CN109816091A

CN109816091A - An Improved Biogeographic Computational Method

Info

Publication number: CN109816091A
Application number: CN201910127880.1A
Authority: CN
Inventors: 李静梅; 吴辉; 张天越
Original assignee: Harbin Engineering University
Current assignee: Harbin Engineering University
Priority date: 2019-02-20
Filing date: 2019-02-20
Publication date: 2019-05-28

Abstract

The present invention provides a kind of improved biogeography calculation method, initializes calculation method parameter, initializes one group of habitat, calculates similarity, and press similarity incremental arrangement, selects preceding S_maxA habitat constitutes initial population, calculates its fitness value, and calculate habitat moves into rate and emigration, executes migration operation；The aberration rate of habitat is calculated, mutation operation is executed, calculates the fitness value of habitat, judge whether calculation method reaches maximum number of iterations, judge that calculation method whether to introduce fireworks calculation method explosion operator, has been introduced into, exports optimal solution, terminate, otherwise, explosion operator is introduced, burst radius and spark quantity is set, generates a series of sparks, spark fitness value is calculated, spark of the fitness value higher than population is selected and is replaced.The present invention starts to improve the initialization mode of population in algorithm iteration, improves the global covering power of algorithm；Fireworks algorithm explosion operator is introduced in the iteration later period, is conducive to jump out local optimum.

Description

A kind of improved biogeography calculation method

Technical field

The invention belongs to deep learning fields, and in particular to a kind of improved biogeography calculation method.

Background technique

Biogeography algorithm (Biogeography-based optimization, BBO) be by Dan Simon in A kind of Swarm Intelligence Algorithm proposed in 2008.The it is proposed of the algorithm is in the Research foundation to biological species migration mathematical model On, it uses for reference other frames for freeing intelligent optimization algorithm and is formed.(such as: genetic algorithm, particle with other Swarm Intelligence Algorithms Group's algorithm etc.) it compares, more superior performance is shown, thus obtained domestic and international algorithm research and many of engineering field The concern of person.

Biogeography mathematical model essentially describes how species generate, become extinct and migrate.In a region, such as Some habitat of fruit is very suitable to biological existence, in the habitat with regard to existence suitability degree with higher, can support more Species.For the habitat with higher suitability degree, it will be eventually exhibited as surging species saturation of the same race, a part of option of species is moved Out；If the suitability degree of a habitat is always maintained at lower level, since certain natural calamity may cause this The species extinction of one habitat also results in largely moving into for other species at this time.Meanwhile there is variation in habitat population.

By taking the Species migration of single habitat as an example: when species are 0, the emigration μ of species is 0, and the rate λ of moving into is most Greatly；When species reach S_maxWhen, the rate λ that moves into of species is 0, and emigration μ reaches maximum.Define P_sFor a certain habitat tool There is the probability of species S, then the probability P in t to t+ Δ t_sThe variable be

P_s(t+ Δ t)=P_s(t)(1-λ_sΔt-μ_sΔt)+P_s-1(t)λ_s-1Δt+P_s+1(t)μ_s+1Δt

In formula: λ_sAnd μ_sRate and emigration are moved into for species of the habitat when species are S.

Assuming that Δ t is sufficiently small, so that the entry/leave more than a kind of species is ignored, then when there is the following formula in Δ t → 0 It sets up

When Species migration model is linear transport model, have

In formula, E and I are respectively that maximum emigration and maximum move into rate.

BBO algorithm can be expressed as a triple:

BBO=(I, ψ, T)

In formula, I is one and generates the function of the initial ecosystem, and calculates the fitness of each habitat；ψ is ecology System transfer function, calculate each habitat first moves into rate and emigration, then adjusts habitat according to suitability degree, finally Mutation operation is executed, and recalculates the fitness value of habitat；T indicates stopping criterion.

Summary of the invention

The object of the present invention is to provide a kind of improved biogeography calculation methods, overcome standard biological geography Learning algorithm has the shortcomings that later period convergence is short of power, easily falls into local optimum.

The object of the present invention is achieved like this:

A kind of improved biogeography calculation method, concrete implementation step are as follows:

Step 1. initializes calculation method parameter, including maximum species number S_max, maximum move into rate E_max, maximum emigration I_max, maximum aberration rate m_maxAnd maximum number of iterations；

Step 2. initializes one group of habitat, and habitat quantity is much larger than maximum species number S_max；

Step 3. calculates the similarity of each habitat, and similarity incremental arrangement is pressed in habitat；

Step 4. selects preceding S_maxA habitat constitutes initial population, calculates the fitness value of initial population；

What step 5. calculated each group of habitat moves into rate and emigration, executes migration operation；

Step 6. calculates the aberration rate of each group of habitat, executes mutation operation；

The fitness value of step 7. calculating habitat；

Step 8. judges whether calculation method reaches maximum number of iterations, reaches, and goes to step 9, otherwise goes to step 5；

Step 9. judges whether calculation method has been introduced into fireworks calculation method explosion operator, has been introduced into, exports optimal solution, Terminate, otherwise goes to step 10；

Step 10. introduces explosion operator, sets burst radius and spark quantity, generates a series of sparks；

Step 11. calculates spark fitness value, selects spark of the fitness value higher than population and is replaced, goes to step 5.

The similarity of habitat is improved by Euclidean distance formula in the step 3, and specific formula for calculation is

Wherein, SIMILAR_ijFor habitat X_iAnd X_jSimilarity value, N be habitat dimension, x_ikAnd x_jkFor habitat x_i With habitat x_jThe corresponding value in kth dimension.

Calculation method restrains for the first time in the step 10, introduces fireworks calculation method explosion operator at this time, generates a system Column spark and the habitat for preferentially replacing fitness difference in population, continue iteration, export most when restraining for second of calculation method Excellent solution.

