Subpopulation Diversity Based Selecting Migration Moment in Distributed Evolutionary Algorithms

01/05/2017
by   Chengjun Li, et al.
University of Technology Sydney
0

In distributed evolutionary algorithms, migration interval is used to decide migration moments. Nevertheless, migration moments predetermined by intervals cannot match the dynamic situation of evolution. In this paper, a scheme of setting the success rate of migration based on subpopulation diversity at each interval is proposed. With the scheme, migration still occurs at intervals, but the probability of immigrants entering the target subpopulation will be determined by the diversity of this subpopulation according to a proposed formula. An analysis shows that the time consumption of our scheme is acceptable. In our experiments, the basement of parallelism is an evolutionary algorithm for the traveling salesman problem. Under different value combinations of parameters for the formula, outcomes for eight benchmark instances of the distributed evolutionary algorithm with the proposed scheme are compared with those of a traditional one, respectively. Results show that the distributed evolutionary algorithm based on our scheme has a significant advantage on solutions especially for high difficulty instances. Moreover, it can be seen that the algorithm with the scheme has the most outstanding performance under three value combinations of above-mentioned parameters for the formula.

READ FULL TEXT VIEW PDF
POST COMMENT

Comments

There are no comments yet.

Authors

page 1

page 2

page 3

page 4

12/18/2013

Comparative analysis of evolutionary algorithms for image enhancement

Evolutionary algorithms are metaheuristic techniques that derive inspira...
09/08/2011

Analysis of Speedups in Parallel Evolutionary Algorithms for Combinatorial Optimization

Evolutionary algorithms are popular heuristics for solving various combi...
05/04/2018

An Operating System Level Data Migration Scheme in Hybrid DRAM-NVM Memory Architecture

With the emergence of Non-Volatile Memories (NVMs) and their shortcoming...
08/28/2021

What can phylogenetic metrics tell us about useful diversity in evolutionary algorithms?

It is generally accepted that "diversity" is associated with success in ...
01/07/2016

NodIO, a JavaScript framework for volunteer-based evolutionary algorithms : first results

JavaScript is an interpreted language mainly known for its inclusion in ...
06/04/2018

Ring Migration Topology Helps Bypassing Local Optima

Running several evolutionary algorithms in parallel and occasionally exc...
02/05/2020

Exploring Maximum Entropy Distributions with Evolutionary Algorithms

This paper shows how to evolve numerically the maximum entropy probabili...
This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

I Introduction

As stochastic search methods based on population, evolutionary algorithms (EAs) have been applied successfully in many fields [1, 2]. Nevertheless, for nontrivial problems, EAs tend to premature convergence. Fortunately, using a parallel EA (PEA) often leads to a superior numerical performance because EAs are naturally prone to parallelism [1]. Among types of PEAs, distributed EAs (DEAs), which can be readily implemented in distributed memory MIMD computers  [1], are most popular [1, 3].

In a DEA, a large population is divided into a number of subpopulations. The parallel operator, migration, exchanges individuals between subpopulations at intervals [4]. As a result, global convergence behavior is improved. The first idea of DEA can be traced back to [5]. So far, many papers focus on DEAs. In recent years, the application range of DEA is wider and wider for more and more suitable parallel computing environments being available. For instance, in [6, 7, 8, 9, 10], there are DEAs for problems of different fields. On such a background, it is meaningful to improve migration operator for enhancing DEAs. As one way for this purpose, diversity based migration has been studied in some papers (e.g. [11, 12, 13, 14]).

The main motivation of this paper is as follows. According to literature, diversity based migration do help DEAs to get better solutions. However, existing related studies only focus on selecting fit individuals for migration. In fact, diversity based migration may also be realized by selecting migration moments, which is determined only by the parameter, migration interval in traditional DEAs. Nevertheless, such a topic is rarely seen in literature.

A scheme of setting the success rate of migration based on subpopulation diversity at each interval is proposed in this paper for DEAs. The idea behind it is selecting migration moments based on not only interval but also subpopulation diversity. More precisely, migration still occurs at intervals, but the probability of immigrants entering the target subpopulation, , will be decided by the diversity of this subpopulation, , according to a formula building the relationship of them. In detail, at migration intervals, is computed first. Then, is calculated according to the formula. At the probability, , immigrants enter the subpopulation .

An experiment is conducted on eight instances of the Travelling Salesman Problem (TSP) from [15] to compare the outcomes of the DEA with the proposed scheme and those of a traditional DEA. The only difference between them is that the proposed scheme is used only in the former. For the formula in the scheme, two parameters, and , are required. In our experiment, the common parameters in both algorithms are set the same value. Moreover, the algorithm based on the scheme is tested under nine value combinations of and . Results show that the DEA based on our scheme can get better and more stable solutions than the traditional one. Further, under some value combinations of and , the advantage of the algorithm with our scheme is much more remarkable.

The rest of this paper is organized as follows. Related researches are described in Section 2. In Section 3, the proposed scheme is presented. Then, experimental results are shown and analyzed in Section 4. Finally, a conclusion and a prospect are dealt with in Section 5.

Ii Related Researches

Ii-a Distributed Evolutionary Algorithms

A DEA can be considered as the upgrade of a EA to enhance solving ability. In [4], it is expressed as the Algorithm 1 whose italic part is the migration operation.

