Research map / Probabilistic and scientific modeling / Causal discovery
Sample Complexity Bounds for Score-Matching: Causal Discovery and Generative Modeling
Zhenyu Zhu, Francesco Locatello, Volkan Cevher
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 |
|---|---|---|---|
| increase C_m from 1 to 512 (ablation) | SCORE with ReLU score network · d=100, n=100 | SHD: 2941.0±29.5 -> 13.8 ± 1.8 | Improved |
| increase C_m from 1 to 4 (ablation) | SCORE with ReLU score network · d=100, n=100 | SHD: 2941.0±29.5 -> 2900.6 ± 80.8 | No clear effect |
| increase C_m from 16 to 64 (ablation) | SCORE with ReLU score network · d=100, n=100 | SHD: 1512.4 ± 283.6 -> 55.0 ± 16.0 | Improved |
| increase C_m from 128 to 512 (ablation) | SCORE with ReLU score network · d=100, n=100 | SHD: 23.9 ± 4.6 -> 13.8 ± 1.8 | Improved |
| increase number of samples n from 5 to 160 (ablation) | SCORE with ReLU score network · d=10, C_m=1 | SHD: 31.7±2.1 -> 13.0± 4.0 | Improved |
| increase n from 20 to 40 (ablation) | SCORE with ReLU score network · d=10, C_m=1 | SHD: 23.3±2.7 -> 23.0±4.0 | No clear effect |
| increase number of variables d from 5 to 100 (ablation) | SCORE with ReLU score network · n=10, C_m=1 | SHD: 4.5 ± 2.0 -> 2923.7 ± 38.5 | Hurt |
| increase d from 10 to 20 (ablation) | SCORE with ReLU score network · n=10, C_m=1 | SHD: 29.6 ± 2.2 -> 124.3 ± 4.6 | Hurt |
Closest papers by meaning
- Score-informed Neural Operator for Enhancing Ordering-based Causal Discovery (NeurIPS 2025 · similarity 0.87)
- Neural Network-Based Score Estimation in Diffusion Models: Optimization and Generalization (ICLR 2024 · similarity 0.86)
- Algorithm- and Data-Dependent Generalization Bounds for Diffusion Models (NeurIPS 2025 · similarity 0.86)
- Since Faithfulness Fails: The Performance Limits of Neural Causal Discovery (ICML 2025 · similarity 0.86)
- Optimal Kernel Choice for Score Function-based Causal Discovery (ICML 2024 · similarity 0.86)
- Differentiable Constraint-Based Causal Discovery (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.
