Use equivalent sample size instead of a fixed alpha to define priors

bmihaljevic / bnclassify

Learning Discrete Bayesian Network Classifiers from Data

18 stars 10 forks source link

Open bmihaljevic opened 5 years ago

bmihaljevic commented 5 years ago

A fixed alpha limits to K2-type priors, while equivalent sample size allows for score-equivalence and thus seems more natural.