A1012
Title: Identifying functional shape outliers through complexity-based mixture models
Authors: Enea Bongiorno - Universita del Piemonte Orientale (Italy) [presenting]
Kwo Lik Lax Chan - Universita del Piemonte Orientale (Italy)
Aldo Goia - Universita' del Piemonte Orientale (Italy)
Abstract: The purpose is to address the problem of shape outliers in the functional data setting by interpreting them as contamination originating from high-complexity components within a suitable mixture model framework. The study has three main objectives. First, it introduces a notion of complexity grounded in the concept of small-ball probability. Second, it formally defines a complexity-based mixture model and explores its theoretical consequences on small ball probabilities. Third, it proposes an algorithm to decompose such mixtures into distinct components, thereby enabling the implicit detection of potential contamination in functional datasets. The practical utility of the proposed approach is illustrated through an empirical case study.
