Development of the fdakma R package for the joint alignment and clustering of functional data: application to neuronal spike trains data

Patriarca, Mirco

Development of the fdakma R package for the joint alignment and clustering of functional data: application to neuronal spike trains data

Monday 22nd July 2013

Sangalli, L.

Vantini, S.

In this work we present the basics of functional data analysis, pointing out the latest developments in this new branch of statistics and in particular in the subfield of functional data registration, in order to give a solid theoretical framework for the solution of the registration problem. Three registration methods are presented, including the K-Mean Alignment method, able to jointly cluster and align functional data, and the one based on Fisher-Rao metric. The K-Mean Alignment technique will be used to analyze a dataset containing spike trains intensities, i.e., data coming from brain impulses of a monkey doing a particular task using his hand. In the last part of this work we develop an R package called fdakma. This package, available on CRAN, resumes the K-Mean Alignment method and allows user to jointly cluster and align functional data using different metrics/semimetrics and different types of warping functions. KEYWORDS: Functional data analysis; Clustering; Alignment; Time warping; Similarity; Spike trains.