Technion, IEM faculty - Operations Research seminar
Speaker: Ronny Luss
Title: Isotonic Recursive Partitioning
Date: 17/01/2011
Time: 12:30
Place: Bloomfield-527
Abstract:  <>

Abstract: Isotonic regression is a widely studied nonparametric
approach for fitting monotonic models to data and has been
studied from both theoretical and practical aspects. We present
an algorithm, which we term Isotonic Recursive Partitioning
(IRP), for isotonic regression based on recursively partitioning
the covariate space through solution of progressively smaller
``best cut'' subproblems. This creates a regularized sequence of
isotonic models of increasing model complexity that converges to
the global isotonic regression solution. The models along the
sequence are often more accurate than the unregularized isotonic
regression model because of the complexity control they offer.
We quantify this complexity control through estimation of
degrees of freedom along the path. Furthermore, we show that IRP
for the classic l2 isotonic regression can be generalized to
convex differentiable loss functions such as Huber's loss.
Success of the regularized models in prediction and IRP's
favorable computational properties are demonstrated through a
series of simulated and real data experiments.
Technion Math. Net (TECHMATH)
Editor: Michael Cwikel   <> 
Announcement from:  <>