Learning decision trees
Parameter learning:
- As with tabular representations
- Multinomial distribution at each leaf
- Counts are at the level of leaves
-
Structure learning
- Define the MDL or marginal likelihood
- General structure similar to scores of Bayesian networks