Missing data imputation through generative topographic mapping as a mixture of t-distributions: Theoretical developments

The Generative Topographic Mapping (GTM) was originally conceived as a probabilistic alternative to the well-known, neural network-inspired, Self-Organizing Map (SOM). The GTM can also be interpreted as a constrained mixture of distributions model. In recent years, much attention has been directed t...

Descripción completa

Detalles Bibliográficos
Autor: Vellido Alcacena, Alfredo|||0000-0002-9843-1911
Tipo de recurso: informe técnico
Fecha de publicación:2004
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/83308
Acceso en línea:https://hdl.handle.net/2117/83308
Access Level:acceso abierto
Palabra clave:Missing data
Outliers
Generative topographic mapping
Student multivariate t-distributions
Robust imputation
Data visualization
Àrees temàtiques de la UPC::Informàtica::Intel·ligència artificial
Descripción
Sumario:The Generative Topographic Mapping (GTM) was originally conceived as a probabilistic alternative to the well-known, neural network-inspired, Self-Organizing Map (SOM). The GTM can also be interpreted as a constrained mixture of distributions model. In recent years, much attention has been directed towards Student t-distributions as an alternative to Gaussians in mixture models due to their robustness towards outliers. In this report, the GTM is redefined as a constrained mixture of t-distributions: the t-GTM, and the Expectation-Maximization algorithm that is used to fit the model to the data is modified to provide missing data imputation.