Research map / Learning theory and algorithms / Sample complexity
The Fisher Dimension: Instance-Dependent Complexity for Causal Discovery
Luong Doan, Khanh Nguyen Quoc, Duc Hai Nguyen, Mai Phan Quoc Hung, Phong Ho, Nhung Duong, Tuan Do
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 Fisher dimension predictor instead of structural proxies for tree graphs | Fisher dimension · Tree graphs (d=10) | Spearman correlation with n*: 0.88 | Improved |
| use Fisher dimension predictor instead of structural proxies for chain graphs | Fisher dimension · Chain graphs (d=10) | Spearman correlation with n*: 0.93 | Improved |
| use average Markov blanket size instead of Fisher dimension | Fisher dimension · Erdos-Renyi graphs (d=10) | Spearman correlation with n*: 0.37 -> 0.58 | Hurt |
| move from structured families (tree, chain) to Erdos-Renyi graphs | Fisher dimension predictor · d=10 graphs | Spearman rho: 0.91 -> 0.66 | Hurt |
| move from chain to Erdos-Renyi graphs | Fisher dimension predictor · d=10 graphs | Pearson r: 0.92 -> 0.55 | Hurt |
| Erdos-Renyi low-F cell versus medium-F cell | Fisher dimension prediction · Erdos-Renyi graphs | fraction within factor 2: 55% -> 80% | Mixed |
| increase graph size d from 5 to 20 | PC algorithm · Chain graphs | empirical sample complexity n*: 1,267 -> 4,367 | Hurt |
| increase graph size d from 5 to 20 | PC algorithm · Erdos-Renyi graphs | empirical sample complexity n*: 863 -> 10,000 | Hurt |
| increase graph size d from 5 to 20 | PC algorithm · Tree graphs | empirical sample complexity n*: 1,383 -> 7,800 | Hurt |
| overall agreement of prediction across all cells | Fisher dimension prediction · all graph families | fraction within factor 2: 85% | Improved |
Closest papers by meaning
- Latent Hierarchical Causal Structure Discovery with Rank Constraints (NeurIPS 2022 · similarity 0.88)
- Scores for Learning Discrete Causal Graphs with Unobserved Confounders (AAAI 2024 · similarity 0.88)
- Causal Structure Discovery from Distributions Arising from Mixtures of DAGs (ICML 2020 · similarity 0.87)
- Standardizing Structural Causal Models (ICLR 2025 · similarity 0.87)
- Coarsening Linear Non-Gaussian Causal Models with Cycles (NeurIPS 2026 · similarity 0.87)
- Learning Mixtures of Unknown Causal Interventions (NeurIPS 2024 · similarity 0.86)
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.