1:  Generate at random a population, , of individuals
2:  Divide into , …, subpopulations
3:  Define a neighborhood structure for , , …,
4:  repeat {Execute in parallel the next steps for , , …, }
5:     Apply, during generations, stochastic operators of EA
6:     Send individuals to neighboring subpopulations
7:     Receive individuals from neighboring subpopulations
8:  until the stop criterion is not fulfilled
Algorithm 1 Traditional DEA

Main parallel parameters in DEAs are introduced as below:

  • Migration topology (“neighborhood structure”in Algorithm 1), it describes which subpopulations send individuals to which subpopulations [16].

  • Migration strategy, it defines how to select emigrants and choice replaced individuals in each subpopulation.

  • Migration rate (“r”in Algorithm 1), it specifies the quantity of individuals which emigrate from each subpopulation in one migration round.

  • Migration interval (“i”in Algorithm 1), it is the every certain number of generations between a migration round and the next one [16].

  • Subpopulation size, it is the quantity of individuals in each subpopulation.

  • Quantity of subpopulation (“n”in Algorithm 1), it is the number of subpopulations.

On one hand, semi-isolation can maintain difference among subpopulations. On the other hand, for subpopulations, at intervals, migration is expected to afford individuals whose fitness are close to that of local ones but have some different building blocks [17]. After migration, immigrants participating in variation operations can help to resist premature convergence if they do have some new building blocks. In this way, stagnation is postponed.

Ii-B Diversity Based Migration in DEAs

Diversity refers to differences among individuals, which can be measured at the genotype or phenotype levels [18]. Thus, diversity of a colony shows the its convergence degree. Though diversity measures at the latter level often demands much less time consumption, they are not as accurate as ones at the former one because the difference on fitness is not equivalent to that on chromosome structure.

Some instances on diversity based migration from literature are listed as below. Their diversity measures are all at the genotype level. In [11], a weighted sum of fitness and diversity measure is used in a DEA for selecting emigrants and replaced individuals. Results showed a reduction in the time taken to find the optimal solution across a range of benchmark problems. In the DEA presented by [12], the subpopulation representative what is an individual having the lowest average of distance to other all individuals in this subpopulation is found by calculation. Then, it is selected to be the emigrant. Furthermore, if migration rate is set more than one, some other individuals are chosen as emigrants based on the distance from it. This strategy has been tested on a number of problems and has consistently outperformed the standard migration strategy. The scheme in [13] only allows different enough individuals migrate to a subpopulation in order to keeping diversity of this subpopulation. To realize this scheme, a method for evaluating the similarity of two subpopulations is presented. More recently, multikulti methods, are proposed in [14] for selecting individuals in source subpopulation different enough to ones in target subpopulation . In order to do this, information on the composition of source subpopulation is required when emigrants are selected. Different ways of providing this information in a concise manner are considered. The results of experiments prove the usefulness of the multikulti strategies. Besides, The success of this kind of strategies is explained via the measurement of entropy as a representation of population diversity. In conclusion, existing schemes for diversity based migration concentrate on selecting individuals for migration with the guide of diversity.

Iii Proposed Scheme Description

Iii-a Preliminary

Migration may bring new building blocks for target subpopulation, while increases the similarity between source subpopulation and target one inevitably. Hence, there are many occasions not fit for migration during a run. For instance, in a subpopulation, migration may not be required when getting better individuals by variation operators is easy. In such a case, migration have little positive influence but still decreases the difference between subpopulations. Anyhow, migration moments predetermined by intervals cannot match the dynamic situation of evolution. Consequently, inappropriate migration cannot be avoided effectively.

Iii-B Scheme to Solve Existing Problem

Our thinking is that more migration chances should be provided when subpopulation diversity is low. In general, the computation for subpopulation diversity is time-consuming in most EAs. Consequently, it is not fit to decide migration moments only based on subpopulation diversity since, in such a plan, the computation for subpopulation diversity has to be executed frequently. Also, setting a value of subpopulation diversity as the threshold for migration is not adopted by us. Two of reasons are listed below. Firstly, the fitting value is difficult to find and may vary much from one occasion to another. Besides, migration may not occur at all in some periods, such as the initial stage.

In our scheme, the attempt of migration still occurs at intervals, but the probability of immigrants entering the target subpopulation, , will be computed according to the formula that builds the relationship between subpopulation diversity, , and the probability, . At first, the simplest formula to describe the relationship between and , Formula 1, is designed by us.

(1)

In the initial stage of a run, the value of is close to the maximum, 1. According to Formula 1, is near to 0 at this time. Then, decreasing monotonically makes increasing monotonically. In the end, approaches 1 since has come to the minimum, 0. Then, Formula 2, where , and Formula 3, are considered.

(2)
(3)

Finally, Formula 4 is adopted since the three former ones can be regarded as special cases of it.

(4)

Essentially, and are just used to adjust the relationship between and .

The flowchart of a DEA with our scheme is shown as Algorithm 2. The scheme is in the underlined and italic text of Algorithm 2.

