How can I use optim.jl to find some values

Specific Domains Optimization (Mathematical)
1

I have some working code that i cant optmization to find some x[1] x[2] x[3] x[4] constants.
A link for all the code is:

https://drive.google.com/drive/folders/11cqtyUnWAoAYWdWyWOd6fbu4TVejbBlM?usp=sharing

aRe = [1.28; 1.47; 1.67; 1.95; 2.26; 2.38; 2.68; 2.92; 3.16; 3.77; 4.47]
aΩ11e = [2.52; 2.26; 2.07; 1.92; 1.90; 2.08; 2.31; 2.53; 2.68; 2.51; 2.27]

function dΩ11(x,g,sigma)
R = 0.388/sigma
T=0.119
α = x[3]
ϕ = 0.9*(1-1/(1+(R/x[1])^x[2]))

θ(g) = ϕ*exp(-x[3]*g^2/T)
γ = x[4]


g^5*θ(g)*Qb(g,sigma,α)*exp(-g^2/T) + g^5*(1-θ(g))*γ*Qa(g,sigma)*exp(-g^2/T) +
g^5*(1-θ(g))*(1-γ)*Qc(g,sigma)*exp(-g^2/T)

end

function Ω11(x,sigma)

1/T^3*quadgk(g -> dΩ11(x,g,sigma) ,0.01 , 1.51, atol=1e-3)[1]

end

function X2(x)
soma = 0
for (i,sigma) in enumerate(aσ)

    soma = soma + (Ω11(x,sigma)-aΩ11e[i])^2
end
soma

end

lower = [1.0; 0 ; 0.0 ; 0.7 ]
upper = [5.0; 30.0; 1.0 ; 1.0]
initial_x = [3; 20; 0.5 ; 1.0 ]
opt = optimize(x->X2(x), initial_x, lower, upper,NelderMead())
x = Optim.minimizer( opt )
Optim.minimum( opt )

Is there any error in the code ?

2

Please read this post. That will facilitate everyone to help you.

3 Likes
3

The template that I use for this kind of request, and for bug reports, is:

  1. What I did

  2. What I expected to happen when I did that

  3. What actually happened

You’ve told us 1. To help you, we need to know 2 and 3 as well. Did Optim throw an error? Does this function have a known minimum that Optim didn’t find? Or is there some other reason that you suspect there is an error in the code?

3 Likes
4

I have a range for alpha, but i use lower and upper values in that range, but optim do not respect it …
i will upgrade the code

5

Oh right. I think you’re trying to optimise a multivariate function, but using the syntax for a univariate one. Here’s the relevant part of the manual. I haven’t used Fminbox myself, and I don’t know if it works with Nelder-Mead.

6

yes, it works … but i do not know what happening because the lower and upper limits does not be respect by the program

7

try to run and see it

8

1
1755×526 76.6 KB

9

Please post code (complete and working that people can copy and paste), not screenshots.

10

Hello man … i cannot post my entire code because needs a .dat files …
a link from google drive in the beggining have my entire code.
i use interpolate, so i have .dat files

can you know how to solve it ?

11

I am just guessing but it seems that you did not use the correct syntax for Fminbox, which should be
Fminbox(GradientDescent()), not Fminbox{GradientDescent}()

12

even if i take off the FminBox, leave only the lower and upper, the algoritm fails to respect the limits

13

what are this error ?

image
image1396×445 35.1 KB

14

It looks like you are confusing the argument order in your optimize call - the example given in the docs is

inner_optimizer = GradientDescent()
results = optimize(f, g!, lower, upper, initial_x, Fminbox(inner_optimizer))

(note the order of lower, upper, and initial in comparison to your call)

15

two questions :

  1. why cannot use NelderMead method ?

  2. in my x[3] value i cannot put 0.0 , but my range include the 0.0 value … why the optim do not respect that?

image
image1363×336 25.6 KB

16

As mentioned by David above it’s very hard to help you without an MWE, please consider the advice in Please read: make it easier to help you - #81

1 Like