Forrest, Daniel L. K. (2001-11-30)
Recent work has shown that AdaBoost can be viewed as an algorithm that
maximizes the margin on the training data via functional gradient descent. Under
this interpretation, the weight computed by AdaBoost, for each hyp ...