Research map / Probabilistic and scientific modeling / Variational and generative sampling
Amortized Monte Carlo Integration
Adam Golinski, Frank Wood, Tom Rainforth
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 |
|---|---|---|---|
| Use three separately trained proposals to estimate positive, negative, and normalization terms | SNIS-style amortized expectation estimation · 1D Gaussian tail integral | ReMSE: about 10^3 more samples to achieve the same relative error | Improved |
| Use three separately trained proposals to estimate positive, negative, and normalization terms | SNIS-style amortized expectation estimation · 5D Gaussian tail integral | ReMSE | Improved |
| Use three separately trained proposals to estimate positive, negative, and normalization terms | SNIS-style amortized expectation estimation · Cancer treatment planning | ReMSE: N=10^4; N=2 | Improved |
| Use three separately trained proposals instead of the mixture proposal in SNIS | SNIS q_m · 1D Gaussian tail integral | ReMSE: about 10^3 more samples to achieve the same relative error | Improved |
| Use three separately trained proposals instead of the mixture proposal in SNIS | SNIS q_m · 5D Gaussian tail integral | ReMSE | Improved |
| Use three separately trained proposals instead of the mixture proposal in SNIS | SNIS q_m · Cancer treatment planning | ReMSE | Improved |
| Use AMCI rather than the theoretical SNIS error bound | SNIS estimator · 5D Gaussian tail integral | ReMSE | Hurt |
| Reuse samples from q_1 to estimate the denominator, setting beta above zero (ablation) | AMCI · Tail integral, low-mismatch case | ReMSE | No clear effect |
| Reuse samples from q_1 to estimate the numerator, setting alpha below one (ablation) | AMCI · Tail integral, low-mismatch case | ReMSE: N=M=64 | Improved |
| Reuse samples from q_1 to estimate the numerator, setting alpha below one (ablation) | AMCI · Tail integral, high-mismatch case | ReMSE: N=M=64 | No clear effect |
| Train using a rotating synthetic training and validation dataset rather than drawing every batch directly from the model (ablation) | AMCI proposals · Training procedure | training convergence | Improved |
| Use SNIS q_1 rather than SNIS q_2 | SNIS · Tail integral | ReMSE | Hurt |
Closest papers by meaning
- Leveraging Self-Consistency for Data-Efficient Amortized Bayesian Inference (ICML 2024 · similarity 0.87)
- All-in-one simulation-based inference (ICML 2024 · similarity 0.86)
- Direct Amortized Likelihood Ratio Estimation (AAAI 2024 · similarity 0.86)
- Consistency Models for Scalable and Fast Simulation-Based Inference (NeurIPS 2024 · similarity 0.85)
- Compositional simulation-based inference for time series (ICLR 2025 · similarity 0.85)
- To discretize continually: Mean shift interacting particle systems for Bayesian inference (NeurIPS 2026 · 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.
