Research map / Probabilistic and scientific modeling / Variational and generative sampling
Approximate Knowledge Compilation by Online Collapsed Importance Sampling
Tal Friedman, Guy Van den Broeck
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 the partially compiled SDD proposal instead of dummy or true-marginal proposals (ablation) | Collapsed compilation with FD or MinEnt · 50-20 | Hellinger distance: 2.37e-04 -> 3.72e-07; 1.77e-04 -> 3.72e-07 | Improved |
| Use the partially compiled SDD proposal instead of dummy or true-marginal proposals (ablation) | Collapsed compilation with MinEnt · 50-20 | Hellinger distance: 3.29e-04 -> 2.10e-08; 1.31e-03 -> 2.10e-08 | Improved |
| Set the circuit size threshold to 100k rather than 10k or 1m (ablation) | Collapsed compilation on 50-20 · 50-20 | Hellinger distance: 7.33e-5 -> 3.21e-7 -> 7.53e-6 | Mixed |
| Use MinEnt rather than FD for variable selection (ablation) | Collapsed compilation · DBN | Hellinger distance: 1.24e-1 -> 1.83e-2 | Improved |
| Use RBVar or MinEnt variable selection rather than the compared baselines | Collapsed compilation · DBN | Hellinger distance: 6.27e-3; 1.83e-2 versus baselines 6.27e-1 to 6.39e-1 | Improved |
| Use collapsed compilation rather than EDBP or samplesearch on frust16 | Collapsed compilation (FD) · frust16 | Hellinger distance: 4.73e-3 -> 5.96e-6; 6.23e-3 -> 5.96e-6 | Improved |
| Increase the samplesearch treewidth limit from 10 to 12 (ablation) | IJGP-Samplesearch · linkage | Hellinger distance: 4.93e-2 -> 1.10e-3 | Improved |
Closest papers by meaning
- Sound Abstraction and Decomposition of Probabilistic Programs (ICML 2018 · similarity 0.87)
- Total Variation Distance Meets Probabilistic Inference (ICML 2024 · similarity 0.85)
- Large Language Bayes (NeurIPS 2025 · similarity 0.84)
- Efficient and Accurate Explanation Estimation with Distribution Compression (ICLR 2025 · similarity 0.84)
- On the Hardness of Approximating Distributions with Tractable Probabilistic Models (NeurIPS 2025 · similarity 0.84)
- BIRD: A Trustworthy Bayesian Inference Framework for Large Language Models (ICLR 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.
