(Downloads - 0)
For more info about our services contact : help@bestpfe.com
Table of contents
Abstract
Résumé
Acknowledgments
Contents
1 Introduction
1.1 Algorithm Selection
1.2 Exploratory Landscape Analysis
1.3 Key Objective of the Thesis
1.4 Outline of the Thesis
2 Contributions of the Thesis
2.1 Combining Fixed-Budget Regression Models
2.2 Impact of Hyper-Parameter Tuning
2.3 Personalized Performance Regression
2.4 Adaptive Landscape Analysis
2.5 Trajectory-Based Algorithm Selection
I The Background
3 Black-Box Optimization
3.1 Black-Box Optimization Algorithms
3.1.1 CMA-ES
3.1.2 Modular CMA-ES
3.1.3 Additional Algorithms
3.2 Algorithm Performance Measures
3.3 Problem Collections
3.3.1 BBOB Test Suite
4 Exploratory Landscape Analysis
4.1 ELA Features
4.2 Choice of Features
4.3 Feature Computation
5 Algorithm Selection
5.1 Per-Instance Algorithm Selection
5.2 From Performance Regression to Algorithm Selection
5.3 State of the Art
5.4 Performance Assessment of Algorithm Selectors
II Contributions
6 Combining Fixed-Budget Regression Model
6.1 Preliminaries
6.2 Fixed-Budget Performance Regression
6.2.1 Impact of Feature Selection
6.3 Fixed-Budget Algorithm Selection
6.3.1 Impact of the Threshold Value and the Feature Portfolio
6.3.2 Impact of the Algorithm Portfolio
6.3.3 Impact of the Sample Size for Feature Extraction
6.4 Conclusions
7 Impact of Hyper-Parameter Tuning
7.1 Preliminaires
7.2 Performance Regression Quality of Different Models
7.3 ELA-Based Algorithm Selection
7.4 Sensitivity Analyses
7.5 Conclusions
8 Personalized Performance Regression
8.1 Preliminaries
8.2 Personalized Machine Learning Models
8.3 Use-Case: ELA-Based Fixed-Budget Performance Regression
8.3.1 Experimental Setup
8.3.2 BIPOP-CMA-ES Performance Prediction
8.4 Conclusions
9 Adaptive Landscape Analysis
9.1 Preliminaries
9.2 “Zooming In” into the Landscapes
9.3 Conclusions
10 Trajectory-Based Performance Regression
10.1 Preliminaries
10.2 Supervised Machine Learning for Performance Regression
10.3 Comparison with Global Feature Values
10.4 Sensitivity Analyses
10.5 Conclusions
11 General Conclusions and Outlook
Bibliography