The beneficial effects of the present invention are: a kind of improved biogeography calculation method proposed by the present invention, in algorithm Iteration starts to improve the initialization mode of population, calculates the similarity between habitat, the habitat group for selecting similarity small At initial population, the global covering power of algorithm is improved；Fireworks algorithm explosion operator is introduced in the iteration later period, by fitness height Habitat generate a series of sparks as fireworks, and replace the habitat of fitness difference, be conducive to jump out local optimum.

Detailed description of the invention

Fig. 1 is algorithm flow chart of the invention.

Fig. 2 is the optimal solution for three kinds of algorithms that Schwefel2.22 function is test object.

Fig. 3 is three kinds of convergence speed of the algorithm figures that Schwefel2.22 function is test object.

Specific embodiment

The present invention will be further described with reference to the accompanying drawing:

Embodiment 1

Standard biological geography algorithm has the shortcomings that later period convergence is short of power, easily falls into local optimum, and the present invention mentions A kind of improved biogeography algorithm EBBO out, starts the initialization mode for improving population in algorithm iteration, calculating is dwelt The similarity between ground is ceased, the habitat for selecting similarity small forms initial population, improves the global covering power of algorithm；? The iteration later period introduces fireworks algorithm explosion operator, generates a series of sparks for the high habitat of fitness as fireworks, and replace The habitat of fitness difference, is conducive to jump out local optimum.

A kind of the step of improved biogeography calculation method, is as follows:

It is Schwefel 2.22 that step 1., which selectes global optimization test function, and optimization aim is the functional minimum value, i.e., Min (Schwefel 2.22), function dimension n are 50, function expression are as follows:

S.t.n=50

Step 2. initialization algorithm parameter, including maximum species number S_maxRate E is moved into for 50, maximum_maxIt is moved for 0.95, maximum Extracting rate I_maxIt is 0.95, maximum aberration rate m_maxIt is 0.05, the number of iterations was 2000 generations；

Step 3. initializes one group of habitat, and habitat quantity is maximum species number S_max×2；

Step 4. calculates the similarity of each habitat, and similarity incremental arrangement is pressed in habitat；

Step 5. selects preceding S_maxA habitat constitutes initial population, calculates the fitness value of initial population；

What step 6. calculated each group of habitat moves into rate and emigration, executes migration operation；

Step 7. calculates the aberration rate of each group of habitat, executes mutation operation；

The fitness value of step 8. calculating habitat；

Step 9. judges whether algorithm reaches maximum number of iterations, reaches, and goes to step 9, otherwise goes to step 5；

Step 10. judges that algorithm whether to introduce fireworks algorithm explosion operator, has been introduced into, exports Schwefel2.22's Functional value, algorithm terminate, and otherwise go to step 10；

Step 11. introduces explosion operator, set burst radius as 40, and spark quantity is 20, generates a series of sparks；

Step 12. calculates spark fitness value, selects spark of the fitness value higher than population and is replaced, goes to step 5；

Step 13. repeats above step 10 times, is averaged as final output as a result, to reduce error.

Table 1 is the particle swarm algorithm (Particle Swarm Optimization, PSO) and standard biological for comparison The parameter setting of geography algorithm:

The setting of 1 PSO and BBO algorithm parameter of table

In terms of computational accuracy, as shown in Fig. 2, to be that 110.5630, BBO is acquired optimal for the optimal solution that acquires of PSO algorithm Solution 33.507, the optimal solution that algorithm EBBO provided by the invention is acquired are 10.451, and inventive algorithm EBBO computational accuracy has Very big promotion.In terms of convergence rate, as shown in figure 3, EBBO convergence speed of the algorithm is substantially better than PSO and BBO algorithm.

Claims

1. an improved biogeographic calculation method, is characterized in that, concrete realization step is:

Step 1. Initialize the calculation method parameters, including the maximum number of species S _max , the maximum immigration rate E _max , the maximum emigration rate I _max , the maximum mutation rate m _max and the maximum number of iterations;

Step 2. Initialize a set of habitats, and the number of habitats is much larger than the maximum number of species S _max ;

Step 3. Calculate the similarity of each habitat, and arrange the habitats according to the increasing similarity;

Step 4. Select the first S _max habitats to form the initial population, and calculate the fitness value of the initial population;

Step 5. Calculate the in-migration rate and out-migration rate of each group of habitats, and perform migration operations;

Step 6. Calculate the mutation rate of each group of habitats and perform mutation operations;

Step 7. Calculate the fitness value of the habitat;

Step 8. Determine whether the calculation method reaches the maximum number of iterations, and then go to step 9, otherwise go to step 5;

Step 9. Determine whether the calculation method has introduced the explosion operator of the fireworks calculation method, if it has been introduced, output the optimal solution, end, otherwise go to step 10;

Step 10. Introduce the explosion operator, set the explosion radius and the number of sparks, and generate a series of sparks;

Step 11. Calculate the fitness value of sparks, select sparks whose fitness value is higher than that of the population and replace them, and go to step 5.

2. a kind of improved biogeography calculation method according to claim 1, is characterized in that: in described step 3, the similarity of habitat is improved by Euclidean distance formula, and concrete calculation formula is

Among them, SIMILAR _ij is the similarity value of habitats X _i and X _j , N is the habitat dimension, and x _ik and x _jk are the corresponding values of habitat _xi and habitat x _j in the kth dimension.

3. A kind of improved biogeography calculation method according to claim 1, it is characterized in that: in described step 10, calculation method converges for the first time, at this moment, introduce firework calculation method explosion operator, produce a series of sparks and The habitats with poor fitness in the population are replaced by the best, and the iteration is continued, and the optimal solution is output when the calculation method converges for the second time.