|
NeuroCOLT
Technical Report NC-TR-01-098
2001-098
On
the Convergence of Leveraging
Gunnar
Ratsch
Sebastian Mika
Manfred
K. Warmuth
We
give an unified convergence analysis of ensemble learning methods
including e.g.\ AdaBoost, Logistic Regression and the
Least-Square-Boost algorithm for regression. These methods have in
common that they iteratively call a base learning algorithm, which
returns hypotheses that are then linearly combined. We show that these
methods are related to the \emph{Gauss-Southwell method} known from
numerical optimization and show \emph{linear convergence} for all
these methods. Our analysis includes $\ell_1$ regularized cost functions
leading to a clean and general way to regularize ensemble learning.
Download
PDF
|