Dataset Viewer

The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Data for Causal Relationship Probability Estimation on CPDAGs

Dataset Structure

Synthetic Data

Generated from random DAGs (Erdos-Renyi, Scale-Free, Barabasi-Albert) with linear Gaussian SEMs. Each .pkl file contains estimated CPDAGs, CI confidence scores, and ground-truth causal relationship probabilities.

Directory Description Node Range
synthetic/train/ Training set (19,500 graphs) n=5-50
synthetic/val/ Validation set (1,500 graphs) n=5-50
synthetic/test_small/ Test Group I n=5-15
synthetic/test_medium/ Test Group II n=15-30
synthetic/test_large/ Test Group III n=30-50
synthetic/test_ood/ Test Group IV (out-of-distribution) n=100-300

Real-World Data

Directory Description Source
real/sachs/ Sachs protein signaling network (n=11) Sachs et al., 2005
real/dream4/ DREAM4 gene regulatory networks (n=100) Marbach et al., 2010

Semi-Real Data

Directory Description
semi_real/ 8 bnlearn networks (asia, child, insurance, water, sachs, barley, hailfinder, win95pts) with parametric observations

Robustness Evaluation Data

Directory Description
robustness/cross_sem/ Non-linear and non-Gaussian SEM variants
robustness/cross_algo/ CPDAGs estimated by alternative algorithms (GES)

File Format

All .pkl files are Python pickle files loadable with pickle.load(). Each file contains a list of dictionaries with keys including cpdag, confidence, ground_truth, and graph metadata.

License

This dataset is released under the CC-BY-4.0 license.

Downloads last month
8