Abstract
The success of evolutionary search depends on adequate parameter settings. Ill conditioned strategy parameters decrease the success probabilities of genetic operators. Proper settings may change during the optimization process. The question arises if adequate settings can be found automatically during the optimization process. Evolution strategies gave an answer to the online parameter control problem decades ago: self-adaptation. Self-adaptation is the implicit search in the space of strategy parameters. The self-adaptive control of mutation strengths in evolution strategies turned out to be exceptionally successful. Nevertheless, for years self-adaptation has not achieved the attention it deserves. This paper is a survey of self-adaptive parameter control in evolutionary computation. It classifies self-adaptation in the taxonomy of parameter setting techniques, gives an overview of automatic online-controllable evolutionary operators and provides a coherent view on search techniques in the space of strategy parameters. Beyer and Sendhoff’s covariance matrix self-adaptation evolution strategy is reviewed as a successful example for self-adaptation and exemplarily tested for various concepts that are discussed.
Similar content being viewed by others
Notes
Typically, genetic operators are crossover/recombination, mutation, inversion, gene deletion and duplication.
According to a tradition in ES, μ is the size of the parental population, while λ is the size of the offspring population.
Originally, σ denotes step sizes in ES. We use this symbol for all kinds of strategy variables in this paper.
The concept of global multi-recombination will be explained in Sect. 4.2.
For log-normal mutation of step sizes, see Sect. 4.1.2.
The index j denotes the index of the j-th ranked individual of the λ offspring individuals with regard to fitness f(x j ).
The Sphere problem is the continuous optimization problem to find an \({\bf x} \in {\mathbb{R}}^N\) minimizing \(f({\bf x}):=\sum_{i=1}^{N} x_{i}^{2}.\)
Optimal progress is problem-dependent and can be stated theoretically on artificial functions [12].
The success probability p s of an evolutionary operator is the relation of solutions with better fitness than the original solution f(x) and all generated solutions.
OneMax is a pseudo-binary function. For a bit string \({\bf x} {\in}\{0,1\}^N\), maximize \(f({\bf x}):=\sum_{i=1}^{N}x_i \quad\hbox{ with } {\bf x} \in \{0,1\}^N\) optimum x * = (1, ..., 1) with f(x *) = N.
3-SAT is a propositional satisfiability problem, each clause contains k = 3 literals; we randomly generate formulas with N = 50 variables and 150 clauses; the fitness function is: \({\frac{\#\hbox{of true clauses}} {\#\hbox{of all clauses}}}.\)
Schwefel’s problem 2.40 [69]: Minimize f(x): = − ∑ 5i=1 x i , constraints \(g_{j}({\bf x}):=x_{j} {\geq}0, \hbox{for } j=1,\ldots,5\) and \(g_{j}({\bf x})=-\sum\nolimits^{5}_{i=1}(9+i)x_{i}+50000 {\geq}0, \hbox{for } j=6\), minimum x * = (5000, 0, 0, 0, 0)T with f(x *) = − 5000.
TR (tangent problem): minimize \(f({\bf x}):=\sum_{i=1}^{N} x_{i}^{2}\), constraints \(g({\bf x}):=\sum_{i=1}^{N} x_{i}-t>0, \qquad t \in {\mathbb{R}} \qquad \hbox{(tangent)},\) for N=k and t=k the minimum lies at: \({\bf x}^{*}=(1,\ldots,1)^{T},\hbox{with }\quad f({\bf x}^{*})=k\), TR2 means N = 2 and t = 2.
difference between the optimum and the best solution \(|f({\bf x}^{*})-f({\bf x}^{best})|.\)
Low innovation rates are caused by variation operators that produce offspring not far away from their parents, e.g. by low mutation rates.
