Bgorven

Statistical understanding of AdaBoost

Hello Bgorven,

On March 12, 2014, you made a substantial change to the AdaBoost article which among other things added the following:

Specifically, in the case where all weak learners are known a priori, AdaBoost corresponds to a single iteration of the backfitting algorithm in which the smoothing splines are the minimizers of $\sum _{i=1}^{n}e^{-Y_{i}{\hat {\mu }}(x_{i})}+\infty \int _{x_{1}}^{x_{n}}{\hat {\mu }}''(x)^{2}\,dx$, that is: ${\hat {\mu }}_{i}$ fits an exponential cost function and is linear with respect to the observation.

Is the infinity in the formula a typo? Also, do you have a source for this? I tried finding information connecting AdaBoost to backfitting, but could not find one.

I am going to add {{fact}} for now.

« D. Trebbien (talk) 19:32, 8 May 2016 (UTC)