CFE 2019: Start Registration
View Submission - CMStatistics
Title: Component-wise gradient boosting for functional shape regression Authors:  Almond Stoecker - Humboldt University of Berlin (Germany) [presenting]
Sonja Greven - Humboldt University of Berlin (Germany)
Abstract: In 2D, the shape of an object, such as a bone or a body cell, may be represented either by a collection of prominent points (landmark shape) or by its outline (functional shape), which are considered modulo the shape preserving transformations of translation, rotation and scaling. Component-wise gradient boosting for regression with shape responses is developed allowing for modular specification of multiple linear or smooth covariate effects in an additive predictor. Boosting is particularly attractive in this context due to its step-wise procedure and inherent regularization, which provides automated variable selection and is thus well suited for complex model scenarios potentially involving many covariates.