References
Angeline PJ (1995) Adaptive and self-adaptive evolutionary computations. In: Palaniswami M, Attikiouzel Y (eds) Computational intelligence a dynamic systems perspective. IEEE Press, New York, pp 152–163
Arnold DV, Brauer D (2008) On the behaviour of the (1+1)-ES for a simple constrained problem. In: Proceedings of the 10th conference on parallel problem solving from nature—PPSN X, pp 1–10
Auger A (2003) Convergence results for (1, λ)-SA-ES using the theory of ϕ-irreducible markov chains. In: Proceedings of the evolutionary algorithms workshop of the 30th international colloquium on automata, languages and programming
Bäck T (1998) An overview of parameter control methods by self-adaption in evolutionary algorithms. Fundam Inf 35(1–4):51–66
Bartz-Beielstein T (2006) Experimental research in evolutionary computation: the new experimentalism. Natural computing series. Springer, April
Bartz-Beielstein T, Lasarczyk C, Preu M (2005) Sequential parameter optimization. In: McKay B, et al (eds) Proceedings of the IEEE congress on evolutionary computation—CEC, vol 1. IEEE Press, pp 773–780
Bartz-Beielstein T, Preuss M (2006) Considerations of budget allocation for sequential parameter optimization (SPO). In: Paquete L, et al. (eds) Workshop on empirical methods for the analysis of algorithms, proceedings, Reykjavik, Iceland, pp 35–40
Bäck T (1991) Self-adaptation in genetic algorithms. In: Proceedings of the 1st European conference on artificial life—ECAL, pp 263–271
Bäck T (1992) The interaction of mutation rate, selection, and self-adaptation within a genetic algorithm. In: Proceedings of the 2nd conference on parallel problem solving from nature—PPSN II, pp 85–94
Bäck T, Schütz M (1996) Intelligent mutation rate control in canonical genetic algorithms. In: Foundation of intelligent systems, 9th international symposium, ISMIS ’96. Springer, pp 158–167
Berlik S (2004) A step size preserving directed mutation operator. In: Proceedings of the 6th conference on genetic and evolutionary computation—GECCO, pp 786–787
Beyer H-G (2001) The theory of evolution strategies. Springer, Berlin
Beyer H-G, Deb K (2001) On self-adaptive features in real-parameter evolutionary algorithms. IEEE Trans Evol Comput 5(3):250–270
Beyer H-G, Meyer-Nieberg S (2006) Self-adaptation on the ridge function class: first results for the sharp ridge. In: Proceedings of the 9th conference on parallel problem solving from nature—PPSN IX, pp 72–81
Beyer HG, Schwefel HP (2002) Evolution strategies—a comprehensive introduction. Nat Comput 1:3–52
Beyer HG, Sendhoff B (2008) Covariance matrix adaptation revisited—the cmsa evolution strategy. In: Proceedings of the 10th conference on parallel problem solving from nature—PPSN X, pp 123–132
Coello CA (2000) Use of a self-adaptive penalty approach for engineering optimization problems. Comput Ind 41(2):113–127
Davis L (1989) Adapting operator probabilities in genetic algorithms. In: Proceedings of the 3rd international conference on genetic algorithms, San Francisco, Morgan Kaufmann Publishers Inc, pp 61–69
de Landgraaf W, Eiben A, Nannen V (2007) Parameter calibration using meta-algorithms. In: Proceedings of the IEEE congress on evolutionary computation—CEC, pp 71–78
DeJong K (2007) Parameter setting in EAs: a 30 year perspective. In: Parameter setting in evolutionary algorithms, studies in computational intelligence. Springer, pp 1–18
Eiben A, Schut MC, de Wilde A (2006) Is self-adaptation of selection pressure and population size possible? A case study. In: Proceedings of the 9th conference on parallel problem solving from nature—PPSN IX, pp 900–909
Eiben AE, Hinterding R, Michalewicz Z (1999) Parameter control in evolutionary algorithms. IEEE Trans Evol Comput 3(2):124–141
Eiben AE, Michalewicz Z, Schoenauer M, Smith JE (2007) Parameter control in evolutionary algorithms. In: Parameter setting in evolutionary algorithms, studies in computational intelligence. Springer, pp 19–46
Fogarty TC (1989) Varying the probability of mutation in the genetic algorithm. In: Proceedings of the 3rd international conference on genetic algorithms, San Francisco, Morgan Kaufmann Publishers Inc, pp 104–109
Fogel DB, Fogel LJ, Atma JW (1991) Meta-evolutionary programming. In: Proceedings of 25th asilomar conference on signals, systems & computers, pp 540–545
georg Beyer H, Arnold DV (2003) Qualms regarding the optimality of cumulative path length control in csa/cma-evolution strategies. Evol Comput 11
