Degree of population diversity - a perspective on premature convergence in genetic algorithms and its Markov chain analysis

LeungYee,GaoYong,XuZong-Ben

1997-01-01

IEEE Transactions on Neural Networks

Abstract:In this paper, a concept of degree of population diversity is introduced to quantitatively characterize and theoretically analyze the problem of premature convergence in genetic algorithms (GAs) wi...

What problem does this paper attempt to address?

Quantitative Analysis and Prevention of Genetic Algorithm Premature Convergence

WANG Min-le,GAO Xiao-guang,LIU Gang

DOI: https://doi.org/10.3321/j.issn:1001-506x.2006.08.037

2006-01-01

Abstract:To aim at the problem of genetic algorithm premature convergence,the new definition of premature convergence is given,and a new fuzzy index for measuring population maturity degree is presented on the basis of fuzzy system theory.Finally,the method of adaptively adjusting crossover probability and mutation probability with population maturity degree is proposed to prevent premature convergence,and its validity is verified through a simulation experiment.
Characteristic Analysis and Prevention on Premature Convergence in Genetic Algorithms

Zongben Xu,Yong Gao

DOI: https://doi.org/10.1007/bf02916943

1997-01-01

Abstract:The identification and characteristics of premature convergence in genetic algorithms (GAs) are investigated. Through a detailed quantitative analysis on the search capability and the degree of population diversity, the cause of premature convergence in GAs is recognized, and attributed to the maturation effect of the GAs: The minimum schema deduced from current population, which is the largest search space of a GA, converges to a homogeneous population in probability 1 (so the search capability of the GA decreases and premature convergence occurs). It is shown that, as quantitative features of the maturation effect, the degree of population diversity converges to zero with probability 1, and the tendency for premature convergence is inversely proportional to the population size and directly proportional to the variance of the fitness ratio of zero allele at any gene position of the current population. Based on the theoretical analysis, several strategies for preventing premature convergence are suggested. A specific GA formulation that converges assuredly to the global optimum without appearance of premature convergence is proposed.
A Novel Strategy to Control Population Diversity and Convergence for Genetic Algorithm.

Dongyang Li,Weian Guo,Yanfen Mao,Lei Wang,Qidi Wu

DOI: https://doi.org/10.1007/978-3-319-61824-1_39

2017-01-01

Abstract:Genetic algorithm (GA), an efficient evolutionary algorithm inspired from the science of genetics, attracts the worldwide attention for several decades. This paper tries to strengthen the search ability of the population in GA in the way of improving the distance among individuals by introducing a new solution updating strategy based on the theory of Cooperative Game. The simulation is done using fourteen benchmark functions, and the results demonstrate that this modified genetic algorithm works efficiently.
On the Mechanism and Convergence of Genetic Algorithm

Zhi-gang YU,Shen-min SONG,Guang-ren DUAN

DOI: https://doi.org/10.3321/j.issn:1001-0920.2005.09.003

2005-01-01

Abstract:A new approach based on decomposition of solution space according to difference of genotype is presented to analyze quantitatively the evolutionary course of populations of GA. The analysis yields new insight into the properties of the three phases, mutation, crossover, and fitness selection of a genetic algorithm by representing them as acting on the solution space. The probability distribute of populations over solution space can be calculated by this approach. Theoretically, the capability of finding the global optimum is proved, and a necessary and sufficient condition is obtained namely, the conclusion of building block hypothesis is proved. Meanwhile, under crossover or mutation operator applied to static optimization problems, the probability distribution of populations over solution space can be calculated by means of a finite Markov chain model, and the probability of GA converging to the global optimum also can be estimated. Finally, the reasons about GA deceptive problem and premature convergence are discussed.
A Novel Genetic Algorithm Based on Individual and Gene Diversity Maintaining and Its Simulation

Xiaojun Xing,Qiuling Jia,Zhigang Ling,Dongli Yuan

DOI: https://doi.org/10.1109/ical.2007.4339049

2007-01-01

Abstract:In order to overcome premature convergence in SGA, a novel adaptive genetic algorithm based on diversity maintaining is proposed. First, variance of all individuals' fitness is used to measure individual diversity in a population and to adjust crossover probability adaptively. Second, to restrain the lack of effective genes in certain loci, mutation probabilities of all alleles in each locus vary adaptively depending on gene diversity in corresponding locus. We compare the performance of the DMAGA with that of the simple genetic algorithm (SGA) and AGA in optimizing several complex functions. The simulation result shows that the novel GA can obtain higher precision solution and avoid local optima.
Study on Premature Convergence in Assembly Sequence Genetic Planning

