Research map / Probabilistic and scientific modeling / Variational and generative sampling
Compositional Score Modeling for Simulation-Based Inference
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 PF-NPSE with m=6 instead of the compared methods | SBI posterior estimation · Gaussian/Gaussian, Gaussian/Mixture of Gaussians, SIR, Lotka–Volterra | squared MMD | Helped |
| Use F-NPSE rather than NRE with HMC | Multimodal posterior example · Illustrative multimodal example | posterior mode coverage | Helped |
| Increase the simulator-call budget | Compared SBI methods · Gaussian/Gaussian, Gaussian/Mixture of Gaussians, SIR, Lotka–Volterra | squared MMD | Helped |
| Use PF-NPSE or F-NPSE instead of NRE | SBI with multiple observations · Gaussian/Gaussian, Gaussian/Mixture of Gaussians, SIR, Lotka–Volterra | squared MMD | Helped |
| Vary PF-NPSE subset size m from 1 through 30 (ablation) | PF-NPSE · Gaussian/Gaussian, Gaussian/Mixture of Gaussians, SIR, Lotka–Volterra; 22 observations | squared MMD | Mixed |
| Use intermediate subset sizes m=3 or m=6 rather than either endpoint (ablation) | PF-NPSE · Gaussian/Gaussian, Gaussian/Mixture of Gaussians, SIR, Lotka–Volterra | squared MMD | Helped |
| Constrain the score network to a conservative parameterization (ablation) | NPSE, F-NPSE, and PF-NPSE · Gaussian/Gaussian, Gaussian/Mixture of Gaussians, SIR, Lotka–Volterra | squared MMD | No clear effect |
| Vary annealed Langevin sampler parameters while retaining sufficient steps (ablation) | PF-NPSE · Gaussian/Gaussian, Gaussian/Mixture of Gaussians, SIR, Lotka–Volterra | squared MMD | No clear effect |
| Use the alternative sampler without unadjusted Langevin dynamics instead of annealed Langevin dynamics (ablation) | F-NPSE and PF-NPSE · Gaussian/Gaussian, Gaussian/Mixture of Gaussians, SIR, Lotka–Volterra | squared MMD | Hurt |
| Increase the number of Langevin steps per noise level (ablation) | PF-NPSE · Gaussian/Gaussian, Gaussian/Mixture of Gaussians, SIR, Lotka–Volterra | squared MMD | Helped |
| Combine individual-observation posterior scores using F-NPSE | F-NPSE · Weinberg simulator | posterior concentration around the true parameter | Helped |
Closest papers by meaning
- Sequential Neural Score Estimation: Likelihood-Free Inference with Conditional Score Based Diffusion Models (ICML 2024 · similarity 0.90)
- Compositional simulation-based inference for time series (ICLR 2025 · similarity 0.88)
- Learning Likelihood-Free Reference Priors (ICML 2025 · similarity 0.86)
- CoLT: The conditional localization test for assessing the accuracy of neural posterior estimates (NeurIPS 2025 · similarity 0.86)
- Simultaneous identification of models and parameters of scientific simulators (ICML 2024 · similarity 0.86)
- Multilevel neural simulation-based inference (NeurIPS 2025 · similarity 0.85)
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.