Algorithm parameter of solve ignored

Using Problem-based linear solver linprog. Trying to pass Algorithm interior-point. Algorithm used is dual-simplex.

x1 = optimvar('x1');  % 
x2 = optimvar('x2');  % 
x3 = optimvar('x3');  % 
x4 = optimvar('x4');  % 
x5 = optimvar('x5');  % 

prob = optimproblem('Objective',10*x1 + 2*x2 + 4*x3 + 8*x4 - x5,'ObjectiveSense','min');  
options = optimoptions('linprog','Algorithm','interior-point');
% Constraints 
prob.Constraints.cons1 =    x1 + 4*x2 - x3            >= 16;
prob.Constraints.cons2 =  2*x1 +   x2       + x3      >= 4;
prob.Constraints.cons3 =  3*x1 +              x4 + x5 >= 8;
prob.Constraints.cons4 =    x1 +            2*x4 - x5 >= 20;
prob.Constraints.cons5 =  x1 >= 0;
prob.Constraints.cons6 =  x2 >= 0;
prob.Constraints.cons7 =  x3 >= 0;
prob.Constraints.cons8 =  x4 >= 0;
prob.Constraints.cons9 =  x5 >= 0;

prob.optimoptions(options);
problem = prob2struct(prob);

[sol, fval, exitflag, output] = solve(prob)

Optimal solution found.
sol = 
    x1: 0
    x2: 4.0000
    x3: 0
    x4: 10.0000
    x5: 0

fval = 88
exitflag = OptimalSolution
output = 
         iterations: 7
    constrviolation: 3.5527e-15
            message: 'Optimal solution found.'
          algorithm: 'dual-simplex'
      firstorderopt: 1.4211e-14
             solver: 'linprog'

Answer 1

[Alan Weiss]

You are passing the options incorrectly. See this example on the solve function reference page.

Alan Weiss

MATLAB mathematical toolbox documentation

Answer 2

[Jon Bebeau]

Alan, I tried all that... Using the example in the documentation: I get an Error:

options = optimoptions('intlinprog','Display','off');
problem = prob2struct(prob);

sol = solve(prob,'Options',options)

I get an Error:

Error using optim.problemdef.OptimizationProblem/solve
Options is not a valid solver. Use 'linprog' or 'intlinprog' instead.

In my output the above error, both lines, are in red but MATLAB won't cut and paste the error message.

[Steven Lord]

The version of the example in the documentation for release R2017b has a slightly different call to solve.

sol = solve(prob,options)

The Optimization Toolbox Release Notes state that the syntax for solve has changed since R2017b. The page to which Alan linked is for the most recent release (currently R2018b) where the "solve(prob, options)" syntax has been removed.

[Alan Weiss]

You are undoubtedly using R2017b. The syntax for passing options to solve changed in R2018a. You should use

sol = solve(prob,options)

and remove the statement prob.optimoptions(options);.

I am a little confuse why you have a prob2struct call, but that is not directly relevant.

Alan Weiss

MATLAB mathematical toolbox documentation

[Jon Bebeau]

Alan, The prob2struct is left over from my many, many attempts to find a solution. You are right using the solve function the prob2struct call is superfluous.

Steve Lord has already commented with the proper syntax for R2017b and this is the version I have.

[Jon Bebeau]

Steve, I am using R2017b. Your suggestion fixed the problem. THANKS.