Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Generalized Bayesian Inference for Scientific Simulators via Amortized Cost Estimation
Richard Gao, Michael Deistler, Jakob H. Macke
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| ACE-GBI instead of NPE on real recordings | NPE · Hodgkin-Huxley Allen data | fraction of predictives with more than five spikes: 35.6% -> 66% | Improved |
| raise beta from 25 to 100 in ACE (ablation) | ACE · Hodgkin-Huxley Allen data | fraction of predictives with more than five spikes: 66% -> 96% | Improved |
| evaluate NPE on synthetic prior predictives instead of experimental observations | NPE · Hodgkin-Huxley | average predictive distance: 0.189 vs 0.174 | Improved |
| train NPE with 10x more simulations (1M versus 100K) (ablation) | NPE · Hodgkin-Huxley Allen data | posterior predictive quality | Improved |
| increase beta (more concentrated generalized posterior) (ablation) | ACE · Benchmark tasks | posterior predictive distance | Improved |
| increase beta (ablation) | ACE · Benchmark tasks | C2ST against ground-truth GBI posterior | Hurt |
| ACE instead of NPE/NLE on misspecified observations | NPE and NLE · Benchmark tasks (4 tasks) | average posterior predictive distance | Improved |
| tempering NLE likelihood by beta instead of ACE (ablation) | NLE · Uniform 1D and Linear Gaussian misspecified | posterior predictive distance | Hurt |
| ABC with lower simulation budget (1000 or fewer) (ablation) | ABC · Benchmark tasks | posterior predictive distance | Hurt |
| ACE versus ABC on higher-dimensional parameters | ABC · Linear Gaussian (10D) | posterior predictive distance | Hurt |
| simulation budget 200, 1000 versus 10000 (ablation) | ACE · Benchmark tasks | relative performance across algorithms | No clear effect |
Closest papers by meaning
- Effortless, Simulation-Efficient Bayesian Inference using Tabular Foundation Models (NeurIPS 2025 · similarity 0.89)
- Compositional simulation-based inference for time series (ICLR 2025 · similarity 0.89)
- All-in-one simulation-based inference (ICML 2024 · similarity 0.89)
- Multilevel neural simulation-based inference (NeurIPS 2025 · similarity 0.88)
- Robust Simulation-Based Inference under Missing Data via Neural Processes (ICLR 2025 · similarity 0.87)
- Consistency Models for Scalable and Fast Simulation-Based Inference (NeurIPS 2024 · 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.
