The AEFA algorithm may have problems

[abcdvvvv]

The EGGHOLDER function is not a difficult function, but the AEFA algorithm is unable to find its global optimum, as shown in testset:

MetaheuristicsAlgorithms.jl/test/runtests.jl

Line 18 in 86401c7

@test algo.gbest == [-461.648054120324, -390.06329199871277]

The global optimal solution is at (512,404.2319) and the objective function value is -959.6407. So I was thinking there might be something wrong with the AEFA algorithm, could you check it?

This algorithm was the first one I modified and I wanted to use it as a benchmark, so making sure it was correct first was necessary.

The second problem is that an error occurs when compiling the package, suggesting that the levy function cannot be found. Please check this.

MetaheuristicsAlgorithms.jl/src/SuperbFOA.jl

Line 37 in 86401c7

l = 0.5 * levy(N, dim, 1.5)

[AbdelazimHussien]

I have

The EGGHOLDER function is not a difficult function, but the AEFA algorithm is unable to find its global optimum, as shown in testset:

MetaheuristicsAlgorithms.jl/test/runtests.jl

Line 18 in 86401c7

@test algo.gbest == [-461.648054120324, -390.06329199871277]
The global optimal solution is at (512,404.2319) and the objective function value is -959.6407. So I was thinking there might be something wrong with the AEFA algorithm, could you check it?

This algorithm was the first one I modified and I wanted to use it as a benchmark, so making sure it was correct first was necessary.

The second problem is that an error occurs when compiling the package, suggesting that the levy function cannot be found. Please check this.

MetaheuristicsAlgorithms.jl/src/SuperbFOA.jl

Line 37 in 86401c7

l = 0.5 * levy(N, dim, 1.5)

I have fixed the second issue. But only AEFA who has the 1st problem?

[abcdvvvv]

But only AEFA who has the 1st problem?

Maybe so, I just tested the AEO algorithm. It finds the global optimum -959.640662720851.

Please add the following code to runtest.jl:

@testset "AEO on Eggholder function" begin
    Random.seed!(1)
    lb = [-512, -512]
    ub = [512, 512]
    fun(x) = first(benchmark(x))
    BestF, BestX, _ = AEO(100, 300, lb, ub, 2, fun)

    @test BestF == -959.640662720851
    @test BestX == [512.0, 404.2318050941216]
end

[AbdelazimHussien]

Regarding AEFA
the previous version which is same like AEO works well
Can I return AEFA to the previous version and U design the struct of it again?

[abcdvvvv]

I messed up by modifying the AFEA, which was originally working correctly, so now it’s not operational. Please don’t bother sending the original code back—I’ll handle the fix and have it sorted out today.

[AbdelazimHussien]

Thanks my friend.
I will continue the documentation

[abcdvvvv]

I copied the original AFEA from commit with SHA 4eb8bdd.

MetaheuristicsAlgorithms.jl/src/AEFA.jl

Line 10 in 4eb8bdd

function AEFA(N, max_it, lb, ub, D, benchmark)

Tested with

using Random
Random.seed!(1)
AEFA(100,300,[-512 -512],[512 512],2,benchmark)

where the benchmark function is the current version, eggholder.

And the result is

(-645.8190926628417, [308.28046839622516, -483.7819695003975], Any[-645.8190926628417, -645.8190926628417, -645.8190926628417, -645.8190926628417, -645.8190926628417, -645.8190926628417, -645.8190926628417, -645.8190926628417, -645.8190926628417, -645.8190926628417  …  -645.8190926628417, -645.8190926628417, -645.8190926628417, -645.8190926628417, -645.8190926628417, -645.8190926628417, -645.8190926628417, -645.8190926628417, -645.8190926628417, -645.8190926628417])

This was not global optimum. I tried many times and AFEA continued to fail. So, I think this may be the original AFEA's problem.

[AbdelazimHussien]

I have upload a new file name AEFA_1.jl and I tested it and gives the correct results.

Can U please use it to update the original AEFA

After that I will try to convert all other algorithms to be like AEFA.

Thanks

[abcdvvvv]

I tried your AFEA_1 function, but it still doesn't work. Try this code, but it still can't find the global optimum solution.

using LinearAlgebra
include("benchmark.jl")
include("../src/AEFA_1.jl")
lb = [-512 -512]
ub = [512 512]

sol = AEFA_1(100, 300, lb, ub, 2, benchmark)

Just for test, please modify this line:

MetaheuristicsAlgorithms.jl/src/AEFA_1.jl

Line 17 in 29be188

fitness = [func_num(X[i, :]) for i in 1:npop]

to fitness = [func_num(X[i, :])|>first for i in 1:npop]

For example, some output may like this:

(-697.4176180165719, [-374.9557433355867, -442.1383569416215], [-697.4176180165719, -697.4176180165719, -697.4176180165719, -697.4176180165719, -697.4176180165719, -697.4176180165719, -697.4176180165719, -697.4176180165719, -697.4176180165719, -697.4176180165719  …  -697.4176180165719, -697.4176180165719, -697.4176180165719, -697.4176180165719, -697.4176180165719, -697.4176180165719, -697.4176180165719, -697.4176180165719, -697.4176180165719, -697.4176180165719], [-16.009625441133302, 35.76960222320823, 3.451375572598847, -4.394805759450176, 1.9611612825988916, -12.768015722671107, -4.349943973303746, 1.8054734894143851, -15.764141730706157, -11.131710775399712  …  -86.48023019120947, -86.48023019120949, -86.48023019120949, -86.48023019120949, -86.48023019120947, -86.48023019120947, -86.48023019120947, -86.48023019120944, -86.48023019120949, -86.48023019120944])