If n is the number of samples and there are m attributes then tree learning is O(m* n* log n), a Random forest which optimises for best split is O(T* m* n* log n) where there are T trees.
For extremely randomised trees where we don't optimise for best split locally we drop the linear dependence on n, so it is O(T* m* log n)?
Aucun commentaire:
Enregistrer un commentaire