Leo Breiman
University of California, Berkeley
The talk will cover the development of a suite of "two-eyed" algorithms. Thier purpose is not only to give very accurate predictions, but also to be able to look inside the black box of complex ensemble methods and find out which variaables are instrumental in driving the predictions, locate outliers, and give instructive low dimensional views of the data. This will be illustrated by looking at the random forests program and a recent algorithm developed for predicting multiple dependent outputs.