Research map / Probabilistic and scientific modeling / Variational and generative sampling
Geometry-Aware Probabilistic Circuits via Voronoi Tessellations
Sahil Sidheekh, Sriraam Natarajan
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 |
|---|---|---|---|
| add Voronoi tessellation gating (certified lower bound) | EinsumNet · UCI power | mean test log-likelihood: 0.56 -> 1.73 | Improved |
| add Voronoi tessellation gating (certified lower bound) | EinsumNet · UCI gas | mean test log-likelihood: 5.85 -> 7.13 | Improved |
| add Voronoi tessellation gating (certified lower bound) | EinsumNet · UCI hepmass | mean test log-likelihood: -22.35 -> -22.98 | Hurt |
| add Voronoi tessellation gating (certified lower bound) | EinsumNet · UCI miniboone | mean test log-likelihood: -32.24 -> -32.82 | Hurt |
| add Voronoi tessellation gating (certified lower bound) | HCLT · UCI power | mean test log-likelihood: 0.55 -> 5.10 | Improved |
| add Voronoi tessellation gating (certified lower bound) | HCLT · UCI gas | mean test log-likelihood: 7.22 -> 10.57 | Improved |
| add Voronoi tessellation gating (certified lower bound) | HCLT · UCI hepmass | mean test log-likelihood: -20.36 -> -19.10 | Improved |
| add Voronoi tessellation gating (certified lower bound) | HCLT · UCI miniboone | mean test log-likelihood: -25.43 -> -26.46 | Hurt |
| add hierarchical factorized Voronoi gating (exact inference) | EinsumNet · UCI miniboone | mean test log-likelihood: -32.24 -> -32.18 | No clear effect |
| add hierarchical factorized Voronoi gating (exact inference) | HCLT · UCI gas | mean test log-likelihood: 7.22 -> 7.07 | No clear effect |
| increase data dimensionality for VT bounds | VT-HCLT · UCI hepmass vs power | mean test log-likelihood: 6.77 vs 5.10 (power); -1.61 vs -19.10 (hepmass) | Hurt |
Closest papers by meaning
- Winner-takes-all learners are geometry-aware conditional density estimators (ICML 2024 · similarity 0.88)
- Scaling Probabilistic Circuits via Monarch Matrices (ICML 2025 · similarity 0.87)
- On the Hardness of Approximating Distributions with Tractable Probabilistic Models (NeurIPS 2025 · similarity 0.86)
- On the Expressive Power of Tree-Structured Probabilistic Circuits (NeurIPS 2024 · similarity 0.85)
- On the Relationship Between Monotone and Squared Probabilistic Circuits (AAAI 2025 · similarity 0.84)
- Sum of Squares Circuits (AAAI 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.
