Closed cmencar closed 2 years ago
The formula is from the original paper, page 7: https://arxiv.org/pdf/0811.1679.pdf
If you go down the tree, there are also many duplicated rules, as sub-paths to the terminal nodes are shared between the terminal nodes.
Ch. "INTERPRETABLE MODELS", section "RuleFit", subs. "Theory", subsubs. "step 1". The the number of rules created from an ensemble of M trees should be
K = \sum_{i=1}^M{2^{t_m-1}}
That is, the number of rules grows exponentially with the number of terminal nodes.