Goldberg D (1989) Genetic algorithms in search, optimization and machine learning. Addison Wesley, Reading
Grefenstette J (1986) Optimization of control parameters for genetic algorithms. IEEE Trans Syst Man Cybern 16(1):122–128
Hansen N (2006) An analysis of mutative sigma self-adaptation on linear fitness functions. Evol Comput 14(3):255–275
Harik GR, Goldberg DE (1997) Learning linkage. In: Foundations of genetic algorithms 4. Morgan Kaufmann, pp 247–262
Harik GR, Lobo FG, Sastry K (2006) Linkage learning via probabilistic modeling in the extended compact genetic algorithm (ECGA). In: Scalable optimization via probabilistic modeling, studies in computational intelligence, Springer, pp 39–61
Herdy M (1992) Reproductive isolation as strategy parameter in hierarchically organized evolution strategies. In: Proceedings of the 10th conference on parallel problem solving from nature—PPSN II, pp 207–217
Hesser J, Männer R (1990) Towards an optimal mutation probability for genetic algorithms. In: Proceedings of the 10th conference on parallel problem solving from nature—PPSN I, London, UK, Springer-Verlag, pp 23–32
Hesser J, Männer R (1992) Investigation of the m-heuristic for optimal mutation probabilities. In PPSN, pp 115–124
Hildebrand L (2002) Asymmetrische evolutionsstrategien. PhD thesis, University of Dortmund
Holland JH (1992) Adaptation in natural and artificial systems, 1st edn, MIT Press, Cambridge
Jägersknpper J (2005) Rigorous runtime analysis of the (1+1) es: 1/5-rule and ellipsoidal fitness landscapes. In: Proceedings of the workshop on foundation of genetic algorithms FOGA, pp 260–281
Jägersknpper J (2006) Probabilistic runtime analysis of (1 + λ)es using isotropic mutations. In: Proceedings of the 8th conference on genetic and evolutionary computation—GECCO, New York, ACM, pp 461–468
Joines J, Houck C (1994) On the use of non-stationary penalty functions to solve nonlinear constrained optimization problems with GAs. In: Fogel DB (eds) Proceedings of the 1st IEEE conference on evolutionary computation, Orlando, Florida, IEEE Press, pp 579–584
Jong KAD (1975) An analysis of the behavior of a class of genetic adaptive systems. PhD thesis, University of Michigan
Koumoutsakos P, Muller SD (2006) Flow optimization using stochastic algorithms. Lecture Notes Control Inf Sci 330:213–229
Kramer O (2008) Premature convergence in constrained continuous search spaces. In: Proceedings of the 10th conference on parallel problem solving from nature—PPSN X, Berlin, Springer, to appear
Kramer O (2008) Self-adaptive inversion mutation for combinatorial representations. In: Proceedings of the 2008 international conference on genetic and evolutionary methods, to appear
Kramer O, Barthelmes A, Rudolph G (2009) Surrogate constraint functions for cma evolution strategies. In: Proceedings of the conference on artificial intelligence and automation, page to appear
Kramer O, Koch P (2007) Self-adaptive partially mapped crossover. In: Proceedings of the 9th conference on genetic and evolutionary computation—GECCO, New York, ACM Press, pp 1523–1523
Kramer O, Ting CK, Büning HK (2005) A new mutation operator for evolution strategies for constrained problems. In: Proceedings of the IEEE congress on evolutionary computation—CEC, pp 2600–2606
Kursawe F (1999) Grundlegende empirische Untersuchungen der Parameter von Evolutionsstrategien—Metastrategien. PhD thesis, University of Dortmund
Liang KH, Yao X, Liu Y, Newton CS, Hoffman D (1998) An experimental investigation of self-adaptation in evolutionary programming. In: Proceedings of the 7th international conference on evolutionary programming VII—EP, Berlin, Springer, pp 291–300
Maruo MH, Lopes HS, Delgado MR (2005) Self-adapting evolutionary parameters: encoding aspects for combinatorial optimization problems. In: Proceedings of EvoCOP, pp 154–165
Mercer RE, Sampson JR (1978) Adaptive search using a reproductive metaplan. Kybernetes 7:215–228
Mersch B, Glasmachers T, Meinicke P, Igel C (2006) Evolutionary optimization of sequence kernels for detection of bacterial gene starts. In: ICANN (2), pp 827–836
Meyer-Nieberg S, Beyer HG (2007) Self-adaptation in evolutionary algorithms. In: Lobo FG, Lima CF, Michalewicz Z (eds) Parameter setting in evolutionary algorithms. Springer, Berlin
Mezura-Montes E, Palomeque-Ortiz AG (2009) Self-adaptive and deterministic parameter control in differential evolution for constrained optimization. Constraint-Handl Evol Optim 189:95–120
