Posted by shali on January 26, 19104 at 14:50:02:
Hi, I am doing the simulaiton study for boosting. I tried to apply the simulation method described in "Greedy Function Approximation: A Gradient Boosting Machine" written by J.H. Freidman, but got some trouble. My questions are
1) what is the "haar measure";
2) are those 100 target functions equivalent to 100 trials?
3) I know x is generated from 10-dimensional standard normal,but how to deal with the 7500 sample size? Does it mean that each target function with 10 inputs contains 7500 observations?