Causal Discovery on Synthetic Scale-Free networks d=20
14.52SHDcDCD-PFN
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DCD-PFNdimension (d)=20, zero-shot=true, number of samples (n)=8002026.06 | 14.52 | 82.6 | 30.1 | 43.5 | |
| DirectLiNGAMdimension (d)=20, number of samples (n)=8002026.06 | 15.64 | 98.2 | 18.3 | 30 | |
| GOLEMdimension (d)=20, number of samples (n)=8002026.06 | 15.8 | 98.3 | 17.5 | 28.8 | |
| ICALiNGAMdimension (d)=20, number of samples (n)=8002026.06 | 16.2 | 93 | 14.9 | 25.1 | |
| Notearsdimension (d)=20, number of samples (n)=8002026.06 | 17.12 | 92 | 9.9 | 17.6 | |
| PC (E-KCIT)dimension (d)=20, number of samples (n)=8002026.06 | 18.04 | 17.7 | 10.9 | 13.3 | |
| GESdimension (d)=20, number of samples (n)=8002026.06 | 19.84 | 17 | 14.9 | 15.9 | |
| PC (KCIT)dimension (d)=20, number of samples (n)=8002026.06 | 20.28 | 12.5 | 10.1 | 11.1 | |
| AVICIdimension (d)=20, zero-shot=true, number of samples (n)=8002026.06 | 21.76 | 48.3 | 17.3 | 23.1 | |
| GraNDAGdimension (d)=20, number of samples (n)=8002026.06 | 31 | 29 | 9.9 | 12.3 |