Table 2.

Training and running times of our methods. Training includes training a model on the entire synthetic dataset. Running involves genotype-correcting a single VCF file (96,000 variants) with a pre-trained model. The benchmarks were run on an Intel i5-7200U CPU @ 2.5 GHz.

Training time
Logistic regression	67 s
XGBoost	552 s

Training time
Logistic regression	0.03 s
XGBoost	0.67 s
Restricted Bayesian	0.12 s