Niching methods for genetic algorithms
Niching methods for genetic algorithms
The Design of Innovation: Lessons from and for Competent Genetic Algorithms
The Design of Innovation: Lessons from and for Competent Genetic Algorithms
When Will a Genetic Algorithm Outperform Hill Climbing?
Proceedings of the 5th International Conference on Genetic Algorithms
Modeling Building-Block Interdependency
PPSN V Proceedings of the 5th International Conference on Parallel Problem Solving from Nature
Combining competent crossover and mutation operators: a probabilistic model building approach
GECCO '05 Proceedings of the 7th annual conference on Genetic and evolutionary computation
On the complexity of hierarchical problem solving
GECCO '05 Proceedings of the 7th annual conference on Genetic and evolutionary computation
Linkage Problem, Distribution Estimation, and Bayesian Networks
Evolutionary Computation
Conquering hierarchical difficulty by explicit chunking: substructural chromosome compression
Proceedings of the 8th annual conference on Genetic and evolutionary computation
Learning building block structure from crossover failure
Proceedings of the 9th annual conference on Genetic and evolutionary computation
The gambler's ruin problem, genetic algorithms, and the sizing of populations
Evolutionary Computation
Scalability problems of simple genetic algorithms
Evolutionary Computation
Towards memoryless model building
Proceedings of the 10th annual conference companion on Genetic and evolutionary computation
Large-Scale Optimization of Non-separable Building-Block Problems
Proceedings of the 10th international conference on Parallel Problem Solving from Nature: PPSN X
Cooperation in the context of sustainable search
CEC'09 Proceedings of the Eleventh conference on Congress on Evolutionary Computation
Hi-index | 0.00 |
A major challenge in the field of metaheuristics is to find ways to increase the size of problems that can be addressed reliably. Scalability of probabilistic model building methods, capable to rendering difficult, large problems feasible by identifying dependencies, have been previously explored but investigations had mainly concerned problems where efficient solving is possible with the exploitation of low order dependencies. This is due to the initial-supply population sizing, where the number of samples is lower bounded by the exponential of the order of dependencies covered by the probabilistic model. With an exponentially growing population, the impact of the model building on the overall complexity, can easily exceed the bound for the number of evaluations. In this paper we present a competent methodology, capable of efficiently detecting and combining large modules, even in the case of unfavorable genetic linkage and no intra-block fitness gradient to guide the search or deceptiveness. This is achieved by investing the function evaluations in a model based local-search with strong exploratory power and restricting the model building to a relatively small number of semi-converged samples.