Yu Guo,Xibo Lin,Weili Hu

DOI: https://doi.org/10.1109/wcica.2006.1714331

2006-01-01

Abstract:Focusing on the problem of premature convergence existed in genetic algorithm for assembly sequence planning, the reasons were analyzed from genetic operations. It was pointed out that keeping population diversity and gene schema were the keys to avoid premature convergence. Checking up chromosomes and regulating them with the same mechanism will alter genetic information or destroy schema structure. They will also aggravate premature convergence. The influences of variation probability on population diversity and evolution progress were discussed in detail. To alleviate premature convergence, an extend operator was presented. It can enhance population diversity and local search ability by replacing the bad individuals with the son individuals produced by good individuals dithering. Supplantation operator was adopted to improve the evolution ability. It will check up difference between chromosomes of individuals and add a punish term to the little difference individual at initial evolutions. In order to prevent loss of available gene, gene repair strategy was presented to ill genes. Simulation results for an assembly sequence planning show validity of the presented algorithm.
Study on Diversity of Population in Genetic Algorithms

Dai Guan-zhong

1998-01-01

Abstract:In this paper,variance and entropy are proposed as measures of diversity of pupulation in genetic algorithms. The influence which selection,crossover,and mutation act upon variance and entropy is analyzed. In addition,the different influence between binary and decimal encoding mechanism is also shown.
A Novel Genetic Algorithm Based on Diversity Maintaining and Its Simulation

邢小军,闫建国

DOI: https://doi.org/10.3969/j.issn.1006-9348.2010.04.050

2010-01-01

Abstract:A novel adaptive genetic algorithm based on diversity maintaining is proposed to overcome premature convergence in GAs.Firstly,variance of all individuals' fitness in population is used in measuring individual diversity,and used in the adaptive adjustment of crossover probability.Secondly,by means of computing gene diversity,all alleles' mutation probabilities in each locus are changed adaptively to restrain the lack of effective genes.The simulation result shows that the algorithm can obtain high precision solution and avoid local optima.
A Method to Improve Performance of Genetic Algorithm

Chen Debao,Zhao Chunxia

DOI: https://doi.org/10.3969/j.issn.0465-7942.2005.06.016

2005-01-01

Abstract:Considering diversity of population, a function is presented to describe diversity.Based on the value of diversity function,part-degenerate operator is use to improve the diversity of the new population,and premature is prevented.Imitating crossover principle of biology, according to the distance of the individuals which are used to crossover,they are divide into two parts,and the global genetic is used in the same part,the local genetic is used in the different part,that is to say,the genes in different individual in different part are absorbed in a individual.The improved individual can be product quickly.The simulation results demonstrate that the given method is more efficient to improve convergence speed than the standard genetic algorithm and possibility of premature is deduced.
Genetic Algorithms with Fitness- & Diversity-guided Adaptive Operating Probabilities and Analyses of its Convergence

Li Meiyi,Cai Zixing,Sun Guoyun

2004-01-01

Abstract:The paper has analyzed global convergence properties of adaptive genetic algorithms combining adaptive probabilities of crossover and mutation with diversity-guided crossover and mutation. By means of homogeneous finite Markov chains, it is proved that AGAD, which is present in this paper, and GAD (genetic algorithms with diversity-guided mutation) maintaining the best solution converge to the global optimum, which is the main contributions of this paper. The performance of AGA(adaptive genetic algorithms with adaptive probabilities of crossover and mutation), GAD and AGAD in optimizing several unimodal and multimodal functions has been compared. For multimodal functions, the AGAD converges to the global optimum for fewer generations than AGA and GAD, and it hardly has premature convergence. Key Diversity-guided mutation; Adaptive genetic algorithm; Markov chain analysis; Global convergence.
2 . 2 Population Diversity and Selective Pressure of SGA

