Hello, I am using random forest in my code. And I am not sure that how I should calculate "max_num_of_trees_in_the_forest " that is a parameter of CvRTParams() function.
I know the machine shouldnt memorize all of my train datas .
It's much easier to overfit the data with a larger max_depth than with max_num_of_trees_in_the_forest. In fact, from my experience with random forest it is safe to use a large number of trees eg. in the 30,40,50s. I would plot a graph of "accuracy vs number of trees". You should get a graph that asymptotes off for high number of trees. This should give you an idea of what number to pick.
Asked: 2014-01-21 08:55:28 -0600
Seen: 275 times
Last updated: Jan 21 '14
