior Collapse in Conditional and Hierarchical Variational Autoencoders (ICLR 2024 · similarity 0.88)
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.