R. Busa-Fekete, B. Kégl, T. Éltető & G. Szarvas;
JMLR W&CP 14:37–48, 2011.
Ranking by calibrated AdaBoost
This paper describes the ideas and methodologies that we used in the Yahoo learning-to-rank
Our technique is essentially pointwise with a listwise touch at the last combination step. The main
ingredients of our approach are 1) preprocessing (querywise normalization) 2) multi-class
3) regression calibration, and 4) an exponentially weighted forecaster for model
combination. In post-challenge analysis we found that preprocessing and training AdaBoost
with a wide variety of hyperparameters improved individual models signiﬁcantly, the
ﬁnal listwise ensemble step was crucial, whereas calibration helped only in creating
Page last modified on Wed Jan 26 10:36:50 2011.