Linear Genetic Programming

Front Cover
Springer Science & Business Media, Feb 25, 2007 - Computers - 316 pages

Linear Genetic Programming examines the evolution of imperative computer programs written as linear sequences of instructions. In contrast to functional expressions or syntax trees used in traditional Genetic Programming (GP), Linear Genetic Programming (LGP) employs a linear program structure as genetic material whose primary characteristics are exploited to achieve acceleration of both execution time and evolutionary progress. Online analysis and optimization of program code lead to more efficient techniques and contribute to a better understanding of the method and its parameters. In particular, the reduction of structural variation step size and non-effective variations play a key role in finding higher quality and less complex solutions. This volume investigates typical GP phenomena such as non-effective code, neutral variations and code growth from the perspective of linear GP.

The text is divided into three parts, each of which details methodologies and illustrates applications. Part I introduces basic concepts of linear GP and presents efficient algorithms for analyzing and optimizing linear genetic programs during runtime. Part II explores the design of efficient LGP methods and genetic operators inspired by the results achieved in Part I. Part III investigates more advanced techniques and phenomena, including effective step size control, diversity control, code growth, and neutral variations.

The book provides a solid introduction to the field of linear GP, as well as a more detailed, comprehensive examination of its principles and techniques. Researchers and students alike are certain to regard this text as an indispensable resource.

 

What people are saying - Write a review

We haven't found any reviews in the usual places.

Contents

INTRODUCTION
1
12 Genetic Programming
3
13 Linear Genetic Programming
6
14 Motivation
8
BASIC CONCEPTS OF LINEAR GENETIC PROGRAMMING
13
22 Execution of Programs
25
23 Evolution of Programs
29
CHARACTERISTICS OF THE LINEAR REPRESENTATION
35
76 Initialization of Linear Programs
164
77 Constant Program Length
169
78 Summary and Conclusion
170
A COMPARISON WITH TREEBASED GENETIC PROGRAMMING
173
82 Benchmark Problems
177
83 Experimental Setup
181
84 Experiments and Comparison
185
85 Discussion
190

32 Structural Introns and Semantic Introns
37
33 Graph Interpretation
47
34 Analysis of Program Structure
56
35 Graph Evolution
60
36 Summary and Conclusion
61
A COMPARISON WITH NEURAL NETWORKS
63
42 Benchmark Data sets
64
43 Experimental Setup
65
44 Experiments and Comparison
69
45 Summary and Conclusion
74
LINEAR GENETIC OPERATORS I SEGMENT VARIATIONS
76
51 Variation Effects
78
52 Effective Variation and Evaluation
79
53 Variation Step Size
80
54 Causality
82
55 Selection of Variation Points
86
56 Characteristics of Variation Operators
87
57 Segment Variation Operators
89
58 Experimental Setup
99
59 Experiments
102
510 Summary and Conclusion
118
LINEAR GENETIC OPERATORS II INSTRUCTION MUTATIONS
119
62 Instruction Mutation Operators
121
63 Experimental Setup
129
64 Experiments
131
65 Summary and Conclusion
148
ANALYSIS OF CONTROL PARAMETERS
149
72 Number of Output Registers
156
73 Rate of Constants
157
74 Population Size
159
75 Maximum Program Length
162
86 Summary and Conclusion
191
CONTROL OF DIVERSITY AND VARIATION STEP SIZE
194
92 Structural Program Distance
197
93 Semantic Program Distance
200
94 Control of Diversity
201
95 Control of Variation Step Size
203
96 Experimental Setup
205
97 Experiments
206
98 Alternative Selection Criteria
222
99 Summary and Conclusion
223
CODE GROWTH AND NEUTRAL VARIATIONS
225
101 Code Growth in GP
226
102 Proposed Causes of Code Growth
227
103 Influence of Variation Step Size
229
104 Neutral Variations
230
105 Conditional Reproduction and Variation
232
106 Experimental Setup
233
108 Control of Code Growth
249
109 Summary and Conclusion
259
EVOLUTION OF PROGRAM TEAMS
261
112 Team Evolution
262
113 Combination of Multiple Predictors
265
114 Experimental Setup
273
115 Experiments
276
116 Combination of Multiple Program Outputs
286
117 Summary and Conclusion
287
Epilogue
288
References
291
Index
303
Copyright

Other editions - View all

Common terms and phrases

References to this book

About the author (2007)

Markus Brameier received a PhD degree in Computer Science from the Department of Computer Science at University of Dortmund, Germany,in 2004. From 2003 to 2004 he was a postdoctoral fellow at the Stockholm Bioinformatics Center (SBC), a collaboration between Stockholm University, the Royal Institute of Technology, and Karolinska Institute, in Sweden. Currently he is Assistant Professor at the Bioinformatics Research Center (BiRC) of the University of Aarhus in Denmark. His primary research interests are in bioinformatics and genetic programming.

Wolfgang Banzhaf is a professor of Computer Science at the Department of Computer Science of Memorial University of Newfoundland, Canada, and head of the department since 2003. Prior to that, he served for 10 years as Associate Professor for Applied Computer Science in the Department of Computer Science at University of Dortmund, Germany. From 1989 to 1993 he was a researcher with Mitsubishi Electric Corp., first in MELCO’s Central Research Lab in Japan, then in the United States at Mitsubishi Electric Research Labs Inc., Cambridge, MA. Between 1985 and 1989 he was a postdoc in the Department of Physics, University of Stuttgart, Germany. He holds a PhD in Physics from the University of Karlruhe in Germany. His research interests are in the field of artificial evolution and self-organization studies. He has recently become more involved with bioinformatics.

Bibliographic information