Xinchao Zhao,Xiao-Shan Gao KLMM

2007-01-01

Abstract:Based on certain phenomena from the human society and nature, we propose a binary affinity genetic algorithm (aGA) by adopting the following strategies: the population is adaptively updated to avoid stagnation; the newly generated individuals are guaranteed to survive for certain number of generations in order for them to have enough time to develop their good genes; new individuals and the old ones are balanced to take both of their advantages. In order to quantitatively analyze the selective pressure, the concept of selection degree and a simple linear control equation are introduced. We can maintain the diversity of the evolutionary population by controlling the value of the selection degree. The performance of aGA is further enhanced by incorporating local search strategies.
The Unified Method Analyzing Convergence of Genetic Algorithms

Gongde Guo,Shuanghe Yu

2001-01-01

Abstract:This paper models the homogeneous finite Markov chain of the best individuals in the populations. Based on the limit distribution of transition probability that any state on the chain transfers into the state subset containing all global optima, a precise definition of the global convergence of genetic algorithms is presented regardless of the topologic structures of search landscapes. Two unified criterion theorems judging the global convergence are proposed and proved strictly, which are independent of encoding schemes and selection mechanisms. The results of analyzing the convergence of different genetic algorithms illustrate that the unified criterion theorems are generally practical and convenient.
A Novel Genetic Algorithm Preventing Premature Convergence by Chaos Operator

Liu Juan,Cai Zi-xing,Liu Jian-qin

DOI: https://doi.org/10.1007/s11771-000-0042-8

2000-01-01

Journal of Central South University of Technology

Abstract:This paper proposes a novel genetic algorithm containing chaos operator based on the analysis of population diversity and premature convergence within the framework of Markov chain. This algorithm increases the population size dynamically so as to restore the population diversity and prevent premature convergence effectively. Its validity and superiority are illustrated by two applications.
Improving the Global Convergence of the Genetic Algorithm

何大阔,王福利

DOI: https://doi.org/10.3321/j.issn:1005-3026.2003.06.001

2003-01-01

Abstract:Based on the analysis of the reason of the early convergence of the genetic algo rithm, it was shown that the early convergence of the genetic algorithm is main ly concerned with the distribution of the solutions to the problem, the distribu tion of the individuals of the population and the application of the genetic ope rators. The key to improve the performance of the global convergence of the gen etic algorithms was believed to apply the algorithms to deal with the diversity of population scientifically and recognize the contribution of the individuals t o the global convergence. Several kinds of the individuals, which have much to d o with the performance of the global convergence, have been proposed and combine d with the niche evolution sharing function. A genetic algorithm is suggested to improve the global convergence of the genetic algorithms and to obtain the glob al optimal solution. Simulation results show that the new algorithm has good per formance on the global convergence.
Metric For Measuring The Degree Of Population Divergence Of Gene Expression Programming

Zeng Sanyou,Yan Jingfeng,Zhu Qinglu,Yao Shuzhen,Kang Lishan

2006-01-01

Abstract:Gene Expression Programming (GEP) is a new evolutionary algorithm which can be employed to construct non-linear model for complex data including data from earth science. The diversity of population has impact on the performance of GEP. in this paper, a technique, which measures the diversity of population of GEP, is proposed. Experiment shows that this technique is reasonable. Therefore, it is an optional technique for measuring the diversity of the population in GEP.
Analysing Equilibrium States for Population Diversity

Johannes Lengler,Andre Opris,Dirk Sudholt

DOI: https://doi.org/10.1007/s00453-024-01226-3

IF: 0.909

2024-04-17

Algorithmica

Abstract:Population diversity is crucial in evolutionary algorithms as it helps with global exploration and facilitates the use of crossover. Despite many runtime analyses showing advantages of population diversity, we have no clear picture of how diversity evolves over time. We study how the population diversity of algorithms, measured by the sum of pairwise Hamming distances, evolves in a fitness-neutral environment. We give an exact formula for the drift of population diversity and show that it is driven towards an equilibrium state. Moreover, we bound the expected time for getting close to the equilibrium state. We find that these dynamics, including the location of the equilibrium, are unaffected by surprisingly many algorithmic choices. All unbiased mutation operators with the same expected number of bit flips have the same effect on the expected diversity. Many crossover operators have no effect at all, including all binary unbiased, respectful operators. We review crossover operators from the literature and identify crossovers that are neutral towards the evolution of diversity and crossovers that are not.

computer science, software engineering,mathematics, applied
Study on Population Diversity of TSP in Genetic Algorithms

