Aumentar Tamaño del texto Disminuir Tamaño del texto

Irene Epifanio López, Universitat Jaume I


A tour of functional archetypal analysis


We will take a tour of archetype analysis and a derived method, archetypoid analysis, in particular in the functional case. Archetype analysis is an unsupervised statistical learning technique that lies somewhere in between principal component analysis and cluster analysis.  Archetypal analysis approximates data by means of mixtures of actual extreme cases (archetypoids) or archetypes, which are a convex combination of observations in the data set.  Therefore, its results can be easily interpreted by non-experts. We will review the definition of archetypal analysis in the classic multivariate case and we will extend archetypal analysis to functional data. We will see how to use it to detect (functional) outliers. Applications in very diverse fields and open lines of future work will also be seen.