The Geodise

OptionsMatlab Toolbox

A User's Guide

Credits

Release:            OptionsMatlab v.0.13.0

Version:            OptionsMatlabTutorial.doc 1.6.0

Title:                 Geodise OptionsMatlab Tutorial – A User's Guide

Authors:           Dr Graeme Pound

                        Dr Andrew Price, a.r.price@soton.ac.uk

PI:                    Prof Andy Keane, andy.keane@soton.ac.uk

                        Prof Simon Cox, s.j.cox@soton.ac.uk

Web:                http://www.geodise.org/

Copyright:        Copyright © 2007, The Geodise Project, University of Southampton

Contents

The Geodise OptionsMatlab Toolbox. 1

1 Introduction. 6

2 Installing OptionsMatlab. 9

2.1 Obtain a Gendat license file. 9

2.2 Add OptionsMatlab to the Matlab search path. 9

2.3 View the documentation. 10

3 OptionsMatlab Tutorial 11

3.1 Create a input structure. 11

3.2 Run the search. 12

3.3 View the search histories. 13

3.4 Build and search a Response Surface Model 13

4 Function Reference. 15

Banana problem.. 15

Beam problem.. 17

Bump problem.. 19

optimisationAppendDataPoints . 21

optimisationCropDataPoints . 25

optimisationReplaceDataPoints . 27

optimisationDigest 30

optimisationHistory . 32

optimisationSampleRSM ... 34

optimisationSearchTrace . 39

optimisationTerrain . 42

optimisationTestSuite . 44

optimisationTestSuiteComb . 44

optimisationTestSuiteUncon . 44

optimisationTestSuiteSPM ... 45

optimisationTilePlot 50

optimisationTrace. 55

OptionsMatlab. 57

optjob. 67

optjobparallel 70

optjobparallel2. 72

optum1. 74

Peaks4d problem.. 76

5 Frequently Asked Questions. 78

5.1 Why does Matlab crash when I call OptionsMatlab?. 78

5.2 How do I specify the search method?. 78

5.3 How do I run a Design of Experiments?. 80

5.4 How do I build a Response Surface Model?. 81

5.5 How do I plot my Response Surface Model?. 81

5.6 How do I generate Design of Experiment update points?. 83

5.7 How do I define an unconstrained optimisation?. 84

5.8 How do I write my own objective and constraint functions?. 85

5.9 How do I evaluate a combined objective and constraint function?. 86

5.10 Can OptionsMatlab calculate function evaluations in parallel?. 87

5.11 How do I tune the hyper-parameters for a stochastic process model RSM?. 89

5.12 Can I checkpoint the progress of an optimisation?. 92

5.13 How do I pass Matlab variables to my objective function? . 93

5.14 How do I define discrete design variables?. 93

5.15 How do I restart a Genetic Algorithm?. 95

5.16 What is the meaning of the optional control parameters?. 97

5.17 How do I deal with failed calculations when constructing a response surface model? 107

5.18 How do I build and evaluate a RSM faster?. 108

6 OptionsMatlab Examples. 109

6.1 DoE Direct search. 109

6.2 RSM returning update points. 111

6.3 DoE evaluating candidate points. 111

6.4 RSM using candidate points. 112

6.5 Direct search with checkpointing. 113

6.6 Parallel job submission with userdata. 114

6.7 Hyper-parameter tuning. 116

6.8 User-defined sequential optimiser 116

6.9 Sample a Response Surface Model 117

6.10 Build a stochastic process model RSM with quick tuning. 118

6.11 Search a tuned stochastic process model RSM... 119

6.12 Search the root mean square error of a tuned stochastic process model RSM... 120

6.13 Search the expected improvement of a tuned stochastic process model RSM... 123

6.14 Search the probability of improvement of a tuned stochastic process model RSM... 124

6.15 Search the constrained expected improvement of a tuned stochastic process model RSM 125

6.16 Search the constrained feasibility of improvement of a tuned stochastic process model RSM 127

7 References. 130

introduction

Copyright © 2007, The Geodise Project, University of Southampton