Part (b): The figure above shows the performance (on training and testing sets) of the AdaBoost algorithm (boosting on decision stumps) for the simulated data discussed in Figure 10.2 in the book and using the Matlab code supplied here. This image matches very closely the similar image presented in the book. In this case there is not much overlap in the features space so the ultimate classification performance is relatively good (about 15 percent). Note at no time did the testing error begin to increase.

---

John Weatherwax

Last modified: Sun May 15 08:46:34 EDT 2005