1:  Generate at random a population, , of individuals
2:  Divide into , …, subpopulations
3:  Define a neighborhood structure for , , …,
4:  repeat {Execute in parallel the next steps for , , …, }
5:     Apply, during generations, stochastic operators of EA
6:     Send individuals to neighboring subpopulations
7:     Receive individuals from neighboring subpopulations
8:     Compute
9:     Compute according to the formula
10:     Generate a random number,
11:     if  then
12:        Put received individuals into the subpopulation
13:     else
14:        discard them
15:     end if
16:  until the stop criterion is not fulfilled
Algorithm 2 DEA with proposed scheme

Essentially, our scheme devotes to control the probability of completing migration according to subpopulation diversity at intervals rather than to improve subpopulation diversity.

Iii-C Analysis of Our Scheme on Running Time

Provided that all supopulations in DEAs remains in exact synchronization, the consuming time for the two algorithms can be expressed as below, respectively. Let be the consuming time of all evolutionary operations in one generation, be that of once diversity computation, be that of a migration round, be interval and be total generations. In a DEA based on the proposed scheme, the total consuming time in a run, , can be computed as follow:

(5)

In a DEA without the proposed scheme, be the consuming time of a migration round. Then, the total consuming time in a run, can be computed as follow:

(6)

If above two types of DEA are based on the same EA and same in settings of common parameters, comparisons can be done as follow: Firstly,

(7)

mainly because, in the DEA with the scheme, a migration round requires the maximum time, , only when is satisfied. Moreover, the total consuming time of the former algorithm have a peculiar part,

(8)

Therefore, the difference of these two algorithms on total consuming time in a run can be measured as below:

(9)

The total consuming time of the former is larger than that of the latter because

(10)

That is, consuming time of once diversity computation is much more than the difference in that of a migration round for different algorithms.

In the scheme, subpopulation diversity is computed only at intervals. Although the diversity measure may vary according to chromosome coding, the extra consuming time in a DEA for the proposed scheme will always be acceptable if migration interval, , is large enough. In fact, to maintain the difference between subpopulations, is always large in the majority of DEAs. Therefore, this scheme can be widely used in DEAs.

Iv Experiment Studies

Iv-a Selected Problem and EA for It

The famous NP-hard problem, TSP, is selected for our experiment. [15] has its benchmark instances. Some EA for the TSP have remarkable performance. For instance, the one presented in [19]

outperforms state-of-the-art heuristic algorithms in finding very high-quality solutions on instances with up to 200,000 cities.

It should be stressed that the purpose of our experiments is not to find better solutions of any problem than ever but to test our scheme by comparing the performance of a DEA with it and that of a traditional one. For this purpose, the more powerful a EA is, the larger instances, which demand much more on resource, should be used. In consideration of this, the EA in [20] proposed for years is used in our experiment. It is based on the inver-over operator [21], the selection method that each individual competes with its offspring only and so-called mapping operator [20]. The flow of mapping is as follow. Firstly, two individuals are selected at random. Then, a segment of chromosome in the individual having worse fitness is selected at random. After that, a segment which has the same number of cities and the same first city is searched in the other individual. As soon as it is found, the former segment in the worse individual is replaced by it. Finally, in the worse individual, the other part of the chromosome is adjusted according to the latter steps of partially mapped crossover [22] which is a traditional operator used in EAs for the TSP. It can be seen from the flow that mapping belongs to crossover. In this EA, evolutionary velocity, , which is calculated as the Formula 11, decides whether the mapping operator should be executed.

(11)

In this formula, is the fitness of the current best individual, is the that of the previous one and denotes the generations between the appearance of the previous best individual and that of the current one. Mapping is carried out only when is lower than a threshold value.

Iv-B Diversity Measure in Our DEA

Matrix M in Formula 12 is connection matrix of TSP tour [23].

(12)

In the matrix, is the number of cities and . represents that there is a connection from city  to city  in tour, while denotes that such a connection does not exist. Let and be two individuals. Then, each of them has a connection matrix. Let be the number of rows which are same in the two matrixes. Then, the difference between and can be defined as Formula 13 [23].

(13)

Then, in [23], subpopulation diversity is defined as Formula 14, where is subpopulation size and means combination.

(14)

In this paper, Formula 14 is replaced by Formula 15 in order to decrease computation complexity.

(15)

In Formula 15, is the best individual in subpopulation. represents each individual in it other than . For once computation of subpopulation diversity, this change makes that total times of the calculation for the difference between two individuals greatly reduce from to in each subpopulation. Although the average of distance between pairwise individuals is replaced by that from the best individual to another one may lead to some error in the resulting value of subpopulation diversity, convergence degree can still be reflected with much less computation. As the steps shown in Algorithm 2, of each subpopulation is computed at intervals to obtain each .

Iv-C Experiment to Compare Result

Our experiment is carried on a Drawing TC5000A computing platform. It has 1264 2.6 GHz cores. Its memory capacity is 1.5 TB. In our experiment, and are set different value combinations to find the fit relationship between and . In detail, their values are both get from the set, . In total, there are nine value combinations. Fig. 1-9 are graphs of the function expressed by Formula 4 under different value combination of and , respectively. Then, the outcomes under each value combination are compared with those of a traditional DEA, respectively. The algorithms in the experiment are both based on the EA introduced in Subsection 4.1 and are the same in setting of common parameters. The thinking of setting for these common parameters comes from [24]. In detail, common parameters except migration interval are set the same value in both algorithms. Then, for each instance, the traditional algorithm runs thirty times independently under five equal difference intervals, respectively. Under each of the nine value combinations of and , so does the DEA with our scheme. The value of common parameters except migration interval is listed in Table I.

