Open UnixJunkie opened 3 years ago
this is just a RF variant: at each node, a (feature, value) pair is select out of several random (feature, value) pairs. They don't train on a bootstrap from the training set, but on the whole dataset.
P. Geurts, D. Ernst., and L. Wehenkel, “Extremely randomized trees”, Machine Learning, 63(1), 3-42, 2006.
Should be way faster than training a RF. We'll see about the model performance though.
this is just a RF variant: at each node, a (feature, value) pair is select out of several random (feature, value) pairs. They don't train on a bootstrap from the training set, but on the whole dataset.