#### DMCA

## Escaping Hierarchical Traps with Competent Genetic Algorithms (2001)

Venue: | Proceedings of the Genetic and Evolutionary Computation Conference (GECCO2001 |

Citations: | 101 - 49 self |

### Citations

10034 |
Genetic Algorithms
- Goldberg
- 1989
(Show Context)
Citation Context ...tion is formed. However,sxed, problem-independent, recombination operators have shown to perform quite poorly on problems with interactions among the variables spread across the solutions (Thierens & =-=Goldberg, 199-=-3; Pelikan, Goldberg, & Cantu-Paz, 1998). Moreover, the hierarchical nature of the optimization process has earned only little attention and it has been assumed that genetic algorithms do this automat... |

8894 |
Probabilistic Reasoning in Intelligent Systems
- Pearl
- 1988
(Show Context)
Citation Context ...hat can be used to make the representation of the model more ecient are discussed and a simple greedy algorithm for network construction is brie y described. 2.1 BAYESIAN NETWORKS A Bayesian network (=-=Pearl, 1988-=-) is a directed acyclic graph with the nodes corresponding to the variables in the modeled data set (in our case, to the positions in solution strings). Mathematically, a Bayesian network encodes a jo... |

2839 |
Genetic algorithms in search, optimization and machine learning
- Goldberg
- 2000
(Show Context)
Citation Context ...evel. The proposed algorithm is shown to scale up subquadratically on all test problems. Empirical results are in agreement with recent theory. 1 INTRODUCTION Genetic algorithms (GAs) (Holland, 1975; =-=Goldberg, 1989-=-) combine short partial solutions to form solutions of higher order. New solutions undergo selection and the process is repeated until the entire solution is formed. However,sxed, problem-independent,... |

1089 | An Analysis of the behavior of a class of genetic adaptive systems”, (Doctoral dissertation - Jong - 1975 |

633 | Genetic algorithms with sharing for multimodal function optimization - Goldberg, Richardson - 1987 |

632 | Genetic algorithms for multiobjective optimization: formulation, discussion and generalization - Fonseca, Fleming - 1993 |

396 | Punctuated equilibria: An alternative to phyletic gradualism - Eldredge, Gould - 1972 |

385 | Messy genetic algorithms: Motivation, analysis, and first results - Goldberg, Korb, et al. - 1989 |

355 | Population-based incremental learning: a method for integrating genetic search based function optimization and competitive learning
- Baluja
- 1994
(Show Context)
Citation Context ...be generated by only preserving the proportions of the values of all variables independently of the context. This is the basic principle of the population based incremental learning (PBIL) algorithm (=-=Baluja, 199-=-4), the compact genetic algorithm (cGA) (Harik et al., 1998), and the univariate marginal distribution algorithm (UMDA) (Muhlenbein, 1997). Since these algorithms take into account only contributions ... |

338 | A survey of optimization by building and using probabilistic models
- Pelikan, Goldberg, et al.
- 1999
(Show Context)
Citation Context ...The remainder of this section describes test problems used in our experiments. 5.1 HIERARCHICALLY DECOMPOSABLE FUNCTIONS Hierarchically decomposable functions (HDFs) (Watson, Hornby, & Pollack, 1998; =-=Pelikan & Goldberg, 20-=-00b) are a subclass of general additively decomposable functions (Pelikan, Goldberg, & Cantu-Paz, 1998). HDFs are dened on multiple levels where the input to each level is based on the solutions found... |

312 | From recombination of genes to the estimation of distributions: I. binary parameters. - Muhlenbein, Paaß - 1996 |

307 |
An investigation of niche and species formation in genetic function optimization,
- Deb, Goldberg
- 1989
(Show Context)
Citation Context ...ed class of problems is called hierarchical traps. Empirical results are in agreement with our recent convergence and population sizing theory. 1 Introduction Genetic algorithms (GAs) (Holland, 1975; =-=Goldberg, 1989-=-) combine short order partial solutions to form solutions of higher order. New solutions undergo selection and the process is repeated until the entire solution is formed. However,sxed, problem-indepe... |

283 | A parameter-less genetic algorithm - Harik, Lobo - 1999 |

271 | Learning Bayesian networks with Local Structure - Friedman, Goldszmidt - 1996 |

262 |
Adaptive Selection Methods for Genetic Algorithms
- Baker
(Show Context)
Citation Context ..., but the avoidance of premature convergence. One could, for instance, inject randomly generated individuals into the current population each now and then (Mauldin, 1984), or control the 6 selection (=-=Baker, 1985-=-) somehow to prevent premature convergence. However, this is not a primary purpose of using niching in our work. Various techniques for niching were also proposed in the area of multiobjective optimiz... |

232 | Linkage learning via probabilistic modeling in the ECGA,
- Harik
- 1999
(Show Context)
Citation Context ... optimization algorithm (Pelikan, Goldberg, & Cantu-Paz, 1998), which uses Bayesian networks to model promising solutions and generate the new ones and the extended compact genetic algorithm (ECGA) (H=-=arik, 1999-=-) which uses the minimum description length metric to construct groups of correlated variables and generates new solutions accordingly. For another method using Bayesian networks, see also Etxeberria ... |

