Evolutionary Optimization: the uGP toolkit (eBook)

Preface 6
Contents 9
Chapter 1 Evolutionary computation 12
1.1 Natural and artificial evolution 12
1.2 The classical paradigms 15
1.3 Genetic programming 18
Chapter 2 Why yet another one evolutionary optimizer? 19
2.1 Background 19
2.2 Where to draw the lines 20
2.3 Individuals 21
2.4 Problem specification 23
2.5 Coding Techniques 24
Chapter 3The µGP architecture 26
3.1 Conceptual design 27
3.2 The evolutionary core 27
3.2.1 Evolutionary Operators 28
3.2.2 Population 29
3.3 The Evolutionary Cycle 30
3.3.1 Genetic operator selection 30
3.3.2 Parents selection 31
3.3.3 Offspring Generation 32
3.3.4 Individual Evaluation and Slaughtering 33
3.3.5 Termination and Aging 33
Chapter 4 Advanced features 35
4.1 Self adaptation for exploration or exploitation 35
4.1.1 Self-adaptation inertia 36
4.1.2 Operator strength 36
4.1.3 Tournament size 37
4.2 Escaping local optimums 37
4.2.1 Operator activation probability 38
4.2.2 Tuning the elitism 38
4.3 Preserving diversity 39
4.3.1 Clone detection, scaling and extermination 40
4.3.2 Entropy and delta-entropy computation 40
4.3.3 Fitness holes 41
4.3.4 Population topology and multiple populations 42
4.4 Coping with the real problems 43
4.4.1 Parallel fitness evaluation 44
4.4.2 Multiple fitness 45
Chapter 5 Performing an evolutionary run 46
5.1 Robot Pathfinder 48
5.2 µGPSettings 50
5.3 Population Settings 52
5.4 Library of Constraints 56
5.5 Launching the experiment 60
5.6 µGP Extractor 62
Chapter 6 Command line syntax 64
6.1 Starting a run 64
6.2 Controlling messages to the user 65
6.3 Getting help and information 66
6.4 Controlling logging 66
6.5 Controlling recovery 67
6.6 Controlling evolution 68
6.7 Controlling evaluation 69
Chapter 7 Syntax of the settings file 71
7.1 Controlling evolution 72
7.2 Controlling logging 74
7.3 Controlling recovery 75
Chapter 8 Syntax of the population parameters file 77
8.1 Strategy parameters 77
8.1.1 Base parameters 78
8.1.2 Parameters for self adaptation 81
8.1.3 Other parameters 84
Chapter 9 Syntax of the external constraints file 86
9.1 Purposes of the constraints 86
9.2 Organization of constraints and hierarchy 87
9.3 Specifying the structure of the individual 92
9.4 Specifying the contents of the individual 95
Chapter 10 Writing a compliant evaluator 101
10.1 Information from µGP to the fitness evaluator 101
10.2 Expected fitness format 102
10.2.1 Good Examples 103
10.2.2 Bad Examples 104
Chapter 11 Implementation details 106
11.1 Design principles 106
11.2 Architectural choices 107
11.2.1 The Graph library 108
11.2.2 The Evolutionary Core library 110
11.2.3 Front end 117
11.3 Code organization and class model 117
Chapter 12 Examples and applications 128
12.1 Classical one-max 128
12.1.1 Fitness evaluator 129
12.1.2 Constraints 131
12.1.3 Population settings 133
12.1.4 µGP settings 135
12.1.5 Running 136
12.2 Values of parameters and their influence on the evolution: Arithmetic expressions 137
12.2.1 De Jong 3 137
12.2.2 De Jong 4 Modified 142
12.2.3 Carrom 143
12.3 Complex individuals’ structures and evaluation: Bit-counting in Assembly 149
12.3.1 Assembly individuals representation 149
12.3.2 Evaluator 152
12.3.3 Running 154
Appendix A Argument and option synopsis 155
Appendix B External constraints synopsis 171
References 179