Reasonable BN learning time range

[SEICS]

Hi,

I am training a Bayesian Network (BN) containing 12 attributes with a 10,002,716 rows × 12 columns dataset. The structure of this BN model would be chow-liu tree, where each attribute has maximum one parent. Some of my attributes have relative large number of states (e.g., 3000) so I know training could take a long time. Since I have waited for the training result for an hour and the learning is still continuing, I wonder what is the reasonable BN learning time range for BayesNets.jl. I am a bit concerned the training will cause "out of memory" problem.

Thank you for help!

[tawheeler]

Did you set max_n_parents=1? That should be pretty quick. You could always cut your dataset down drastically and see how long that takes to train.