Evolutionary parameters Mutation rate () Changing during a run according to Formula 16 based on the initial value, 0.02.
Crossover rate 1-
Mapping rate () Changing during a run according to Formula 17 based on the initial value, 0.05.
Threshold value of evolutionary velocity 5000
Parallel parameters Migration topology Ring
Migration strategy Random-random
Migration size 1
Subpopulation size 100
Quantity of subpopulation 16
Terminal criterion 2000 migration rounds having been done
TABLE I: Setting of parameters

This table shows that mutation rate, , and mapping rate, , change during a run according to Formula 16 and Formula  17, respectively.

(16)
(17)

In Formula 16, is the initial . Similarly, is the initial in Formula 17. In both formulas, denotes current generations and represents the maximal ones.

Since the EA used in our experiment can find the optimum solution of many TSP instances smaller than a280, eight much larger instances in [15] from pcb442 to vm1084 are chosen in our experiment. The results of the traditional algorithm are listed in Table II. Those of the DEA based on our scheme under different value combinations of and are listed in Table III-XI

, respectively. In these tables, each instance corresponds to five migration intervals. Under each interval, the average of outcomes and the standard deviation of them is given. Besides, the optimal solution of each instance provided by

[15] is listed in each table. In Table III-XI, results having significant difference with those in Table II

in terms of t-test with 95% confidence are highlighted by

italics and bold.

