Making kernel machines scalable combining matrix approximations and distributed computing

In this work, kernelized binary support vector machines are implemented based on stochastic gradient descent. The Scala library can be used both on a single computing node and on a Spark cluster. Additional tools for parameter tuning, subset selection, and model evaluation are implemented.

Detalhes bibliográficos
Autor: Gerstenlauer, Jakob L.K.
Formato: tesis de maestría
Fecha de publicación:2018
País:España
Recursos:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/118760
Acesso em linha:https://hdl.handle.net/2117/118760
Access Level:acceso abierto
Palavra-chave:Kernel functions
Support vector machines
Machine learning
SVM
support vector machine
kernel methods
map reduce
Apache Spark
Kernel, Funcions de
Aprenentatge automàtic
Àrees temàtiques de la UPC::Informàtica
Descrição
Resumo:In this work, kernelized binary support vector machines are implemented based on stochastic gradient descent. The Scala library can be used both on a single computing node and on a Spark cluster. Additional tools for parameter tuning, subset selection, and model evaluation are implemented.