Research map / Probabilistic and scientific modeling / Variational and generative sampling
Surrogate Likelihoods for Variational Annealed Importance 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 |
|---|---|---|---|
| Increase RAND surrogate size from 64 to 1024 points (ablation) | SL-DAIS · Higgs | ELBO improvement (nats): 211.1 -> 222.4 | Helped |
| Increase RAND surrogate size from 64 to 1024 points (ablation) | SL-DAIS · SUSY | ELBO improvement (nats): 627.5 -> 637.2 | Helped |
| Increase CS-INIT surrogate size from 64 to 1024 points (ablation) | SL-DAIS · Higgs | ELBO improvement (nats): 209.6 -> 218.5 | Helped |
| Increase CS-INIT surrogate size from 64 to 1024 points (ablation) | SL-DAIS · SUSY | ELBO improvement (nats): 625.9 -> 633.7 | Helped |
| Increase CS-FIX surrogate size from 64 to 1024 points (ablation) | SL-DAIS · Higgs | ELBO improvement (nats): 83.0 -> 192.7 | Helped |
| Increase CS-FIX surrogate size from 64 to 1024 points (ablation) | SL-DAIS · SUSY | ELBO improvement (nats): 501.0 -> 555.5 | Helped |
| Replace the likelihood-based RAND surrogate with a neural-network surrogate (ablation) | SL-DAIS · Higgs | ELBO improvement (nats): 222.4 vs 16.8 | Hurt |
| Replace the likelihood-based RAND surrogate with a neural-network surrogate (ablation) | SL-DAIS · SUSY | ELBO improvement (nats): 637.2 vs 2.6 | Hurt |
| Use fixed coreset weights instead of learning surrogate weights (ablation) | SL-DAIS with coreset-selected points · Higgs | ELBO improvement (nats): 209.6 vs 83.0; 218.5 vs 192.7 | Hurt |
| Use fixed coreset weights instead of learning surrogate weights (ablation) | SL-DAIS with coreset-selected points · SUSY | ELBO improvement (nats): 625.9 vs 501.0; 633.7 vs 555.5 | Hurt |
Closest papers by meaning
- MCMC Variational Inference via Uncorrected Hamiltonian Annealing (NeurIPS 2021 · similarity 0.92)
- Variational Bayesian Monte Carlo (NeurIPS 2018 · similarity 0.91)
- Variational Inference with Mixtures of Isotropic Gaussians (NeurIPS 2025 · similarity 0.89)
- Theoretical Guarantees for Variational Inference with Fixed-Variance Mixture of Gaussians (ICML 2024 · similarity 0.89)
- Online Variational Sequential Monte Carlo (ICML 2024 · similarity 0.89)
- Reparameterized Importance Sampling for Robust Variational Bayesian Neural Networks (ICML 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.