Instance Interval
Outcomes
average
Standard
deviation
Optimal
solution
pcb442 150000 50939.3 7.01
200000 50937.7 7.84
250000 50935.0 24.85 50778
300000 50937.5 7.54
350000 50934.6 10.94
p654 50000 34643.4 1.22
60000 34643.4 1.22
70000 34643.5 1.57 34643
80000 34643.0 0.00
90000 34643.1 0.37
d657 200000 49092.7 36.28
250000 49084.5 27.21
300000 49063.9 39.07 48912
350000 49083.2 34.98
400000 49065.3 35.79
u724 150000 42143.3 43.23
200000 42105.0 37.78
250000 42086.1 43.66 41910
300000 42085.6 32.77
350000 42068.8 33.05
rat783 150000 8829.1 11.71
200000 8821.8 9.11
250000 8818.9 7.82 8806
300000 8815.6 5.97
350000 8815.6 7.50
dsj1000 800000 18786697.6 27651.24
1000000 18756550.5 20746.29
1200000 18761931.7 19044.51 18659688
1400000 18762151.6 24886.79
1600000 18757890.0 21787.65
pr1002 500000 259680.9 253.01
600000 259551.0 233.13
700000 259553.3 249.21 259045
800000 259399.3 159.45
900000 259459.9 208.12
vm1084 600000 239785.1 204.25
800000 239773.2 177.70
1000000 239779.5 187.50 239297
1200000 239738.4 178.93
1400000 239739.5 188.81
TABLE II: Results of traditional DEA
Instance Interval
Outcomes
average
Standard
deviation
Optimal
solution
pcb442 150000 50936.8 8.37
200000 50935.6 8.11
250000 50927.8 24.62 50778
300000 50924.8 33.10
350000 50934.5 12.86
p654 50000 34643.1 0.51
60000 34643.1 0.37
70000 34643.0 0.00 34643
80000 34643.1 0.37
90000 34643.1 0.37
d657 200000 49077.9 32.29
250000 49057.9 32.71
300000 49059.6 31.23 48912
350000 49057.0 30.24
400000 49059.1 45.58
u724 150000 42099.2 31.61
200000 42074.6 33.18
250000 42073.3 40.11 41910
300000 42064.9 23.90
350000 42060.3 36.93
rat783 150000 8825.0 7.97
200000 8817.3 7.62
250000 8815.9 8.28 8806
300000 8814.5 6.23
350000 8814.8 5.52
dsj1000 800000 18761927.0 25664.25
1000000 18762398.9 23733.48
1200000 18748623.4 24506.86 18659688
1400000 18745199.9 27584.08
1600000 18744510.3 23523.72
pr1002 500000 259551.6 268.12
600000 259504.9 223.44
700000 259383.7 172.11 259045
800000 259375.9 233.52
900000 259387.2 239.40
vm1084 600000 239736.0 164.15
800000 239676.7 141.30
1000000 239702.1 163.49 239297
1200000 239682.9 179.91
1400000 239637.6 122.91
TABLE III: Results of DEA with our scheme when and
Fig. 1: Graph of function for our scheme when and
Instance Interval
Outcomes
average
Standard
deviation
Optimal
solution
pcb442 150000 50931.7 12.66
200000 50934.3 10.46
250000 50929.5 29.45 50778
300000 50932.7 10.44
350000 50927.1 22.43
p654 50000 34643.0 0.00
60000 34643.0 0.00
70000 34643.0 0.00 34643
80000 34643.0 0.00
90000 34643.0 0.00
d657 200000 49055.4 35.44
250000 49057.3 42.35
300000 49048.0 38.85 48912
350000 49045.9 37.06
400000 49036.6 32.54
u724 150000 42070.7 33.66
200000 42067.9 29.89
250000 42054.6 31.99 41910
300000 42037.9 39.68
350000 42036.7 35.86
rat783 150000 8822.8 9.91
200000 8815.8 7.32
250000 8816.2 7.01 8806
300000 8812.9 5.72
350000 8812.9 8.13
dsj1000 800000 18756031.3 24623.43
1000000 18748806.2 26546.89
1200000 18748989.6 22840.60 18659688
1400000 18735264.4 20742.59
1600000 18738518.6 26541.26
pr1002 500000 259433.6 190.88
600000 259335.0 165.78
700000 259303.5 140.85 259045
800000 259280.7 138.09
900000 259354.0 199.96
vm1084 600000 239683.8 162.64
800000 239665.7 136.93
1000000 239671.2 135.21 239297
1200000 239638.1 145.54
1400000 239623.6 126.22
TABLE IV: Results of DEA with our scheme when and
Fig. 2: Graph of function for our scheme when and
Instance Interval
Outcomes
average
Standard
deviation
Optimal
solution
pcb442 150000 50932.6 11.09
200000 50930.4 10.13
250000 50926.6 27.04 50778
300000 50919.6 32.26
350000 50922.3 23.90
p654 50000 34643.0 0.00
60000 34643.0 0.00
70000 34643.0 0.00 34643
80000 34643.0 0.00
90000 34643.0 0.00
d657 200000 49050.0 31.05
250000 49054.0 41.66
300000 49045.9 34.37 48912
350000 49028.7 31.89
400000 49042.3 36.07
u724 150000 42072.2 31.06
200000 42045.2 27.53
250000 42045.9 30.77 41910
300000 42032.3 29.95
350000 42026.7 26.53
rat783 150000 8828.1 8.83
200000 8817.9 6.69
250000 8813.9 6.77 8806
300000 8815.0 6.24
350000 8813.4 6.44
dsj1000 800000 18741740.2 22220.14
1000000 18739983.0 24819.99
1200000 18738972.7 18942.56 18659688
1400000 18724773.5 27336.00
1600000 18732647.5 20957.86
pr1002 500000 259277.2 136.14
600000 259402.0 201.64
700000 259342.0 189.56 259045
800000 259296.1 180.98
900000 259288.9 132.95
vm1084 600000 239691.4 169.62
800000 239690.1 167.67
1000000 239637.8 128.26 239297
1200000 239638.7 171.76
1400000 239585.6 133.70
TABLE V: Results of DEA with our scheme when and
Fig. 3: Graph of function for our scheme when and
Instance Interval
Outcomes
average
Standard
deviation
Optimal
solution
pcb442 150000 50938.3 6.86
200000 50937.4 8.53
250000 50934.7 10.22 50778
300000 50933.5 27.43
350000 50921.9 35.34
p654 50000 34643.3 1.18
60000 34643.6 2.30
70000 34643.1 0.37 34643
80000 34643.0 0.00
90000 34643.0 0.00
d657 200000 49076.7 42.28
250000 49083.5 33.96
300000 49070.3 32.63 48912
350000 49060.2 29.56
400000 49058.7 42.51
u724 150000 42093.8 39.47
200000 42081.5 31.40
250000 42081.8 30.52 41910
300000 42064.9 28.86
350000 42071.3 30.47
rat783 150000 8825.9 10.50
200000 8818.5 8.52
250000 8816.4 6.64 8806
300000 8814.4 6.72
350000 8814.0 7.31
dsj1000 800000 18773122.2 23162.37
1000000 18761402.1 26171.98
1200000 18755819.7 15957.70 18659688
1400000 18746441.0 23247.20
1600000 18752288.1 19447.99
pr1002 500000 259517.0 273.21
600000 259468.7 236.99
700000 259449.1 199.82 259045
800000 259515.5 272.15
900000 259499.3 226.40
vm1084 600000 239770.3 162.83
800000 239730.6 137.43
1000000 239693.5 151.11 239297
1200000 239694.1 158.66
1400000 239670.1 132.08
TABLE VI: Results of DEA with our scheme when and
Fig. 4: Graph of function for our scheme when and
Instance Interval
Outcomes
average
