Binary Classification on SCM spurious shift (test)
0.733ROC AUCPRISM-Causal
Evaluation Results
| Method | Links | |
|---|---|---|
| PRISM-CausalPrivacy budget (epsilon)=1.0, Allocation strategy=Unif, Evaluation protocol=TSTR, Number of synthetic samples (nsyn)=5000, Feature selection (k)=82026.02 | 0.733 | |
| PRISM-CausalPrivacy budget (epsilon)=1.0, Allocation strategy=Opt, Evaluation protocol=TSTR, Number of synthetic samples (nsyn)=5000, Feature selection (k)=82026.02 | 0.732 | |
| MST (Private-PGM)Privacy budget (epsilon)=1.0, Evaluation protocol=TSTR2026.02 | 0.519 | |
| PRISM-GraphicalPrivacy budget (epsilon)=1.0, Allocation strategy=Opt, Evaluation protocol=TSTR, Number of synthetic samples (nsyn)=5000, Feature selection (k)=82026.02 | 0.513 | |
| PRISM-GraphicalPrivacy budget (epsilon)=1.0, Allocation strategy=Unif, Evaluation protocol=TSTR, Number of synthetic samples (nsyn)=5000, Feature selection (k)=82026.02 | 0.513 | |
| All FeaturesPrivacy budget (epsilon)=1.0, Allocation strategy=Unif, Evaluation protocol=TSTR2026.02 | 0.512 | |
| Independence OnlyPrivacy budget (epsilon)=1.0, Evaluation protocol=TSTR2026.02 | 0.502 | |
| PrivBayes (DataSynthesizer)Privacy budget (epsilon)=1.0, Evaluation protocol=TSTR2026.02 | 0.501 | |
| Corr Top-k (Oracle)Privacy budget (epsilon)=1.0, Evaluation protocol=TSTR2026.02 | 0.49 |