189 | A Bayesian approach to learning Bayesian networks with local structure - Chickering, Heckerman, et al. - 1997 |

154 | MIMIC: Finding Optima by Estimating Probability Densities. - Bonet, Isbell, et al. - 1996 |

144 | Selection in Massively Parallel Genetic Algorithms. - Collins, Jefferson - 2007 |

140 | Multiobjective optimization using the niched pareto genetic algorithm. - Horn, Nafpliotis - 1993 |

131 | Using optimal dependency-trees for combinatorial optimization: Learning the structure of the search space. - Baluja, Davies - 1997 |

121 | Crowding and preselection revisited.” - Mahfoud - 1992 |

121 | The Equation for Response to Selection and Its Use for Prediction. - Mühlenbein - 1997 |

119 | ASPARAGOS: An Asynchronous Parallel Genetic Optimization Strategy - Gorges-Schleuter |

114 | The Bivariate Marginal Distribution Algorithm. In - Pelikan, Muhlenbein - 1999 |

107 | Punctuated equilibria: a parallel genetic algorithm - Cohoon, Hegde, et al. - 1987 |

106 | Schemata, distributions and graphical models in evolutionary optimization. - Muhlenbein, Mahnig, et al. - 1999 |

101 | Linkage Problem, Distribution Estimation, and Bayesian Networks. - Pelikan, Goldberg, et al. - 2000 |

87 | Intelligent Behavior as an Adaptation to the Task Environment - Booker - 1982 |

82 | A naturally occurring niche & species phenomenon: The model and first results - Davidor - 1991 |

71 | Computer-aided gas pipeline operation using genetic algorithms and rule learning, PHD Thesis, - Goldberg - 1983 |

71 | Computer simulations of genetic adaptation: Parallel subcomponent interaction in a multilocus model, Unpublished doctoral dissertation, The - Grosso - 1985 |

67 | Global optimization using bayesian networks,” - Exteberria, Larranaga - 1999 |

61 | Sufficient conditions for deceptive and easy binary functions - Goldberg - 1992 |

51 | Modeling building-block interdependency - Watson, Hornby, et al. - 1998 |

49 | Continuous iterated density estimation evolutionary algorithms within the IDEA framework,
- Bosman, Thierens
- 2000
(Show Context)
Citation Context ...del-building genetic algorithms (PMBGAs) (Pelikan, Goldberg, & Lobo, 2000), estimation of distribution algorithms (EDAs) (Muhlenbein & Paa, 1996), or iterated density estimation algorithms (IDEAs) (Bo=-=sman, 200-=-0). However, estimating a multivariate distribution is not an easy task. There is a trade o between the accuracy of the estimation and its computational cost. To use computationally ecient methods, on... |

46 | Bayesian optimization algorithm, population size, and time to convergence
- Pelikan, Goldberg, et al.
- 2000
(Show Context)
Citation Context ...The remainder of this section describes test problems used in our experiments. 5.1 HIERARCHICALLY DECOMPOSABLE FUNCTIONS Hierarchically decomposable functions (HDFs) (Watson, Hornby, & Pollack, 1998; =-=Pelikan & Goldberg, 20-=-00b) are a subclass of general additively decomposable functions (Pelikan, Goldberg, & Cantu-Paz, 1998). HDFs are dened on multiple levels where the input to each level is based on the solutions found... |

44 | Linkage information processing in distribution estimation algorithms - Bosman, Thierens - 1999 |

42 | Bayesian Optimization Algorithm, Decision Graphs, and Occam’s Razor
- Pelikan, Goldberg, et al.
- 2001
(Show Context)
Citation Context ...The remainder of this section describes test problems used in our experiments. 5.1 HIERARCHICALLY DECOMPOSABLE FUNCTIONS Hierarchically decomposable functions (HDFs) (Watson, Hornby, & Pollack, 1998; =-=Pelikan & Goldberg, 20-=-00b) are a subclass of general additively decomposable functions (Pelikan, Goldberg, & Cantu-Paz, 1998). HDFs are dened on multiple levels where the input to each level is based on the solutions found... |

40 | Finite Markov chain analysis of genetic algorithms with niching,” - Horn - 1993 |

35 | Probabilistic crowding: deterministic crowding with probabilistic replacement - Mengshoel, Goldberg - 1999 |

34 | Hierarchical problem solving by the bayesian optimizationalgorithm. InProceedingsoftheGeneticandEvolutionaryComputation Conference(GECCO
- Pelikan, Goldberg
(Show Context)
Citation Context ...der of this section describes test problems used in our experiments. 5.1 HIERARCHICALLY DECOMPOSABLE FUNCTIONS Hierarchically decomposable functions (HDFs) (Watson, Hornby, & Pollack, 1998; Pelikan & =-=Goldberg, 20-=-00b) are a subclass of general additively decomposable functions (Pelikan, Goldberg, & Cantu-Paz, 1998). HDFs are dened on multiple levels where the input to each level is based on the solutions found... |