Standard
deviation
Optimal
solution
pcb442 150000 50931.5 20.76
200000 50929.5 26.70
250000 50933.9 12.78 50778
300000 50931.7 10.29
350000 50931.2 22.20
p654 50000 34643.1 0.37
60000 34643.1 0.37
70000 34643.4 1.52 34643
80000 34643.0 0.00
90000 34643.0 0.00
d657 200000 49067.4 43.30
250000 49063.7 37.62
300000 49055.5 36.67 48912
350000 49050.3 32.32
400000 49046.9 43.36
u724 150000 42094.9 29.82
200000 42064.9 42.94
250000 42086.1 42067.8 34.59
300000 42066.3 25.19
350000 42054.6 30.69
rat783 150000 8824.0 8.83
200000 8819.7 7.47
250000 8815.1 6.77 8806
300000 8814.0 7.05
350000 8812.5 6.69
dsj1000 800000 18763625.5 21831.34
1000000 18762785.8 24970.92
1200000 18746381.2 20736.39 18659688
1400000 18745791.7 17386.04
1600000 18743302.0 18939.43
pr1002 500000 259435.1 224.69
600000 259417.3 222.38
700000 259359.6 169.53 259045
800000 259341.7 198.52
900000 259317.7 187.90
vm1084 600000 239739.4 183.33
800000 239694.4 158.19
1000000 239691.5 168.34 239297
1200000 239699.0 143.17
1400000 239668.0 129.28
TABLE VII: Results of DEA with our scheme when and
Fig. 5: Graph of function for our scheme when and
Instance Interval
Outcomes
average
Standard
deviation
Optimal
solution
pcb442 150000 50934.0 10.64
200000 50931.3 11.15
250000 50932.7 12.33 50778
300000 50931.5 10.96
350000 50927.6 14.41
p654 50000 34643.0 0.00
60000 34643.0 0.00
70000 34643.0 0.00 34643
80000 34643.1 0.37
90000 34643.0 0.00
d657 200000 49052.7 35.84
250000 49053.1 36.05
300000 49050.2 34.89 48912
350000 49031.9 34.44
400000 49041.6 47.01
u724 150000 42070.6 37.78
200000 42056.0 37.16
250000 42035.9 27.50 41910
300000 42031.8 30.09
350000 42033.0 30.42
rat783 150000 8824.6 9.78
200000 8817.7 8.45
250000 8814.6 6.62 8806
300000 8812.7 6.50
350000 8814.3 6.14
dsj1000 800000 18740584.7 20775.90
1000000 18743355.3 20831.26
1200000 18741427.6 24245.91 18659688
1400000 18729556.2 19250.64
1600000 18733965.3 25656.26
pr1002 500000 259394.7 204.43
600000 259367.8 198.61
700000 259331.4 180.84 259045
800000 259410.4 179.61
900000 259307.4 181.98
vm1084 600000 239724.6 145.35
800000 239717.7 149.77
1000000 239638.4 148.45 239297
1200000 239642.8 116.59
1400000 239644.8 142.91
TABLE VIII: Results of DEA with our scheme when and
Fig. 6: Graph of function for our scheme when and
Instance Interval
Outcomes
average
Standard
deviation
Optimal
solution
pcb442 150000 50933.8 9.27
200000 50936.4 8.19
250000 50934.8 10.53 50778
300000 50936.1 9.49
350000 50937.4 8.42
p654 50000 34643.4 1.52
60000 34643.1 0.37
70000 34643.1 0.51 34643
80000 34643.3 1.14
90000 34643.0 0.00
d657 200000 49085.9 38.80
250000 49095.7 30.77
300000 49075.7 29.14 48912
350000 49065.2 40.71
400000 49066.2 32.46
u724 150000 42113.1 41.00
200000 42097.8 41.37
250000 42084.7 43.12 41910
300000 42076.8 30.95
350000 42077.5 37.93
rat783 150000 8828.0 8.28
200000 8818.5 6.06
250000 8814.8 6.23 8806
300000 8816.0 5.43
350000 8814.6 5.70
dsj1000 800000 18776618.6 27998.28
1000000 18761359.1 26059.44
1200000 18753840.4 22750.80 18659688
1400000 18753787.1 26156.22
1600000 18752323.7 21359.45
pr1002 500000 259622.7 228.65
600000 259523.0 314.84
700000 259479.1 211.93 259045
800000 259500.5 208.99
900000 259519.0 196.58
vm1084 600000 239765.1 122.05
800000 239744.2 134.48
1000000 239774.2 153.76 239297
1200000 239675.7 164.24
1400000 239661.6 125.02
TABLE IX: Results of DEA with our scheme when and
Fig. 7: Graph of function for our scheme when and
Instance Interval
Outcomes
average
Standard
deviation
Optimal
solution
pcb442 150000 50934.5 9.73
200000 50935.6 25.76
250000 50937.1 8.39 50778
300000 50935.8 8.25
350000 50935.2 11.55
p654 50000 34643.3 0.69
60000 34643.1 0.37
70000 34643.0 0.00 34643
80000 34643.0 0.00
90000 34643.1 0.37
d657 200000 49078.2 42.55
250000 49078.7 47.63
300000 49063.7 42.80 48912
350000 49068.5 41.41
400000 49065.7 40.75
u724 150000 42112.0 30.64
200000 42081.4 26.93
250000 42080.1 38.89 41910
300000 42069.0 38.67
350000 42064.8 27.71
rat783 150000 8827.5 8.08
200000 8817.2 4.93
250000 8816.7 8.42 8806
300000 8816.5 6.18
350000 8815.4 5.77
dsj1000 800000 18762595.7 30773.90
1000000 18765975.2 33663.58
1200000 18759551.7 22376.08 18659688
1400000 18755129.5 24348.98
1600000 18753479.1 21177.11
pr1002 500000 259586.6 235.77
600000 259520.7 222.06
700000 259454.7 244.08 259045
800000 259351.2 196.07
900000 259401.3 166.26
vm1084 600000 239746.5 145.48
800000 239690.4 160.28
1000000 239742.5 204.31 239297
1200000 239709.6 160.58
1400000 239715.3 155.95
TABLE X: Results of DEA with our scheme when and
Fig. 8: Graph of function for our scheme when and
Instance Interval
Outcomes
average
Standard
deviation
Optimal
solution
pcb442 150000 50937.6 8.36
200000 50938.5 8.21
250000 50935.4 10.23 50778
300000 50934.7 10.39
350000 50924.3 31.84
p654 50000 34643.1 0.51
60000 34643.0 0.00
70000 34643.0 0.00 34643
80000 34643.0 0.00
90000 34643.0 0.00
d657 200000 49060.4 42.18
250000 49054.0 35.61
300000 49046.5 39.59 48912
350000 49050.9 30.57
400000 49052.1 35.16
u724 150000 42080.9 27.08
200000 42089.8 35.98
250000 42078.1 35.95 41910
300000 42071.6 42.63
350000 42052.2 39.33
rat783 150000 8821.6 9.10
200000 8819.5 8.56
250000 8816.2 6.56 8806
300000 8815.4 6.30
350000 8814.8 5.12
dsj1000 800000 18761976.8 25253.82
1000000 18760042.3 14768.00
1200000 18755479.9 22505.88 18659688
1400000 18748354.6 22603.43
1600000 18746204.2 23893.81
pr1002 500000 259537.5 306.89
600000 259412.2 244.67
700000 259464.5 237.79 259045
800000 259410.7 225.29
900000 259372.2 182.52
vm1084 600000 239789.1 252.07
800000 239706.1 172.80
1000000 239746.1 168.84 239297
1200000 239653.7 146.65
1400000 239660.9 140.98
TABLE XI: Results of DEA with our scheme when and
Fig. 9: Graph of function for our scheme when and

