Research map / Probabilistic and scientific modeling / Variational and generative sampling
Partition Functions from Rao-Blackwellized Tempered Sampling
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 Rao-Blackwellized tempered sampling (RTS) instead of AIS | partition function estimation for large RBMs · binarized MNIST RBMs | estimation accuracy | Helped |
| use RTS instead of AIS or RAISE | partition function estimation · RBMs with 500 and 100 hidden units | bias and RMSE | Helped |
| use RTS instead of competing estimators | partition function estimation · 10-dimensional Gaussian mixture | RMSE | No clear effect |
| use RTS rather than TI or TI-RB with discrete temperatures | partition function estimation · RBM | sensitivity to number of temperatures | Helped |
| use an informative base distribution instead of a uniform base distribution (ablation) | partition function estimators · 784-visible, 500-hidden MNIST RBM | estimation accuracy | Helped |
| use a uniform base distribution instead of an informative base distribution (ablation) | RTS · 784-visible, 500-hidden MNIST RBM | initialization speed | Hurt |
| use RTS rather than AIS with a uniform base distribution | partition function estimation · 784-visible, 500-hidden MNIST RBM | bias | Helped |
| increase the number of inverse temperatures after enough temperatures provide good mixing (ablation) | RTS · RBM | RMSE | No clear effect |
| use a continuous-beta formulation instead of discrete temperatures (ablation) | RTS · partition function estimation | estimation accuracy | Hurt |
| subtract an estimated bias from RTS estimates (ablation) | RTS · partition function estimation | estimation accuracy | No clear effect |
| add control variates (ablation) | RTS · partition function estimation | variance | No clear effect |
| estimate the temperature probabilities from the transition matrix rather than RTS conditional probabilities (ablation) | RTS · RBM with 10 hidden units | RMSE: about 10% | Helped |
Closest papers by meaning
- Fast training and sampling of Restricted Boltzmann Machines (ICLR 2025 · similarity 0.86)
- Scalable Discrete Diffusion Samplers: Combinatorial Optimization and Statistical Physics (ICLR 2025 · similarity 0.85)
- Restricted Spectral Gap Decomposition for Simulated Tempering Targeting Mixture Distributions (NeurIPS 2025 · similarity 0.85)
- Joint Learning of Energy-based Models and their Partition Function (ICML 2025 · similarity 0.85)
- Progressive Inference-Time Annealing of Diffusion Models for Sampling from Boltzmann Densities (NeurIPS 2025 · similarity 0.84)
- BoltzNCE: Learning likelihoods for Boltzmann Generation with Stochastic Interpolants and Noise Contrastive Estimation (NeurIPS 2025 · similarity 0.84)
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.