Research map / Federated learning and privacy / Differential privacy
Towards practical differentially private causal graph discovery
Lun Wang, Qi Pang, Dawn Song
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 sieve-and-examine with sub-sampling (Priv-PC) instead of exponential mechanism (EM-PC) | PC algorithm · all 4 datasets | running time: 10.61 to 32.85 times speedup | Improved |
| process online queries so removed edges skip later tests (ablation) | Priv-PC · all 4 datasets | running time: 5.97 to 27.41 times speedup | Improved |
| add sub-sampling in the sieve process (ablation) | Priv-PC · all 4 datasets | running time: 1.20 to 2.40 times speedup | Improved |
| remove sub-sampling (ablation) | Priv-PC · Earthquake | running time (s): 6.62s -> 11.01s | Hurt |
| remove sub-sampling (ablation) | Priv-PC · Survey | running time (s): 2.13s -> 5.12s | Hurt |
| use sieve-and-examine instead of plain sparse vector technique | PC algorithm · Earthquake, Cancer, Asia, Survey | F1-score | Improved |
| use Priv-PC instead of EM-PC at large privacy budget | PC algorithm · all 4 datasets | F1-score | Improved |
| use Priv-PC instead of EM-PC at small privacy budget (about 0.01 to 0.1) | PC algorithm · all 4 datasets | F1-score | Hurt |
| use sub-sampled Gaussian with moments accountant | PC algorithm · all 4 datasets | F1-score | Hurt |
| compare Priv-PC to SVT-PC runtime | Priv-PC · Earthquake | running time (s): 3.38s vs 6.62s | Mixed |
Closest papers by meaning
- Causal Discovery with Fewer Conditional Independence Tests (ICML 2024 · similarity 0.89)
- On Causal Discovery in the Presence of Deterministic Relations (NeurIPS 2024 · similarity 0.86)
- Stable Differentiable Causal Discovery (ICML 2024 · similarity 0.85)
- Federated Causality Learning with Explainable Adaptive Optimization (AAAI 2024 · similarity 0.85)
- Improving the Variance of Differentially Private Randomized Experiments through Clustering (ICML 2025 · similarity 0.85)
- Federated Causal Discovery from Heterogeneous Data (ICLR 2024 · 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.