Based on the results in Table II-XI, the difficulty of a instance to the two algorithms, DF, can be expressed according to Formula 18, where denotes the average of all solutions for one instance obtained in our experiment and represents the optimal solution provided by [15].

(18)

The value of difficulty is listed in Table XII.

Instance pcb442 p654 d657 u724
Difficulty 0.003048 0.000003 0.003040 0.003842
Instance rat783 dsj1000 pr1002 vm1084
Difficulty 0.001337 0.004947 0.001482 0.001682
TABLE XII: Difficulty of each instance for DEAs

According to Table XII

, the difficulty of each instance can be classified into three levels. That of p654 belongs to the lowest level. Rat783, pr1002 and vm1084 have an intermediate level difficulty. The rest four instances, pcb442, d657, u724, dsj1000, are ones with high difficulty.

It can be seen in Table II-XI that, for the lowest difficulty instance, p654, significant difference between solutions of the DEA based on our scheme and those of the traditional one can be found in no case. For the intermediate difficulty instances, the DEA with the proposed scheme significantly wins in forty-five cases out of one hundred and thirty-five ones (45/135) and statistically loses in two cases (2/135). It should be noted that, for high difficulty instances, the DEA with our scheme yields significantly better outcomes than its peer in ninety cases out of one hundred and eighty ones (90/180). Meanwhile, there are no significant differences in all the rest cases. Also, the tables show that the solutions’ standard deviation of the DEA with our scheme is less than that of the traditional one in two hundred and twenty-six cases out of the all ones (226/360).

Moreover, it can been seen that, under different value combinations of and , the performance of the algorithm with our scheme is significant different. When , or , , for the all eight instances, the algorithm significantly wins in the most cases (27/40) and never statistically loses. When , , the DEA with our scheme also has a good behavior. In detail, it significantly wins in twenty-six cases (26/40) and never statistically loses. On the whole, the winning rate is 135/360, while the losing rate is 2/360. Besides, under the three outstanding value combinations, solutions of the DEA with our scheme have a better standard deviation in eighty-two cases out of one hundred and twenty ones (82/120). This rate is better than that on the whole (226/360). Fig. 1-9 show that, under the three outstanding value combinations, the graph of the function in Formula 4 have the same characteristic which is distinguished from that under the other six ones.

In conclusion, results shows that the DEA based on our scheme has an advantage on solutions. Moreover, a value combination of and which makes the graphing slope of the function in Formula 4 increase monotonically from to zero when is fit for our scheme.

V Conclusion

In this paper, we have presented the scheme of setting the success rate of migration based on subpopulation diversity at each interval for DEAs. Under the control of the scheme, immigrants enter the target subpopulation at a probability, which is the function of subpopulation diversity according to Formula 4, at intervals. In our experiment, eight instances of the TSP are used to test algorithms. In detail, under nine different value combinations of parameters required for Formula 4, outcomes of the algorithm with the scheme are compared with those of the traditional DEA, respectively. The experimental results show that, especially for high difficulty instances, the DEA based on our scheme has a significant advantage on solutions. Moreover, under three value combinations of the parameters making graphing slope of the function in Formula 4 increase monotonically from to zero when , the algorithm with the scheme has most outstanding performance.

Based on the key factor in the behavior of evolutionary computation, diversity, we propose this scheme and discuss its parameters setting. This scheme can be used to improve solutions of DEAs for diversified problems. To apply the scheme, first of all, finding the diversity measure of a certain chromosome coding is necessary. In future, how to use this scheme together with some other diversity based methods should be studied to further improve DEAs. Also, the simplified calculation method for subpopulation diversity to reduce the time complexity used in our scheme should be discussed in theory.

Acknowledgment

The authors would like to thank Dr. Dunhui Xiao and assistant researcher Jian Wang for their valuable suggestions.

