Aprendizaje estadístico para la selección de algoritmos en problemas de optimización
Loading...
Identifiers
Publication date
Authors
Advisors
Tutors
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
En este trabajo se emplearán técnicas de aprendizaje estadístico para predecir el optimizador global que mejor funciona dado un problema de programación matemática no lineal. Antes de nada, se explicará un método de resolución de problemas de programación lineal entera y su adaptación al caso no lineal, donde surgirán nuevas dificultades. Posteriormente, se presentará el problema de aprendizaje estadístico y dos técnicas que permiten ajustar un modelo y crear predicciones: la regresión lineal y las redes neuronales de una sola capa oculta. Estas técnicas permitirán realizar el aprendizaje sobre un conjunto de problemas y obtener los resultados, viendo el desempeño de los distintos optimizadores.
In this work, statistical learning techniques will be used to predict the best performing global optimiser for a non-linear mathematical programming problem. First of all, a method for solving integer linear programming problems and its adaptation to the nonlinear case, where new difficulties will arise, will be explained. Subsequently, the statistical learning problem and two techniques that allow to fit a model and create predictions will be presented: linear regression and single hidden layer neural networks. These techniques will allow learning to be performed on a set of problems and the results to be obtained, looking at the performance of the different optimisers.
In this work, statistical learning techniques will be used to predict the best performing global optimiser for a non-linear mathematical programming problem. First of all, a method for solving integer linear programming problems and its adaptation to the nonlinear case, where new difficulties will arise, will be explained. Subsequently, the statistical learning problem and two techniques that allow to fit a model and create predictions will be presented: linear regression and single hidden layer neural networks. These techniques will allow learning to be performed on a set of problems and the results to be obtained, looking at the performance of the different optimisers.
Description
48 páxs
Keywords
Bibliographic citation
Relation
Has part
Has version
Is based on
Is part of
Is referenced by
Is version of
Requires
Sponsors
Rights
Attribution-NonCommercial-ShareAlike 4.0 International