Mühlenbein H (1992) How genetic algorithms really work: mutation and hillclimbing. In: Proceedings of the 2nd conference on parallel problem solving from nature—PPSN II, pp 15–26
Nannen V, Eiben A (2006) A method for parameter calibration and relevance estimation in evolutionary algorithms. In: Proceedings of the 8th conference on genetic and evolutionary computation—GECCO, New York, ACM Press, pp 183–190
Nannen V, Eiben A (2007) Relevance estimation and value calibration of evolutionary algorithm parameters. In: IJCAI, pp 975–980
Ostermeier A, Gawelczyk A, Hansen N (1994) A derandomized approach to self adaptation of evolution strategies. Evol Comput 2(4):369–380
Ostermeier A, Gawelczyk A, Hansen N (1995) A derandomized approach to self adaptation of evolution strategies. Evol Comput 2(4):369–380
Preuss M, Bartz-Beielstein T (2007) Sequential parameter optimization applied to self-adaptation for binary-coded evolutionary algorithms. In: Parameter setting in evolutionary algorithms, studies in computational intelligence. Springer, pp 91–119
Rechenberg I (1973) Evolutionsstrategie: Optimierung technischer Systeme nach Prinzipien der biologischen Evolution. Frommann-Holzboog, Stuttgart
Rechenberg I (1994) Evolutionsstrategie ’94. Frommann-Holzboog, Stuttgart
Reed J, Toombs R, Barricelli NA (1967) Simulation of biological evolution and machine learning: I. selection of self-reproducing numeric patterns by data processing machines, effects of hereditary control, mutation type and crossing. J Theor Biol 17:319–342
Reinelt G (1991) Tsplib—a traveling salesman problem library. ORSA J Comput 3:376–384
Rosenberg RS (1967) Simulation of genetic populations with biochemical properties. PhD thesis, University of Michigan
Rudolph G (2001) Self-adaptive mutations may lead to premature convergence. IEEE Trans Evol Comput 5(4):410–414
Schaffer JD, Caruana R, Eshelman LJ, Das R (1989) A study of control parameters affecting online performance of genetic algorithms for function optimization. In: Proceedings of the 3rd international conference on genetic algorithms—ICGA 1989, pp 51–60
Schaffer JD, Morishima A (1987) An adaptive crossover distribution mechanism for genetic algorithms. In: Proceedings of the second international conference on genetic algorithms on genetic algorithms and their application, Hillsdale, NJ, USA, L. Erlbaum Associates Inc, pp 36–40
Schwefel H-P (1974) Adaptive Mechanismen in der biologischen Evolution und ihr Einflus auf die Evolutionsgeschwindigkeit. Interner Bericht der Arbeitsgruppe Bionik und Evolutionstechnik am Institut fnr Mess- und Regelungstechnik, TU Berlin
Schwefel HP (1995) Evolution and Optimum Seeking. Sixth-generation computer technology. Wiley Interscience, New York
Semenov MA, Terkel DA (2003) Analysis of convergence of an evolutionary algorithm with self-adaptation using a stochastic lyapunov function. Evol Comput 11(4):363–379
Smith J (2001) Modelling GAs with self adaptive mutation rates. In: Proceedings of the genetic and evolutionary computation conference, pp 599–606
Smith J, Fogarty TC (1996) Recombination strategy adaptation via evolution of gene linkage. In: Proceedings of the IEEE congress on evolutionary computation—CEC, pp 826–831
Smith J, Fogarty TC (1996) Self adaptation of mutation rates in a steady state genetic algorithm. In: Proceedings of the international conference on evolutionary computation—ICEC, pp 318–323
Spears WM (1995) Adapting crossover in evolutionary algorithms. In: McDonnell JR, Reynolds RG, Fogel DB (eds) Proceedings of the fourth annual conference on evolutionary programming, Cambridge, MIT Press, pp 367–384
Stone C, Smith J (2002) Strategy parameter variety in self-adaptation of mutation rates. In: Proceedings of the genetic and evolutionary computation conference—GECCO, San Francisco, Morgan Kaufmann Publishers Inc, pp 586–593
Weinberg R (1970) Computer simulation of a living cell. PhD thesis, University of Michigan
Acknowledgments
The author thanks Günter Rudolph and the anonymous reviewers for their helpful comments to improve the manuscript.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Kramer, O. Evolutionary self-adaptation: a survey of operators and strategy parameters. Evol. Intel. 3, 51–65 (2010). https://doi.org/10.1007/s12065-010-0035-y
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12065-010-0035-y