ROBUST ESTIMATION OF THE MEAN OF A RANDOM MATRIX: A NONASYMPTOTIC STUDY

This thesis is concerned with the estimation of the mean of a random matrix when there are no assumptions about the tail of the distributions that are related to the matrix. More specifically, the estimation procedure contemplates that the distribution of the elements of the random matrix could be h...

Descripción completa

Detalles Bibliográficos
Autor: Roberto Cabal
Tipo de recurso: tesis de maestría
Estado:Versión aceptada para publicación
Fecha de publicación:2020
País:México
Institución:Centro de Investigación en Matemáticas
Repositorio:Repositorio Institucional CIMAT
OAI Identifier:oai:cimat.repositorioinstitucional.mx:1008/1082
Acceso en línea:http://cimat.repositorioinstitucional.mx/jspui/handle/1008/1082
Access Level:acceso abierto
Palabra clave:info:eu-repo/classification/MSC/PROBABILIDAD Y ESTADÍSTICA
info:eu-repo/classification/cti/1
info:eu-repo/classification/cti/12
info:eu-repo/classification/cti/1299
info:eu-repo/classification/cti/129999
Descripción
Sumario:This thesis is concerned with the estimation of the mean of a random matrix when there are no assumptions about the tail of the distributions that are related to the matrix. More specifically, the estimation procedure contemplates that the distribution of the elements of the random matrix could be heavy-tailed. For this reason, we develop concentration inequalities for the estimators around the mean matrix in such a way that the theoretical guarantees give us, for example, valuable information about how to choose the hyperparameters related to the estimator. Of particular interest is the robust estimation of the covariance matrix from a random sample, which has numerous applications in statistical science such as Factor Analysis and Principal Components Analysis. Other famous applications of matrix concentration inequalities are in the fields of Matrix Completion and community detection in Random Graphs Theory.