Research map / Probabilistic and scientific modeling / Variational and generative sampling
Energy-Inspired Models: Learning with Sampler-Induced Distributions
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 a HIS prior instead of a Gaussian prior | VAE · Static MNIST | IWAE log-likelihood lower bound: -89.20 ± 0.08 -> -86.00 ± 0.05 | Helped |
| Use a HIS prior instead of a Gaussian prior | VAE · Dynamic MNIST | IWAE log-likelihood lower bound: -84.82 ± 0.12 -> -82.43 ± 0.05 | Helped |
| Use SNIS instead of a Gaussian prior | ConvHVAE · Static MNIST | IWAE log-likelihood lower bound: -82.43 ± 0.07 -> -81.51 ± 0.06 | Helped |
| Use SNIS instead of a Gaussian prior | ConvHVAE · Dynamic MNIST | IWAE log-likelihood lower bound: -81.14 ± 0.04 -> -80.19 ± 0.07 | Helped |
| Use a ConvHVAE rather than a VAE as the SNIS proposal | SNIS · Static MNIST | IWAE log-likelihood lower bound: -87.65 ± 0.07 -> -81.65 ± 0.05 | Helped |
| Use a ConvHVAE rather than a VAE as the SNIS proposal | SNIS · Dynamic MNIST | IWAE log-likelihood lower bound: -83.43 ± 0.07 -> -79.91 ± 0.05 | Helped |
| Use a ConvHVAE rather than a VAE as the SNIS proposal | SNIS · Fashion MNIST (binarized) | IWAE log-likelihood lower bound: -227.63 ± 0.06 -> -225.35 ± 0.07 | Helped |
| Use HIS rather than SNIS with a weaker small VAE proposal | Energy-inspired model · MNIST (continuous) | IWAE log-likelihood lower bound: -1253.67 ± 0.29 -> -1186.06 ± 6.12 | Helped |
| Use HIS rather than SNIS with a weaker small VAE proposal | Energy-inspired model · Fashion MNIST (continuous) | IWAE log-likelihood lower bound: -2463.60 ± 0.31 -> -2419.83 ± 2.47 | Helped |
| Use HIS rather than SNIS with a weaker small VAE proposal | Energy-inspired model · CelebA | IWAE log-likelihood lower bound: -60115.99 ± 19.75 -> -59711.30 ± 53.08 | Helped |
| Increase the SNIS candidate count K (ablation) | SNIS · VAE prior experiments | IWAE log-likelihood lower bound | Helped |
| Increase the HIS trajectory length T (ablation) | HIS · VAE prior experiments | IWAE log-likelihood lower bound | Helped |
Closest papers by meaning
- Learning Energy-Based Model with Variational Auto-Encoder as Amortized Sampler (AAAI 2021 · similarity 0.91)
- Variational Inference and Model Selection with Generalized Evidence Bounds (ICML 2018 · similarity 0.89)
- Perturb-and-max-product: Sampling and learning in discrete energy-based models (NeurIPS 2021 · similarity 0.89)
- Joint Learning of Energy-based Models and their Partition Function (ICML 2025 · similarity 0.88)
- BoltzNCE: Learning likelihoods for Boltzmann Generation with Stochastic Interpolants and Noise Contrastive Estimation (NeurIPS 2025 · similarity 0.87)
- On scalable and efficient training of diffusion samplers (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.