References

  • [1] E. Alba and M. Tomassini, “Parallelism and evolutionary algorithms,” IEEE Transactions on Evolutionary Computation, vol. 6, no. 5, pp. 443–462, 2002.
  • [2]

    J. Wu, Z. Cai, S. Zeng, and X. Zhu, “Artificial immune system for attribute weighted naive bayes classification,” in

    The 2013 International Joint Conference on Neural Networks (IJCNN)

    .   IEEE, 2013, pp. 1–8.
  • [3]

    E. Cantú-Paz, “A survey of parallel genetic algorithms,”

    Calculateurs paralleles, reseaux et systems repartis, vol. 10, no. 2, pp. 141–171, 1998.
  • [4] F. Herrera, M. Lozano, and C. Moraga, “Hierarchical distributed genetic algorithms,” International journal of intelligent systems, vol. 14, no. 11, pp. 1099–1121, 1999.
  • [5]

    W. Bossert, “Mathematical optimization: Are there abstract limits on natural selection?” in

    The Wistar Institute symposium monograph, vol. 5, 1967, p. 35.
  • [6] U. Tosun, T. Dokeroglu, and A. Cosar, “A robust island parallel genetic algorithm for the quadratic assignment problem,” International Journal of Production Research, vol. 51, no. 14, pp. 4117–4133, 2013.
  • [7] K. Zidi, F. Mguis, P. Borne, and K. Ghedira, “Distributed genetic algorithm for disaster relief planning,” International Journal of Computers Communications & Control, vol. 8, no. 5, pp. 769–783, 2013.
  • [8] F. G. Bulnes, R. Usamentiaga, D. F. Garcia, and J. Molleda, “A parallel genetic algorithm for optimizing an industrial inspection system,” IEEE Latin America Transactions, vol. 11, no. 6, pp. 1338–1343, 2013.
  • [9] I. Rocha, E. Parente, and A. Melo, “A hybrid shared/distributed memory parallel genetic algorithm for optimization of laminate composites,” Composite Structures, vol. 107, pp. 288–297, 2014.
  • [10] O. Devos, G. Downey, and L. Duponchel, “Simultaneous data pre-processing and svm classification model selection based on a parallel genetic algorithm applied to spectroscopic data of olive oils,” Food chemistry, vol. 148, pp. 124–130, 2014.
  • [11] J. Denzinger and J. Kidney, “Improving migration by diversity,” in Evolutionary Computation, 2003. CEC’03. The 2003 Congress on, vol. 1.   IEEE, 2003, pp. 700–707.
  • [12] D. Power, C. Ryan, and R. M. A. Azad, “Promoting diversity using migration strategies in distributed genetic algorithms,” in 2005 IEEE Congress on Evolutionary Computation, vol. 2.   IEEE, 2005, pp. 1831–1838.
  • [13]

    X. Wei, X. Yang, and Y. Deng, “Maintain gene diversity with set pair analysis in parallel genetic programming,” in

    The Proceeding of the 8th International Symposium on Distributed Computing and Applications to Business, Engineering and Science, 2009, pp. 76–79.
  • [14] L. Araujo and J. J. Merelo, “Diversity through multiculturality: Assessing migrant choice policies in an island model,” IEEE Transactions on Evolutionary Computation, vol. 15, no. 4, pp. 456–469, 2011.
  • [15] G. Reinelt, “Tsplib a traveling salesman problem library,” ORSA journal on computing, vol. 3, no. 4, pp. 376–384, 1991.
  • [16] Z. Skolicki and K. De Jong, “The influence of migration sizes and intervals on island models,” in Proceedings of the 7th annual conference on Genetic and evolutionary computation.   ACM, 2005, pp. 1295–1302.
  • [17] J. P. Cohoon, S. U. Hegde, W. N. Martin, and D. Richards, “Punctuated equilibria: a parallel genetic algorithm,” in Genetic algorithms and their applications: proceedings of the second International Conference on Genetic Algorithms: July 28-31, 1987 at the Massachusetts Institute of Technology, Cambridge, MA.   Hillsdale, NJ: L. Erlhaum Associates, 1987., 1987.
  • [18] M. Črepinšek, S.-H. Liu, and M. Mernik, “Exploration and exploitation in evolutionary algorithms: A survey,” ACM Computing Surveys (CSUR), vol. 45, no. 3, p. 35, 2013.
  • [19] Y. Nagata and S. Kobayashi, “A powerful genetic algorithm using edge assembly crossover for the traveling salesman problem,” INFORMS Journal on Computing, vol. 25, no. 2, pp. 346–363, 2013.
  • [20] Z.-H. Cai, J.-G. Peng, W. Gao, W. Wei, and L.-S. Kang, “Improved evolutionary algorithm for the traveling salesman problem.” Jisuanji Xuebao(Chin. J. Comput.), vol. 28, no. 5, pp. 823–828, 2005.
  • [21] G. Tao and Z. Michalewicz, “Inver-over operator for the tsp,” in International Conference on Parallel Problem Solving from Nature.   Springer, 1998, pp. 803–812.
  • [22] D. E. Goldberg and R. Lingle, “Alleles, loci, and the traveling salesman problem,” in Proceedings of an International Conference on Genetic Algorithms and Their Applications, vol. 154.   Lawrence Erlbaum, Hillsdale, NJ, 1985, pp. 154–159.
  • [23] P.-C. Chang, W.-H. Huang, and C.-J. Ting, “Dynamic diversity control in genetic algorithm for mining unsearched solution space in tsp problems,” Expert systems with applications, vol. 37, no. 3, pp. 1863–1878, 2010.
  • [24] C. Li and G. Hu, “Global migration strategy with moving colony for hierarchical distributed evolutionary algorithms,” Soft Computing, vol. 18, no. 11, pp. 2161–2176, 2014.