Research map / Probabilistic and scientific modeling / Variational and generative sampling
Learning Hierarchical Priors in VAEs
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 |
|---|---|---|---|
| Use REWO instead of GECO | VHP · Pendulum | absolute error: 0.53 -> 0.054 | Helped |
| Use the VAE objective with 20-epoch warm-up instead of VHP + REWO (ablation) | VHP · Pendulum | absolute error: 0.054 -> 0.20 | Hurt |
| Use the VAE objective with 200-epoch warm-up instead of VHP + REWO (ablation) | VHP · Pendulum | absolute error: 0.054 -> 0.31 | Hurt |
| Replace the REWO objective with the VAE objective (ablation) | VHP · Pendulum | absolute error: 0.054 -> 0.49 | Hurt |
| Replace the importance-weighted bound in the second stochastic layer with the ELBO (ablation) | VHP + REWO · Pendulum | latent representation quality | Hurt |
| Use the alternative beta-update scheme (ablation) | VHP + REWO · Pendulum | latent representation quality | Hurt |
| Use VHP + REWO instead of VampPrior | VHP · dynamic MNIST test | negative test log-likelihood: 80.42 -> 78.88 | Helped |
| Use VHP + REWO instead of VampPrior | VHP · static MNIST test | negative test log-likelihood: 84.02 -> 82.74 | Helped |
| Use VHP + REWO instead of VampPrior | VHP · Fashion-MNIST test | negative test log-likelihood: 232.78 -> 225.37 | Helped |
| Use VHP + REWO instead of VampPrior | VHP · OMNIGLOT test | negative test log-likelihood: 101.97 -> 101.78 | Helped |
| Use VHP + REWO instead of IWAE (L=1) | VHP · Fashion-MNIST test | negative test log-likelihood: 226.83 -> 225.37 | Helped |
| Use VHP + REWO instead of IWAE | VHP · OMNIGLOT test | negative test log-likelihood: 101.57 -> 101.78 | Hurt |
Closest papers by meaning
- Trading Information between Latents in Hierarchical Variational Autoencoders (ICLR 2023 · similarity 0.92)
- A Contrastive Learning Approach for Training Variational Autoencoder Priors (NeurIPS 2021 · similarity 0.91)
- Variational Autoencoders with Jointly Optimized Latent Dependency Structure (ICLR 2019 · similarity 0.90)
- Beyond Vanilla Variational Autoencoders: Detecting Posterior Collapse in Conditional and Hierarchical Variational Autoencoders (ICLR 2024 · similarity 0.89)
- Variational Learning of Fractional Posteriors (ICML 2025 · similarity 0.88)
- A Statistical Analysis of Wasserstein Autoencoders for Intrinsically Low-dimensional Data (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.