A Support Vector Machine (SVM) for Functional Data Analysis
by
Hong-Li Zeng(Uppsala University)
→
Europe/Stockholm
122:026
122:026
Description
This talk will present the Longitudinal Support Vector Machine (LSVM) algorithm. This is an extension of the standard SVM towards classification of functional data. Functional data or longitudinal data represent observations which are made repeatedly over time. This problem requires modifications of the definition of the margin which underlies the standard SVM. The resulting LSVM results also in a convex optimisation problem, and the dual optimisation problem is derived.
We will present results on a specific application concerning the evolution of brain structures. Here we are interested in classifying the brain structures arising in evolution governed by natural selection, and in evolution governed by artificial selection. That is, we are interested in the differences of brain structures in either evolution regime. This talk then demonstrate how one can use the LSVM for analysis.
Empirical results indicate the efficacy of the LSVM for analysing such functional and evolutionary data.