Table 3 Test-set error for different classifiers and number of boosting iterations. The table displays the test set error averaged over 100 simulation runs for different classifiers and the number of boosting iterations (M; the situation with M =1 is the performance of the base classifier) for the setting with 1000 variables and 50 samples. The difference between the classes was moderate, the correlation structure was exchangeable and there were 10 variables per block, see the “Methods” section for more details

From: Boosting for high-dimensional two-class prediction