Bayesian Optimization in Machine Learning

Bayesian optimization has risen over the last few years as a very attractive approach to find the optimum of noisy, expensive to evaluate, and possibly black-box functions. One of the fields where these functions are common is in machine-learning, where one typically has to fit a particular model by...

Descripción completa

Detalles Bibliográficos
Autor: Jiménez Luna, José
Tipo de recurso: tesis de maestría
Fecha de publicación:2017
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:2117/105999
Acceso en línea:https://hdl.handle.net/2117/105999
Access Level:acceso abierto
Palabra clave:Sequences (Mathematics)
Machine learning
Bayesian
Optimization
Seqüències (Matemàtica)
Classificació AMS::62 Statistics::62L Sequential methods
Àrees temàtiques de la UPC::Matemàtiques i estadística::Estadística matemàtica
Descripción
Sumario:Bayesian optimization has risen over the last few years as a very attractive approach to find the optimum of noisy, expensive to evaluate, and possibly black-box functions. One of the fields where these functions are common is in machine-learning, where one typically has to fit a particular model by minimizing a specified form of loss. In this Master s thesis we first focus on reviewing the most recent literature on Gaussian Processes as well as Bayesian optimiza- tion methods, then we benchmark said methods against several real case machine-learning scenarios and lastly we provide open source software that will allow researchers to apply these strategies in other problems.