34 |
Adaptation in Natural and Arti Systems, Ann Arbor
- Holland
- 1975
(Show Context)
Citation Context ...ems. The proposed class of problems is called hierarchical traps. Empirical results are in agreement with our recent convergence and population sizing theory. 1 Introduction Genetic algorithms (GAs) (=-=Holland, 1975-=-; Goldberg, 1989) combine short order partial solutions to form solutions of higher order. New solutions undergo selection and the process is repeated until the entire solution is formed. However,sxed... |

34 | Evolution in time and space—the parallel genetic algorithm,” in Foundation of Genetic Algorithms, - Muhlenbein - 1992 |

30 | Multimodal deceptive functions - Deb - 1992 |

29 |
Adaptation in natural and arti cial systems. Ann Arbor
- Holland
(Show Context)
Citation Context ...ptive on each level. The proposed algorithm is shown to scale up subquadratically on all test problems. Empirical results are in agreement with recent theory. 1 INTRODUCTION Genetic algorithms (GAs) (=-=Holland, 1975-=-; Goldberg, 1989) combine short partial solutions to form solutions of higher order. New solutions undergo selection and the process is repeated until the entire solution is formed. However,sxed, prob... |

29 | Maintaining diversity in genetic search
- Mauldin
- 1984
(Show Context)
Citation Context ...iple solutions or alternative search regions, but the avoidance of premature convergence. One could, for instance, inject randomly generated individuals into the current population each now and then (=-=Mauldin, 1984-=-), or control the 6 selection (Baker, 1985) somehow to prevent premature convergence. However, this is not a primary purpose of using niching in our work. Various techniques for niching were also prop... |

27 | Genetic Algorithms, Clustering, and the Breaking of Symmetry
- Pelikan, Goldberg
- 2000
(Show Context)
Citation Context ...der of this section describes test problems used in our experiments. 5.1 HIERARCHICALLY DECOMPOSABLE FUNCTIONS Hierarchically decomposable functions (HDFs) (Watson, Hornby, & Pollack, 1998; Pelikan & =-=Goldberg, 20-=-00b) are a subclass of general additively decomposable functions (Pelikan, Goldberg, & Cantu-Paz, 1998). HDFs are dened on multiple levels where the input to each level is based on the solutions found... |

26 | Analysis of recombinative algorithms on a non-separable building-block problem », Foundations of Genetic Algorithms - Watson |

24 |
The Design of Innovation: Lessons from Genetic Algorithms, Lessons for the Real World
- Goldberg
- 1998
(Show Context)
Citation Context ...ibution and interpretation functions. The hierarchical if-and-only-if (HIFF) function (Watson et al., 1998) uses the \if and only if" function on each level. More dicult functions have been propo=-=sed (Goldberg, 1997-=-; Goldberg, 1998; Pelikan & Goldberg, 2000b), where functions deceive the algorithms to a local optimum on each level. Only at the top level it becomes clear which optimum is the global one. In this p... |

24 | Genetic invariance: A new paradigm for genetic algorithm design
- Culberson
- 1992
(Show Context)
Citation Context ...tic crowding method. In probabilistic crowding, the winner of the parent-ospring tournament is chosen by using the probability proportional to thestness. The gene invariant genetic algorithm (GIGA) (C=-=ulberson, 1992-=-) maintains constant univariate frequencies of all values on all positions. For instance, in case of binary strings and uniformly generated initial population, at any point in the run there will be ab... |

23 | Finding multiple solutions in problems of bounded difficulty - Harik - 1994 |

13 | Experimental study of speciation in ecological niche theory using genetic algorithms - Perry - 1984 |

12 | Adaptive search using simulated evolution.” Doctoral dissertation - Cavicchio - 1970 |

10 | Beyond Bayesian networks: Similarity networks and Bayesian multinets. - Geiger, Heckerman - 1996 |

10 | Some experiments in machine learning using vector evaluated genetic algorithms. Doctoral dissertation - Schaer - 1984 |

5 |
15). Four keys to understanding building-block difficulty
- Goldberg
- 1998
(Show Context)
Citation Context ...rpretation functions. The hierarchical if-and-only-if (HIFF) function (Watson et al., 1998) uses the \if and only if" function on each level. More dicult functions have been proposed (Goldberg, 1=-=997; Goldberg, 1998-=-; Pelikan & Goldberg, 2000b), where functions deceive the algorithms to a local optimum on each level. Only at the top level it becomes clear which optimum is the global one. In this paper, only simpl... |

2 | Arti Genetic Adaptation in Computer Control Systems - Hollstein - 1971 |

1 |
A comparative study of scoring metrics in the Bayesian optimization algorithm: Minimum description length and Bayesian-Dirichlet. Unpublished technical report
- Goldberg
- 2000
(Show Context)
Citation Context ...ollins and Jeerson (1991), Davidor (1991), Muhlenbein (1991), and others. The use of spatial separation was also studied in context of the probabilistic model-building genetic algorithms (Pelikan & Go=-=ldberg, 2000-=-b) as a niching and diversity preservation tool. Pelikan and Goldberg divided the population of selected parents in each generation into a number of clusters. A mixture of Gaussians was used to separa... |