Too many output arguments while solving a constrained optimization problem.

Hello,

I'm trying to solve a optimization problem with a couple of non-linear constraints. When solving the problem I get the following error:

Error using writeCompiledFun2VirtualFile

Too many output arguments.

I tried different constraints and functions and I noticed that the error appears whenever I use multiplication and/or division with two optimvars. This might be due ot optimvar only supporting ./ and .* . However, replacing / and * with ./ and .* did not solve the problem.

You can find my code attached.

optimisation

  OptimizationProblem : 

	Solve for:
       K_eq, m_eq, x_max

	minimize :
       -(((1.822 - (K_eq .* x_max)) ./ m_eq) .* x_max).^0.5


	variable bounds:
       0 <= x_max <= 0.0064


Solving problem using fmincon.

Error using optim.problemdef.OptimizationProblem/solve
SOLVE requires a non-empty initial point structure to solve a nonlinear problem.

Error in optimisation (line 103)
prob.solve()

Thanks in advance

[Cris LaPierre]

I modified your post to run your m file. It returns a different error. It's the same error I get if I run your m-file in R2019b.

[Michael Christian Linder]

Hmm I just downloaded my code and run it again. I still get the "too many output vairables" error. I then run the code on a virtual machine (with R2018a) I have access too which resulted in the error:

Error using optim.internal.problemdef.Times.getTimesOperator

At least one argument must be numeric.

Error in *

Error in optimisation (line 37)

K_eq == J_p / (l_l2*l_l2) + (l_l1/l_l2)*(l_l1/l_l2) * J_m,

This error makes more sens. But replacing * with .* doesn't solve the problem (same error)

Answer 1

[Matt J]

You are reporting multiple different errors in multiple different places from multiple different Matlab versions, so it is a little hard to parse your question. Basically, though, in earlier version of Matlab, the problem-based framework was limited in scope to linear programming. Non-linear OptimizationVariables expressions were simply not supported. As I demonstrated in my other answer, the code runs fine on more recent Matlab versions, once you supply an initial guess. If you cannot upgrade to a more recent version, however, you will have to resort to the Optimization Toolbox's solver-based optimization tools.

[Michael Christian Linder]

Thanks. I wasn't aware of those differences between the versions. Updating to 2021a solved the problem.

Answer 2

[Matt J]

I get no problem in R2020b once the last line is modified to,

prob.solve(x0)

[Matt J]

The Matlab Online engine doesn't seem to have any problem either (R2021a).

optimisation

  OptimizationProblem : 

	Solve for:
       K_eq, m_eq, x_max

	minimize :
       -(((1.822 - (K_eq .* x_max)) ./ m_eq) .* x_max).^0.5


	variable bounds:
       0 <= x_max <= 0.0064


Solving problem using fmincon.

Local minimum possible. Constraints satisfied.

fmincon stopped because the size of the current step is less than
the value of the step size tolerance and constraints are 
satisfied to within the value of the constraint tolerance.

ans = struct with fields:
     K_eq: 4.0684e-06
     m_eq: 1.5679e-11
    x_max: 4.1737e-04