LI Jun-hua,LI Ming,YUAN Li-hua

2008-01-01

Abstract:Genetic algorithms make it possible to explore a far greater range of potential solutions to a problem than do conventional programs,So it is a widely used optimization method.However,low diversity is often mentioned as the reason for poor performance in genetic algorithms.To find a solution to TSP with GA,this paper study on population diversity in genetic algorithms.This paper examines measures of diversity in genetic algorithms.Results show the phenotype diversity definitions is not good at reflecting the diversity or diversity's change of TSP population.The proposed method build the connect matrix of all individuals,1and then compute the entropy of the matrix to show the diversity or the diversity's change.Results indicate that the novel method not only efficient but reduce the time complexity.
A Real Coding Parallel Genetic Algorithm Based on Diversity of Population

刘胜,李高云,孙天英

DOI: https://doi.org/10.3969/j.issn.1673-4785.2008.05.007

2008-01-01

Abstract:In order to improve the convergence performance of genetic algorithms,a function measuring population diversity on the basis of the degree of individual adaptability was needed.This measuring function must reflect the evolutionary stage of the genetic algorithm and forecast premature trends appropriately.We designed a migration operator and crossover operator based on diversity of population functions,and made dynamic adjustments on crossover and mutation probabilities.This structured a parallel genetic algorithm with real coding and a multi-layer migration operator.Comparative experiments were made on benchmark functions.The results showed that this algorithm is obviously superior to other genetic algorithms in overcoming problems such as prematurity and slow convergence.
Adaptive genetic algorithms based on generational gap information

Yu Guo,Xibo Lin,Weili Hu

2006-01-01

Abstract:2 Start Computer Science & Technology Co. Ltd , Fuzhou 350002 , China) Abstract : Focusing on the problems existed in some adaptive genetic algorithms that cannot take account of population features and are difficult to converge to optimal solution steadily , the reasons of stopping evolution or degeneration are analyzed in view of fitness average and population diversity. According to the fitness average and population diversity , a novel adaptive genetic algorithms based on generational gap information is proposed. By using the information of fitness difference and diversity difference between two neighbour generations , a strategy for genetic probability adjusting is designed. By this adaptive genetic algorithm , good diversity can be remained and premature can be avoided effectively. Convergence of the algorithm is proved. Simulation results show that the proposed algorithms can make the population remain good evolutivity and convergence.
Adaptive and Parallel Genetic Algorithm Based on Solving Premature Convergence

ZHOU Yuanhui,LU Yuchang,SHI Chunyi

DOI: https://doi.org/10.3321/j.issn:1000-0054.1998.03.022

1998-01-01

Abstract:Premature convergence is still the preeminent problem in genetic algorithms (GAs). Some excellent individuals or genes are lost due to premature convergence, which causes local optimum. This paper presents an improved genetic algorithm based on multiple populations evolution and adaptive parameter adjusting. Experimental results show that this method alleviates the problem of premature convergence and improves the efficiency of searching.

Degree of population diversity - a perspective on premature convergence in genetic algorithms and its Markov chain analysis

Quantitative Analysis and Prevention of Genetic Algorithm Premature Convergence

Characteristic Analysis and Prevention on Premature Convergence in Genetic Algorithms

A Novel Strategy to Control Population Diversity and Convergence for Genetic Algorithm.

On the Mechanism and Convergence of Genetic Algorithm

A Novel Genetic Algorithm Based on Individual and Gene Diversity Maintaining and Its Simulation

Study on Premature Convergence in Assembly Sequence Genetic Planning

Study on Diversity of Population in Genetic Algorithms

A Novel Genetic Algorithm Based on Diversity Maintaining and Its Simulation

A Method to Improve Performance of Genetic Algorithm

Genetic Algorithms with Fitness- & Diversity-guided Adaptive Operating Probabilities and Analyses of its Convergence

2 . 2 Population Diversity and Selective Pressure of SGA

The Unified Method Analyzing Convergence of Genetic Algorithms

A Novel Genetic Algorithm Preventing Premature Convergence by Chaos Operator

Improving the Global Convergence of the Genetic Algorithm

Metric For Measuring The Degree Of Population Divergence Of Gene Expression Programming

Analysing Equilibrium States for Population Diversity

Study on Population Diversity of TSP in Genetic Algorithms

A Real Coding Parallel Genetic Algorithm Based on Diversity of Population

Adaptive genetic algorithms based on generational gap information

Adaptive and Parallel Genetic Algorithm Based on Solving Premature Convergence