Searching for optima in non-stationary environments

Trojanowski, Krzysztof; Michalewicz, Zbigniew

doi:10.1109/cec.1999.785498

Cited by 93 publications

(65 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The second type of dynamic problem generators starts from a predefined fitness landscape, usually constructed in n-dimensional real space [5], [16], [23], [28]. This stationary landscape is composed of a number of component landscapes (e.g., cones), each of which can change independently.…”

Section: Review Of Existing Generatorsmentioning

confidence: 99%

Experimental study on population-based incremental learning algorithms for dynamic optimization problems

2005

View full text Add to dashboard Cite

Evolutionary algorithms have been widely used for stationary optimization problems. However, the environments of real world problems are often dynamic. This seriously challenges traditional evolutionary algorithms. In this paper, the application of PopulationBased Incremental Learning (PBIL) algorithms, a class of evolutionary algorithms, for dynamic problems is investigated. Inspired by the complementarity mechanism in nature a Dual PBIL is proposed, which operates on two probability vectors that are dual to each other with respect to the central point in the genotype space. A diversity maintaining technique of combining the central probability vector into PBIL is also proposed to improve PBIL's adaptability in dynamic environments. In this paper, a new dynamic problem generator that can create required dynamics from any binary-encoded stationary problem is also formalized. Using this generator, a series of dynamic problems were systematically constructed from several benchmark stationary problems and an experimental study was carried out to compare the performance of several PBIL algorithms and two variants of standard genetic algorithm. Based on the experimental results, we carried out algorithm performance analysis regarding the weakness and strength of studied PBIL algorithms and identified several potential improvements to PBIL for dynamic optimization problems.

show abstract

Section: Review Of Existing Generatorsmentioning

confidence: 99%

Experimental study on population-based incremental learning algorithms for dynamic optimization problems

2005

View full text Add to dashboard Cite

show abstract

“…Implicit memory schemes use redundant encodings, e.g., diploid genotype [10,12,18], to store information for EAs to exploit during the run. In contrast, explicit memory uses precise representations but splits an extra storage space to explicitly store information from a current generation and reuse it later [4,11,17].…”

Section: Memory Schemes For Dynamic Environmentsmentioning

confidence: 99%

Memory Based on Abstraction for Dynamic Fitness Functions

Richter

Yang

2008

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. This paper proposes a memory scheme based on abstraction for evolutionary algorithms to address dynamic optimization problems. In this memory scheme, the memory does not store good solutions as themselves but as their abstraction, i.e., their approximate location in the search space. When the environment changes, the stored abstraction information is extracted to generate new individuals into the population. Experiments are carried out to validate the abstraction based memory scheme. The results show the efficiency of the abstraction based memory scheme for evolutionary algorithms in dynamic environments.

show abstract

“…As to the memory organization, there exist two mechanisms: local mechanism where the memory is individual-oriented and global mechanism where the memory is population-oriented. Trojanowski and Michalewicz [22] introduced a local memory approach, where for each individual the memory stores a number of its ancestors. When the environment changes, the current individual and its ancestors are re-evaluated and compete together with the best becoming the active individual while the others stored in the memory.…”

Section: Explicit Memory For Eas In Dynamic Environmentsmentioning

confidence: 99%

“…For example, for the direct memory scheme the whole memory individuals may enter the new population as in [13] or compete with the population individuals for the new population as in [3], while for the associative memory scheme only the associated memory individual(s) [21] or new individuals created by the associated environmental information [26,30] may enter the new population. And for the local memory organization scheme the best ancestor of an active individual competes with it to become active in the population [22], while for the global memory scheme the best memory individual(s) may compete with all individuals in the population.…”

Section: Explicit Memory For Eas In Dynamic Environmentsmentioning

confidence: 99%

“…Typical examples are genetic algorithms (GAs) based on multiploidy representations [8,9,12,20], structured encoding [7], or dualism mechanisms [24,29]. Explicit memory schemes use precise representations but split an extra memory space to explicitly store useful information, e.g., good solutions [2,13,15,22] and/or environmental information [21,25], from the current generation for reuse in later generations or environments.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Explicit Memory Schemes for Evolutionary Algorithms in Dynamic Environments

Yang

2007

Studies in Computational Intelligence

View full text Add to dashboard Cite

Summary. Problem optimization in dynamic environments has atrracted a growing interest from the evolutionary computation community in reccent years due to its importance in real world optimization problems. Several approaches have been developed to enhance the performance of evolutionary algorithms for dynamic optimization problems, of which the memory scheme is a major one. This chapter investigates the application of explicit memory schemes for evolutionary algorithms in dynamic environments. Two kinds of explicit memory schemes: direct memory and associative memory, are studied within two classes of evolutionary algorithms: genetic algorithms and univariate marginal distribution algorithms for dynamic optimization problems. Based on a series of systematically constructed dynamic test environments, experiments are carried out to investigate these explicit memory schemes and the performance of direct and associative memory schemes are campared and analysed. The experimental results show the efficiency of the memory schemes for evolutionary algorithms in dynamic environments, especially when the environment changes cyclically. The experimental results also indicate that the effect of the memory schemes depends not only on the dynamic problems and dynamic environments but also on the evolutionary algorithm used.

show abstract

Searching for optima in non-stationary environments

Cited by 93 publications

References 8 publications

Experimental study on population-based incremental learning algorithms for dynamic optimization problems

Experimental study on population-based incremental learning algorithms for dynamic optimization problems

Memory Based on Abstraction for Dynamic Fitness Functions

Explicit Memory Schemes for Evolutionary Algorithms in Dynamic Environments

Contact Info

Product

Resources

About